Description:TECHNICAL FIELD
The present disclosure relates generally to the field of animal husbandry and, more specifically, to an electronic system and a method for efficiently estimating methane emissions from ruminants from an enteric fermentation.
BACKGROUND
Methane emissions from ruminants, particularly those resulting from enteric fermentation, act as a significant source of greenhouse gases (GHGs) in the livestock sector. The methane emissions have a potent effect on global warming due to methane's high Global Warming Potential (GWP). Addressing the issue of the methane emissions is crucial not only for environmental management but also for ensuring the long-term sustainability of farming communities worldwide.
Currently, estimating the methane emissions relies on sophisticated and expensive techniques, such as respiration chambers, the SF6 tracer method, remote sensing, an infrared thermography, an isotopic analysis, portable gas sensors, and mask methods. While effective, these conventional methods present challenges due to their complexity and the need for specialized expertise to operate and interpret the results. For instance, the respiration chambers allow for precise measurement of methane output from individual ruminant animals, but they are limited in sample size and may not replicate real-world conditions accurately.
Furthermore, the high costs associated with equipment procurement and maintenance make widespread adoption of these systems and methods difficult, especially for small-scale farmers and researchers in resource-constrained settings. The Intergovernmental Panel on Climate Change (IPCC) Tier 1 approach, which provides country-wise and region-wise emission factors, does not account for the varying characteristics and feed combinations of individual cattle, further limiting its applicability.
As the agricultural sector is increasingly recognized as both a contributor to and a potential solution for mitigating environmental challenges, there is a growing need for innovative and more accessible approaches to accurately estimate the methane emissions from the ruminants through the enteric fermentation. Alternative methods that are less complex, more cost-effective, and require less specialized expertise may facilitate wider adoption and enable more effective management of the methane emissions from livestock operations, particularly in resource-constrained settings.
Therefore, in light of the foregoing discussion, there exists a need to overcome the aforementioned drawbacks associated with the conventional systems and the conventional methods applied for the methane emission estimation by the ruminants.
SUMMARY
The present disclosure provides an electronic system and a method implemented in the electronic system for efficiently estimating methane emissions from ruminants from an enteric fermentation. The present disclosure provides a solution to the existing problem of low adoption, infeasibility, and complexity of existing systems applied for estimating the methane emissions from the ruminants from the enteric fermentation, and how to technically reduce the high cost associated with equipment procurement and maintenance of these systems and increase adoptability of the methods. An aim of the present disclosure is to provide a solution that overcomes at least partially the problems encountered in the prior art and provide an improved electronic system applied for the methane emission estimation from the ruminants and an improved method implemented in the electronic system for the methane emission estimation from the ruminants from the enteric fermentation. The electronic system of the present disclosure provides a new way of estimating the methane emission from the ruminants by monitoring and management of the feed given to the ruminants. The system uses a simplified approach that is accurate, reliable, affordable, and easy to use. The simplified approach leverages machine learning models trained on the ruminant datasets in a unique manner to significantly reduce the complexity of the methane emission estimation compared to conventional systems used for estimating the methane emissions from the ruminant from the enteric fermentation.
One or more objectives of the present disclosure is achieved by the solutions provided in the enclosed independent claims. Advantageous implementations of the present disclosure are further defined in the dependent claims.
In one aspect, the present disclosure provides an electronic system for estimating methane emissions from ruminants. The electronic system comprising a hardware controller configured to receive input data related to a first ruminant animal from a user, via a first-user interface. The input data is stored in a master database and includes animal characteristics and feed data comprising types and quantities of feed materials consumed by the first ruminant animal. The hardware controller is further configured to access a nutritional database comprising nutritional data for the feed materials. The hardware controller is further configured to process the input data using the nutritional data from the nutritional database to derive feature data representing the overall nutritional content of the feed materials consumed by the first ruminant animal. The hardware controller is further configured to estimate methane emissions from the first ruminant animal based on the derived feature data using a machine learning model trained on ruminant data. The machine learning model outputs a personalized estimate of methane emissions per day for the first ruminant animal. The hardware controller is further configured to generate an alert based on the estimated methane emission from the first ruminant animal. The alert is indicative of an intervention required to reduce the estimated methane emissions if the estimated methane emission exceeds a predetermined threshold value.
The electronic system provides personalized estimates of the methane emissions per day for each individual ruminant animal, considering a breed of the ruminant animals, characteristics, and specific feed consumption. This level of granularity is challenging to achieve with conventional electronic systems. Further, the seamless integration of the electronic system on data on the ruminant animal characteristics, feed composition, and breed-specific emission factors associated with each individual ruminant enables a comprehensive analysis that accounts for multiple variables influencing the methane emissions. A user-friendly interface accessible through websites or mobile devices makes the adoption of the electronic system easy for farmers. The easy adoption makes a scalable solution for the methane emission management. The electronic system serves as a decision-support tool, aiding farmers in selecting appropriate feed combinations and management practices that may potentially reduce and manage the methane emissions from their ruminant animals. The electronic system allows farmers to input data on a regular basis, which helps monitor methane emissions in real-time. Further, real-time monitoring allows for timely interventions and adjustments to mitigate the environmental impact of the methane emissions. The automated nature of the electronic system, as it works on machine learning models, ensures efficient and rapid processing of the input data. The rapid processing of the input data helps in the estimation of methane emission without the labour-intensive processes associated with conventional methods. The machine learning capabilities of the electronic system allow for continuous improvement and refinement of the methane emission estimation model as more data is collected, potentially enhancing its accuracy over time.
In another aspect, the present disclosure provides the method implemented in an electronic system for estimating the methane emissions from the ruminants. The method includes receiving, by a hardware controller, input data related to a first ruminant animal. The input data is stored in a master database and comprising animal characteristics and feed data comprising types and quantities of feed materials consumed by the first ruminant animal. The method further includes accessing, by the hardware controller, a nutritional database comprising nutritional data for the feed materials. The method further includes processing, by the hardware controller, the input data using the nutritional data from the nutritional database to derive feature data representing the overall nutritional content of the feed materials consumed by the first ruminant animal. The method further includes estimating, by the hardware controller, methane emissions from the first ruminant animal based on the derived feature data using a machine learning model trained on ruminant data. The machine learning model outputs a personalized estimate of methane emissions per day for the first ruminant animal. The method further includes generating, by the hardware controller, an alert based on the estimated methane emission from the first ruminant animal. The alert is indicative of an intervention required to reduce the estimated methane emissions if the estimated methane emission exceeds a predetermined threshold value.
The method achieves all the advantages and technical effects of the electronic system of the present disclosure.
It is to be appreciated that all the aforementioned implementation forms can be combined. All steps which are performed by the various entities described in the present application as well as the functionalities described to be performed by the various entities are intended to mean that the respective entity is adapted to or configured to perform the respective steps and functionalities. It will be appreciated that features of the present disclosure are susceptible to being combined in various combinations without departing from the scope of the present disclosure as defined by the appended claims. Additional aspects, advantages, features and objects of the present disclosure would be made apparent from the drawings and the detailed description of the illustrative implementations construed in conjunction with the appended claims that follow.
BRIEF DESCRIPTION OF THE DRAWINGS
The summary above, as well as the following detailed description of illustrative embodiments, is better understood when read in conjunction with the appended drawings. For the purpose of illustrating the present disclosure, exemplary constructions of the disclosure are shown in the drawings. However, the present disclosure is not limited to specific methods and instrumentalities disclosed herein. Moreover, those skilled in the art will understand that the drawings are not to scale. Wherever possible, like elements have been indicated by identical numbers.
Embodiments of the present disclosure will now be described, by way of example only, with reference to the following diagrams wherein:
FIG. 1 is a block diagram illustrating an electronic system for efficiently estimating methane emissions from ruminants, in accordance with an embodiment of the present disclosure;
FIG. 2 is a flowchart of a series of operations that illustrates a process of training and working of the electronic system for the methane estimation from the ruminants, in accordance with an embodiment of the present disclosure; and
FIG. 3 is a flowchart of a method for efficiently estimating the methane emissions from the ruminants, in accordance with an embodiment of the present disclosure.
DETAILED DESCRIPTION OF EMBODIMENTS
The following detailed description illustrates embodiments of the present disclosure and ways in which they can be implemented. Although some modes of carrying out the present disclosure have been disclosed, those skilled in the art would recognize that other embodiments for carrying out or practicing the present disclosure are also possible.
As used throughout this disclosure, the word “may” is used in a permissive sense (i.e., meaning having the potential to), rather than the mandatory sense (i.e., meaning must). Similarly, the words “include”, “including”, and “includes” mean including but not limited to.
The phrases “at least one”, “one or more”, and “and/or” are open-ended expressions that are both conjunctive and disjunctive in operation. For example, each of the expressions “at least one of A, B and C”, “at least one of A, B, or C”, “one or more of A, B, and C”, “one or more of A, B, or C” and “A, B, and/or C” means A alone, B alone, C alone, A and B together, A and C together, B and C together, or A, B and C together.
The term “a” or “an” entity refers to one or more of that entity. As such, the terms “a” (or “an”), “one or more” and “at least one” can be used interchangeably herein. It is also to be noted that the terms “comprising”, “including”, and “having” can be used interchangeably.
The term “automatic” and variations thereof, as used herein, refers to any process or operation done without material human input when the process or operation is performed. However, a process or operation can be automatic, even though performance of the process or operation uses material or immaterial human input, if the input is received before performance of the process or operation. Human input is deemed to be material if such input influences how the process or operation will be performed. Human input that consents to the performance of the process or operation is not deemed to be “material”.
The present subject matter may have a variety of modifications and may be embodied in a variety of forms, and specific embodiments will be described in more detail with reference to the drawings. It should be understood, however, that the embodiments of the present subject matter are not intended to be limited to the specific forms, but include all modifications, equivalents, and alternatives falling within the spirit and scope of the present subject matter.
FIG. 1 is a block diagram illustrating an electronic system for efficiently estimating methane emissions from ruminants, in accordance with an embodiment of the present disclosure. With reference to FIG. 1, there is shown a block diagram that includes an electronic system 100 (herein after referred to as the system 100). The system 100 includes a hardware controller 104 and a memory 106. In an implementation, the system 100 further includes a machine learning model 110. The hardware controller 104 is communicatively coupled with the memory 106 and the machine learning model 110. The system 100 may be used to estimate the methane emissions from ruminant animals based on cattle characteristics and a feed material provided to them. Specifically, the system 100 generates tailored nutrition recommendation for the ruminant animal to reduce the methane emission. In some examples, the system 100 may be used for cattle farming or other animal management or farming scenarios. Some other farming scenarios may include, but not limited to, poultry farming and aqua farming.
In an implementation, the hardware controller 104, the memory 106, and the machine learning model 110 may be implemented on a same server, such as a server 102. In some implementations, the system 100 further includes a storage device 112. In some implementations, the storage device 112 is configured to store a master database 114 and a nutritional database 116. In some implementations, the master database 114 may be stored in the same server, such as the server 102.
In some implementations, the storage device 112 is communicatively coupled to the server 102, via a communication network 118. The server 102 may be communicatively coupled to a plurality of user devices, such as a user device 120, via the communication network 118. A mobile application or a web platform may be accessible from the user device 120. The user device 120 includes a first user interface 122 and a second user interface 124. The second user interface 124 further includes an alert 126. The user device 120 may be configured to receive input data 108 related to a first ruminant animal from a user, via the first user interface 122.
The present disclosure provides the system 100 designed to estimate methane emissions from ruminant animals. The system 100 uses the hardware controller 104 to process the input data 108 about the ruminant animal and its feed, extracting nutritional features and accessing databases with relevant information. The machine learning model 110 then calculates the methane emissions based on these features. If the methane emissions exceed a threshold, alerts prompt interventions. Additional features include forming a methane emission database, suggesting alternative feed options, interfacing with carbon credit devices, and visualizing emissions over time. The system 100 may further train and periodically update the machine learning model 110 for improved accuracy, ensuring precision in estimating the methane emissions and aiding environmental sustainability and animal welfare practices. The synergy between the hardware controller 104, databases (i.e. the master database 114 and the nutritional database 116), and the machine learning model 110 enables the system 100 to process and analyze the input data 108 effectively, providing valuable insights for farmers and dairy and beef producing companies to manage and reduce the methane emissions from their cattle.
The server 102 includes suitable logic, circuitry, interfaces, and code that may be configured to communicate with the user device 120 via the communication network 118. In an implementation, the server 102 may be a master server or a master machine that is a part of a data center that controls an array of other cloud servers communicatively coupled to it for load balancing, running customized applications, and efficient data management. Examples of the server 102 may include, but are not limited to a cloud server, an application server, a data server, or an electronic data processing device.
The hardware controller 104 refers to a computational element that is operable to respond to and processes instructions that drive the system 100. The hardware controller 104 may refer to one or more individual processors, processing devices, and various elements associated with a processing device that may be shared by other processing devices. Additionally, the one or more individual processors, processing devices, and elements are arranged in various architectures for responding to and processing the instructions that drive the system 100. In some implementations, the hardware controller 104 may be an independent unit and may be located outside the server 102 of the system 100. Examples of the hardware controller 104 may include but are not limited to, a hardware processor, a digital signal processor (DSP), a microprocessor, a microcontroller, a complex instruction set computing (CISC) processor, an application-specific integrated circuit (ASIC) processor, a reduced instruction set (RISC) processor, a very long instruction word (VLIW) processor, a state machine, a data processing unit, a graphics processing unit (GPU), and other processors or control circuitry.
The memory 106 refers to a volatile or persistent medium, such as an electrical circuit, magnetic disk, virtual memory, or optical disk, in which a computer may store data or software for any duration. Optionally, the memory 106 is a non-volatile mass storage, such as a physical storage media. Examples of implementation of the memory 106 may include, but are not limited to, an Electrically Erasable Programmable Read-Only Memory (EEPROM), Dynamic Random-Access Memory (DRAM), Random Access Memory (RAM), Read-Only Memory (ROM), Hard Disk Drive (HDD), Flash memory, a Secure Digital (SD) card, Solid-State Drive (SSD), and/or CPU cache memory.
The input data 108 refers to a set of details specific to the ruminant animal which are essential for accurate estimation of the methane emissions. The input data 108 comprises animal characteristics and feed data comprising types and quantities of feed materials consumed by the ruminant animals (such as the first ruminant animal). The animal characteristics may include species of the ruminant animals (such as cattle, sheep, or goats), the breed or genetic lineage of the ruminant animals, country of origin of the ruminant animals, its current production stage (milking or dry), last calving measured in days, the lactation cycle measured in numbers, live weight in kilograms and the daily milk production measured in litres per day. The breed may influence factors like size, metabolism, and methane production. The metabolism governs how efficiently animals digest their food. In the case of ruminant livestock such as cattle, digestion occurs in multiple stages within the rumen, where microbes break down fibrous plant material. The efficiency of the digestion process may vary between breeds due to differences in metabolic rates and digestive physiology.
The feed data includes information about the types of feed materials, the quantity consumed by the ruminant animal, and their nutritional content. The feed materials may include dry fodder (for example: hay), green fodder (for example: grass), silage, grains, concentrates, plant byproducts, salts, minerals and supplements, with amounts measured in kilograms per day. Further, the nutritional composition of each feed material, including protein, fiber, carbohydrates, fats, vitamins, and minerals, is also considered. Additionally, the feeding schedule, which indicates the frequency and timing of feedings throughout the day, is considered, with variations based on factors such as the ruminant animal's age, lactation cycle, last calving and feeding management practices.
The machine learning model 110 refers to a software component or an integrated circuitry comprising an AI (artificial intelligence) engine that incorporates an artificial intelligence model, designed to estimate the methane emissions from the ruminant animal. The machine learning model 110 is trained on ruminant data, including animal characteristics, feed data, and derived and recorded methane emissions. The training process involves preprocessing the ruminant data to handle missing values and outliers, deriving feature data, and training the machine learning model 110 to map the derived features to recorded methane emissions.
In some implementations, the machine learning model 110 may utilize a linear multiple regression operation to perform the methane emission estimation. The machine learning model 110 processes the input data 108, which includes information on the animal characteristics, the feed composition, and the nutritional content, to generate personalized estimates of methane emissions per day for individual ruminant animals.
The storage device 112 may be any storage device that stores data and applications without any limitation thereto. In an implementation, the storage device 112 may be a cloud storage, or an array of storage devices. The master database 114 and the nutritional database 116 within the system 100 serve as a repository encompassing distinct datasets for the functionality of system 100.
The master database 114 is the primary repository for all data collected directly from farmers. The master database 114 serves as the central hub for raw data related to cattle characteristics, feed information, and other relevant details. The master database 114 undergoes various data processing steps to ensure cleanliness and organization before training the machine learning model 110. Additionally, the estimated methane conversion factor (Ym) and methane emission are stored in the master database 114 for future reference and analysis. Essentially, the master database 114 forms the foundation for subsequent data processing, modelling of the machine learning model 110 and decision-making processes within the system 100.
The nutritional database 116 includes a nutritional data for the feed materials. The nutritional data comprises details regarding dry matter, Crude Protein (CP), Crude Fiber (CF), Neutral Detergent Fiber (NDF), Acid Detergent Fiber (ADF), Ether Extract (EE), Ash content, lignin, Gross Energy (GE), Metabolizable Energy (ME), Total Dry Matter Intake Weight (DMI) of the feed materials. Further, the dry matter represents everything contained in a sample of feed materials except water. The dry matter includes protein, fiber, fat, minerals, and the like. In practice for measurement of the nutritional data, a sample of feed materials is taken. The dry matter is total weight of the sample of feed materials minus the weight of water in the sample of feed materials, expressed as a percentage.
The CP is a measurement of the total protein content in the sample of feed materials and is expressed as the percentage of the dry matter. The CP includes both true protein and non-protein nitrogen compounds. The CF represents the portion of the sample of feed materials that consists of cellulose, hemicellulose, and lignin. The CF is measured as a percentage of the dry matter and provides an indication of the fiber content in the feed materials. The NDF is a fraction of the cell wall components in the sample of feed materials, including hemicellulose, cellulose, and lignin. The NDF is expressed as a percentage of the dry matter and reflects the structural fiber content of the sample of feed materials. The ADF represents the cellulose and lignin content of the sample of feed materials and is expressed as a percentage of the dry matter. The ADF provides a measure of the indigestible fiber content in the sample of feed materials. The EE also known as crude fat, refers to the fat content present in the sample of feed materials. The EE is extracted using diethyl ether or petroleum ether and is expressed as a percentage of the dry matter.
The GE represents the total energy content of the feed material and is measured as the heat released when the feed undergoes complete combustion. In an example, the GE may be expressed in megajoules per kilogram of dry matter. The ME is the energy available to the ruminant animal for maintenance, growth, and production purposes after accounting for energy losses in faeces, urine, and gases. In an example, the ME may be expressed as megajoules per kilogram of the dry matter. The ME is one of the main indicators of the energy value of the feed materials.
The DMI refers to the combined weight of all the feed materials consumed by cattle after removing the moisture content. The measurement accounts for the solid components of the feed, such as fiber, proteins, and carbohydrates, and is calculated by summing up the dry matter weights of each feed material fed to the cattle.
Further, the nutritional data for the feed materials may be obtained using some credible sources already available. For example, nutritional data from credible sources involves accessing information that has been thoroughly researched, validated, and published by experts in the field of the ruminant animal nutrition and agriculture. Research papers, academic journals, and scientific publications are primary sources where findings from studies and experiments related to the nutritional data for the feed materials are documented. Additionally, databases maintained by reputable institutions dedicated to animal nutrition and agriculture serve as valuable repositories of comprehensive nutritional data for the feed materials. The databases aggregate information from multiple sources, providing a centralized and reliable resource for accessing nutritional data content in the feed materials.
The methane production in the rumen is driven by microbial fermentation. The microbial populations produce methane as a byproduct of their metabolic processes during the breakdown of feed components. Breeds with different metabolic rates may harbour distinct microbial communities in their rumen, leading to variations in the methane production. For example, higher metabolic rates often correspond to increased energy requirements, which may cause the ruminants to consume larger quantities of feed. With more feed entering the rumen, there is a greater substrate available for microbial fermentation. This, in turn, may lead to elevated methane production. The ruminants with higher metabolic rates may exhibit increased physical activity. This elevated activity level may influence the rate of feed passage through the digestive system and the extent of rumen fermentation. More physical activity may result in a more efficient mixing of rumen contents, enhancing microbial fermentation and the methane production.
The communication network 118 includes a medium (e.g., a communication channel) through which the user device 120 communicates with the server 102. The communication network 118 may be a wired or wireless communication network. Examples of the communication network 118 may include, but are not limited to, a Local Area Network (LAN), a wireless personal area network (WPAN), a Wireless Local Area Network (WLAN), a wireless wide area network (WWAN), a cloud network, a Long-Term Evolution (LTE) network, a plain old telephone service (POTS), a Metropolitan Area Network (MAN), and/or the Internet.
The user device 120 refers to an electronic computing device operated by a user. In some implementations, the user device 120 may be configured to receive the input data 108 related to a first ruminant animal from a user, via a first user interface 122. Examples of the user device 120 may include, but not limited to, a mobile device, a smartphone, a desktop computer, a laptop computer, a Chromebook, a tablet computer, a robotic device, or other user devices.
The first user interface 122 is an interface through which the user interacts with the system 100 to provide the input data related to the first ruminant animal. The first user interface 122 includes elements such as forms, fields, buttons, and menus that allow the user to input information about the first ruminant animal's characteristics and the feed data. The first user interface 122 may be accessed via the user device 120.
The second user interface 124 refers to a computing device utilized for displaying visualizations of the estimated methane emissions over a period of time for the first ruminant animal. The user device 120 is connected to the hardware controller 104 and allows users to interact with the output given by the system 100. The second user interface 124 may include devices such as a computer monitor, tablet, smartphone, or any other display device capable of presenting visual information to the user. The second user interface 124 enables users to view and analyze the estimated methane emission data conveniently, facilitating informed decision-making regarding interventions to reduce the methane emissions from the ruminant animals.
In operation, the hardware controller 104 is configured to receive the input data 108 related to the first ruminant animal from the user, via the first user interface 122. The input data 108 is stored in the master database 114 and comprising the animal characteristics and the feed data comprising types and quantities of the feed materials consumed by the first ruminant animal. The hardware controller 104 operates by enabling user interaction through the first user interface 122. The first user interface 122 allows easy customization of feed combinations, making the input data 108 process accessible to individuals with varying levels of technical knowledge. The hardware controller 104 serves as a central component of the system 100, facilitating the reception of the input data 108 pertaining to the first ruminant animal from users, for example, the farmer via an application installed in their smartphones. The input data 108 is stored in the master database 114 for further processing.
For example, consider a dairy farmer who aims to estimate the methane emissions of his Holstein cow. The farmer accesses the first user interface 122 of the system 100 on his smartphone, where he encounters a series of fields prompting him to fill out information about his cow’s characteristics and the feed data. Through the first user interface 122, the farmer enters specific details such as the country where his cow is situated , the species of his ruminant (e.g., cow), and its precise breed (e.g., registered Holstein). The farmer also selects the current production stage of his cow, specifying whether it is in a milking phase or undergoing a dry period. Further, the farmer provides information about the feed materials consumed by his cow, last calving, lactation cycle, milk production, and live weight of the cow. Through this interactive process facilitated by the first user interface 122 of the system 100, the farmer effectively communicates input data 108 about his Holstein cow to the system 100.
The hardware controller 104 is further configured to access a nutritional database 116 comprising nutritional data for the feed materials. In some implementations, the system 100 estimates methane emissions from ruminants by accessing the nutritional database 116 that contains the nutritional data for the feed materials. The system 100 processes the input data 108 using these nutritional contents in the backend in the form of the nutritional database 116. By accessing the nutritional database 116 of nutritional data for feed materials, the hardware controller 104 may consider the specific attributes of each feed material that contribute to the methane emissions. The selection of specific attributes by the hardware controller 104 allows for a more precise estimation of the methane emissions based on the nutritional composition of the feed provided to the ruminants.
For example, consider two different feed materials commonly fed to the ruminant animals, alfalfa and corn silage. The alfalfa has a higher dry matter and the CP compared to corn silage. For example, the alfalfa has a dry matter of 90% and the CP of 20%, while corn silage has a dry matter of 70% and the CP of 8%. When accessing the nutritional database 116, the system 100 considers these nutritional data of the feed materials. For example, as the alfalfa has a higher CP than corn silage, the alfalfa may result in higher methane emissions due to increased nitrogen content, which may lead to greater microbial activity in the rumen of the ruminants. Similarly, the system 100 considers the dry matter of the feed materials. Higher dry matter content in alfalfa means less moisture, resulting in a higher concentration of nutrients per unit weight. This can affect methane emissions as well. Now, if a ruminant animal is primarily fed alfalfa, the system 100 may predict higher methane emissions compared to an animal primarily fed corn silage, reflecting the effect of differences in nutritional composition between the two feed materials on the methane emission estimation.
The hardware controller 104 is further configured to process the input data 108 using the nutritional data from the nutritional database 116 to derive feature data representing the overall nutritional content of the feed materials consumed by the first ruminant animal. In the process of estimating methane emissions from ruminants, the hardware controller 104 utilizes the data stored in the nutritional database 116, which contains information about the nutritional contents of feed materials given to the ruminants. In an implementation, in order to create the nutritional database 116, the nutritional content data in the feed materials is collected from credible sources and organized into a spreadsheet named “feed data”. The spreadsheet serves as a data table and includes columns representing nutritional compositions.
For example, the feed material like wheat bran. The nutritional composition of the wheat bran, including attributes like dry matter, CP, and CF, is documented in the feed data created from the spreadsheet. When the system 100 receives the input data 108 from the farmer through the first user interface 122, specifying the types and quantities of feed materials consumed by their ruminant, the system 100 matches the names of these feed materials with those in the data table of feed data.
Upon matching the names, the hardware controller 104 is configured to retrieve the corresponding nutritional data from the feed data. The hardware controller 104 then processes the “feed data” to derive the feature data representing the overall nutritional content of the feed materials consumed by the first ruminant. By incorporating the nutritional data of the feed materials into the estimation process, the system 100 enhances its ability to provide precise results.
In an implementation, in order to derive the feature data, the hardware controller 104 is further configured to convert the feed data to dry matter weights using the nutritional data to calculate overall percentages and overall absolute weights of crude protein, crude fiber, neutral detergent fiber, ether extract, and ash for the feed material consumed by the first ruminant animal.
In an exemplary feed mixture, various ingredients like alfalfa, corn silage, and soybean meal may be present with different moisture content. The hardware controller 104 accesses nutritional data that includes the dry matter percentages for each ingredient. For example, the alfalfa may have a dry matter percentage of 90%, the corn silage may have a dry matter percentage of 25%, and soybean meal may have a dry matter percentage 88%. Using this nutritional data, the hardware controller 104 calculates the dry matter weights for each ingredient in the feed mixture. For example, for 100 kg of alfalfa, the dry matter weight will be 90 kg, for 100 kg of corn silage dry matter weight will be 25 kg, and for the soybean meal dry matter weight will be 88 kg.
Further, the hardware controller 104 proceeds to calculate the overall percentages and absolute weights of key nutritional components such as the CP, the CF, the NDF, the EE, the ADF, total gross energy of the feed and ash. After obtaining the nutritional content of each ingredient per kilogram of dry matter, the hardware controller 104 proceeds to calculate the overall percentage and the absolute weights. For alfalfa, the crude protein content is 20%, crude fiber is 25%, neutral detergent fiber is 35%, ether extract is 5%, and ash is 8%. The hardware controller 104 further iterates through each feed mixture, considering the dry matter weights of each ingredient. Further, the hardware controller 104 then calculates the weighted average of each nutritional component across all ingredients in the mixture to obtain the overall percentage.
For example, a feed mixture present in the form of its dry matter weight, this dry matter weight of the feed mixture is used for the calculation of the overall percentage. The dry matter weight of the feed mixture consists of 50% alfalfa, 30% corn silage, and 20% soybean meal. The hardware controller 104 calculates the overall percentage of crude protein by multiplying the crude protein content of each ingredient by its proportion in the feed mixture. As for alfalfa, the crude protein content is 20%, crude fiber is 25%, neutral detergent fiber is 35%, ether extract is 5%, and ash is 8%. For corn silage, crude protein content is 8%, crude fiber content is 3%, neutral detergent fiber content is 30%, ether extract is 1%, and ash is 6%. For soybean meal, crude protein is 48%, Crude fiber is 3%, Neutral detergent fiber is 12%, Ether extract is 1%, and ash is 7%.
To calculate the overall protein content in the feed mixture, the hardware controller 104 is configured to multiply the protein content of each ingredient by its proportion in the feed mixture. For example, if alfalfa constitutes 50% of the mixture and its protein content is 20%, the contribution to overall protein content is 50%×20%=10%. If corn silage constitutes 30% of the mixture and its protein content is 8%, the contribution to overall protein content is 30%×8%=2.4%. If soybean meal constitutes 20% of the mixture and its protein content is 48%, the contribution to overall protein content is 20%×48%=9.6%.
The hardware controller 104 is configured to add the calculated values to obtain the total protein content contributed by the alfalfa, the corn silage and the soybean and finally divides the total protein content by the total dry matter weight of the feed mixture to get the overall protein content of the feed. The above process is repeated for each nutritional component, resulting in the calculation of overall percentages of crude protein, crude fiber, neutral detergent fiber, ether extract, and ash for the feed material consumed by the first ruminant animal.
Further, the hardware controller 104 calculates the absolute weights of each nutritional component across all ingredients in the feed mixture to obtain the overall absolute weights of the feed combination. For example, if the feed mixture contains 10 kg of paddy straw and 20 kg of lucerne. The weight of paddy straw and lucerne is given in terms of “as fed basis” (i.e. with moisture). In order to calculate the absolute weight of CP in the paddy straw and lucerne weights given in “as fed basis” (i.e. with moisture) need to be converted into dry matter weights. As for paddy straw, dry matter (% as fed) is 90%, the CP(% as dry matter) content is 4%. For lucerne, dry matter (% as fed) is 20%, and the CP(% as dry matter) content is 20%. The hardware controller 104 converts weights given in “as fed basis” into dry matter weights by multiplying weight in ‘‘fed as basis’’ by dry matter ( % as fed). The dry matter weight for paddy straw becomes (10 kg)×0.9=9kg. The dry matter weight for lucerne becomes (20 kg)×0.2=4 kg.
To calculate the absolute CP content in the feed mixture, the hardware controller 104 is configured to multiply the dry matter weight of each ingredient by proportion of the CP( in fractions) in the feed component and add all the absolute CP corresponding to different feed components of the feed mixture. For example, paddy straw constitutes the CP(% as dry matter) content of 4%. The absolute weight CP in paddy straw is 9×0.04=0.36. The absolute weight CP in lucerne is 4×0.2=0.8. In order to obtain the absolute CP content in the feed mixture. The absolute weight CP in paddy straw and the absolute weight CP in lucerne are added, i.e. 0.36+ 0.8=1.16 Kgs. The hardware controller 104 is further configured to estimate the methane emissions from the first ruminant animal based on the derived feature data using a machine learning model trained on ruminant data, wherein the machine learning model 110 outputs a personalized estimate of methane emissions per day for the first ruminant animal. The ruminant data in the training dataset includes characteristics of ruminant animals, diet, and derived and recorded methane emissions corresponding to diet. The characteristics of the ruminant animal may include details like species, breed, age, weight, gender, and health status. The diet may include data about the types of feed consumed, their nutritional composition, quantity, and feeding schedule. Further, derived and recorded methane emissions may contain measurements of methane emissions associated with each animal with respect to characteristics and diet, along with the methodology used for measurement of recorded methane emission. By integrating data collection, feature derivation, database referencing, and machine learning modelling, the hardware controller 104 provides a personalized estimate of methane emissions for the individual ruminant animal.
After obtaining the derived feature data, the hardware controller 104 generates methane conversion factor (Ym) using Ym equation. The methane conversion factor (Ym) represents the percentage of gross energy intake by the ruminant animal that is converted into methane. The derivation of the methane emission factor involves a series of steps. The hardware controller 104 gathers necessary inputs. In an example, the necessary inputs include the Total Dry Matter Intake (DMI), Ether Extract (EE) content of the feed, and Neutral Detergent Fiber (NDF) content of the feed. The methane conversion factor (Ym) is calculated using the Ym equation, which is a mathematical formula that incorporates necessary inputs. The methane conversion factor (Ym) is:
{[ 76.0 + (13.5 x DMI) – (9.55 x EE) + (2.24 x NDF)] x 0.05565} /(DMI x 18.45).
The resulting value provides insight into the methane production efficiency of the ruminant animal based on its diet and intake. Higher methane conversion factor (Ym) values indicate a greater proportion of gross energy converted into methane, which may have implications for greenhouse gas emissions and environmental sustainability.
The hardware controller is further configured to estimate methane emissions based on the derived methane conversion factor (Ym). In an example, the IPCC energy equation may be used to get the methane emission in grams per head per day. GE is taken in megajoules per kilogram. Methane emission = {(GE) x (methane conversion factor (Ym)) x 1000}/ 55.65.
In some implementations, for certain feed materials in different regions or countries, if the GE value is not available from credible sources, the GE for the feed material might be derived using equations suitable from credible sources for each feed material. For example, using equations such as GE (Mcal/kg) = [(CP x 0.056) + (EE x 0.094) + (100 - CP – EE - Ash) x 0.042], the GE may be derived with reasonable accuracy using other variables available to us of that particular feed material. The GE may be then subsequently converted to GE (MJ/kg) using suitable factors.
The methane conversion factor (Ym) equation may change based on different countries and based on breeds and also maybe new findings of equations from different credible sources that we can substitute it in. Hence the methane emissions may change accordingly as per the methane conversion factor (Ym). Once the methane conversion factor (Ym) and methane emission are estimated, the methane conversion factor (Ym) and methane emission data are stored back in the master database 114. In an implementation, the hardware controller 104 is further configured to form a methane emission database for the estimated methane emissions from the first ruminant animal. The estimated emission from the first ruminant animal is used as a repository for further training of the machine learning model 110. Training the machine learning model 110 on a methane emission database derived from actual estimates helps reduce bias in the machine learning model's predictions. The reduction in bias ensures that the machine learning model 110 is grounded in real-world data, leading to more reliable and unbiased estimates of methane emissions from ruminant animals.
In some implementations, the research and experimentation may be conducted based on different countries or regions by collecting samples of experimental data, which includes ruminant characteristics (for e.g.: species of ruminant animal, breed of ruminant animal, live weight of ruminant animal, body condition score of ruminant animal, health statistics of ruminant animal, etc), the temperature and humidity readings of the experimental project area under project conditions, the feed materials types and quantities fed to the cattle on a per-day basis. The nutritional profiles of the feed materials fed to the cattle may be obtained through experiments, also, the methane production of the ruminant under consideration may be captured on a per-day basis through the use of equipment. Using the data collected from the conducted experiments, regression equations of the methane produced may be derived for that sample. The derived regression equations of the methane produced are substituted in the methane conversion factor (Ym) equation. The Ym equation may be used to derive the methane production values using the IPCC equation for the data collected from farmers on ruminants or dairy or beef companies of that particular country or region. The master database 114 based on the country and its unique ruminant breeds may be formed based on derived methane production values. Further, the master database 114 may be used to form the machine learning model 110 that estimates the individual methane emission of ruminants on a per-day basis of that country or region for new framer data. In an implementation, the machine learning model 110 comprises a linear multiple regression model trained to map the derived feature data associated with different ruminant breeds. The linear multiple regression is a statistical method utilized to examine the relationship between dependent variables (the methane emissions) and the multiple independent variables such as animal characteristics (e.g., breed, lactation cycle), the feed data (e.g., types and quantities of dry matter weight of feed materials consumed), and the nutritional content of the feed materials (e.g., crude protein, crude fiber, and ether extract) can in ruminant animals.
The incorporation of multiple independent variables allows the linear multiple regression for a more comprehensive analysis of the factors influencing the methane emissions from the ruminant animals. The linear multiple regression models enable the examination of how changes in various nutritional contents (for example, total dry matter intake weight, the CP, the EE, and the NDF), in combination with animal characteristics, affect methane production. This method facilitates the identification of significant predictors and their respective contributions to methane emission levels. Additionally, linear multiple regression provides insights into the strength and direction of these relationships, helping researchers and practitioners better understand and manage methane emissions in livestock production systems.
The hardware controller 104 is further configured to train the machine learning model 110 by obtaining a training dataset comprising the ruminant data, including the animal characteristics, the feed data, and corresponding recorded methane emissions. By incorporating diverse information such as species, breed, age, body weight, feed types, quantities, and nutritional content, along with the methane emission measurements, the machine learning model 110 may identify patterns and correlations that may not be apparent through manual analysis alone.
The hardware controller 104 is further configured to train the machine learning model 110 by preprocessing the training dataset to handle missing data and outliers. The hardware controller 104 scans the training dataset to identify any missing values or null entries in the data. In an example, various techniques, such as mean, median, mode imputation, or advanced imputation methods like K-nearest neighbours (KNN) or interpolation, may be employed to fill in missing values with estimated values. By addressing missing data, the training process of the machine learning model 110 becomes less prone to bias. The addressing of missing data ensures that valuable information is not disregarded, thereby improving the accuracy of the predictions of the machine learning model 110.
The outliers, which are data points significantly different from other observations in the training dataset, are identified using statistical methods or domain knowledge. Outliers can be handled by removing them from the dataset, transforming them, or replacing them with more appropriate values. Managing outliers prevents them from disproportionately influencing the training process and model performance. This leads to a more reliable and generalized model that better represents the underlying data distribution.
The hardware controller 104 in further configured to train the machine learning model 110 by deriving training feature data and the methane emission factors associated with each individual ruminant breeds from the training dataset. By processing the feed data using the nutritional data from the nutritional database 116, the hardware controller 104 may derive features representing the overall nutritional content of the feed materials consumed by each ruminant in the training dataset. These derived features, such as overall percentages and absolute weights of crude protein, crude fiber, neutral detergent fiber, ether extract, and ash, capture the combined nutritional profile of the feed consumed, which may significantly impact the methane emissions.
By incorporating individual cattle-specific methane conversion factors and methane emission based on equations from credible sources taking different variables into the training process allows the machine learning model 110 to account for inherent differences in methane production among various breeds, potentially improving the accuracy of the methane emission estimates. By deriving training feature data representing the overall nutritional content of the feed materials, the machine learning model 110 may effectively learn the relationship between feed composition and the methane emissions, which is crucial for accurate methane emission estimation.
The hardware controller 104 is further configured to train the machine learning model 110 by mapping the derived training feature data associated with different ruminant breeds to the recorded methane emissions. By explicitly training the machine learning model 110 to map the derived nutritional features and each cattle-specific methane conversion factor to the recorded methane emissions, the machine learning model 110 may learn the intricate relationships between the feature data and the methane emissions. The feature mapping process allows the machine learning model 110 to capture the complex interactions and dependencies between feed composition, cattle characteristics, and methane production, leading to more accurate methane emission estimates. The mapping process may be scaled to incorporate additional relevant features or methane emission factors as they become available. For example, if new information about the influence of environmental factors or management practices on the methane emissions is discovered, the machine learning model 110 may be retrained with additional features to further improve its performance. By analyzing the mapping between the input features and the target, insights into the relative importance of different nutritional components and cattle characteristics in influencing the methane emissions may be interpreted. The interpretability may help in future research, feed formulation strategies, and breeding programs aimed at reducing methane emissions.
In an implementation, the machine learning model 110 comprises a linear multiple regression algorithm. The linear multiple regression algorithm includes multiple predictor variables, allowing for a complex analysis that examines how various factors, such as cattle characteristics and feed composition, influence methane emissions simultaneously. This approach provides detailed insights into the relationships between methane emissions and each independent variable while controlling for the effects of other predictors, thus isolating each variable's impact. It enhances predictive power by forecasting methane emissions based on known values of independent variables, and quantifies the influence of each predictor, highlighting the most impactful factors. Multiple regression is efficient, handling different types of variables and functional forms in a single model, and enables modelling of interactions between variables. Additionally, it supports statistical hypothesis testing and provides metrics like R-squared to assess model fit, making it a robust tool for accurate and detailed methane emission estimation and decision-making in livestock management.
The hardware controller 104 is further configured to periodically retrain the machine learning model 110 using continuously collected ruminant data and the estimated methane emissions to improve the accuracy of the machine learning model 110 over time. The retraining process involves taking the existing dataset and incorporating new inputs from the farmers. The new inputs undergo data processing and transformation, resulting in the derived feature. The derived feature passes through the pre-trained machine learning model 110, which predicts personalized methane emission estimates for the ruminant animal. As new data is continuously collected from the ruminants and their estimated methane emissions, it is important to incorporate this information into the machine learning model 110. By periodically retraining, the machine learning model 110 can adapt to changes in the training dataset and capture any evolving patterns between the ruminant data and the methane emissions. Further, by incorporating new data and retraining, the machine learning model 110 becomes more adept at capturing the complex relationships between ruminant characteristics, feed combinations, and methane emissions. The continuous improvement in accuracy also enhances the overall effectiveness and reliability of the system 100.
In some implementations, the derived features facilitate in retraining the machine learning model 110, and a new machine learning model with better metrics will be obtained. The new machine learning model, when subjected to new data from the farmers again which, may be subjected to data processing and transformation in the backend and provide the latest more accurate estimation of methane emissions.
In some implementations, in order to preprocess the training dataset, the hardware controller 104 is further configured to remove unnecessary data columns and rows based on predefined criteria. The removal of unnecessary data columns and rows helps to streamline the training dataset and focus on relevant information for training of the machine learning model 110. This process reduces the complexity and dimensionality of the training dataset, leading to more efficient training of machine learning models. Removing irrelevant columns and rows also helps eliminate noise and redundant information.
In some implementations, in order to preprocess the training dataset, the hardware controller 104 is further configured to handle null values in the animal characteristics and the feed data. The null values refer to the absence of data in a particular field of the training dataset. The null values occur due to various reasons such as data entry errors into the system 100, equipment malfunction, or simply because the information is not available or not collected for certain feed data. Addressing the null values ensures that the training dataset for the machine learning model 110 is complete and accurate. The null values may introduce bias into the analysis if not properly handled. Instead of discarding the training dataset with null values, handling them allows for the retention and utilization of valuable training data.
In some implementations, in order to preprocess the training dataset, the hardware controller 104 is further configured to combine similar breed names into representative categories. By consolidating similar breed names into broader categories, the number of distinct features in the training dataset is reduced, which prevents the machine learning model 110 from becoming overly complex and mitigates the risk of overfitting.
Aggregating similar breeds into representative categories facilitates the machine learning model 110 to learn more general patterns shared across broader breed categories, leading to more accurate predictions. Some rare or less represented breeds may have insufficient data for training the machine learning model 110. Grouping them with more common breeds ensures that these less prevalent categories still contribute to the model's learning process. Also, grouping similar breeds into broader categories may simplify the interpretation of the machine learning model 110 results as it facilitates understanding trends and insights at a higher level, facilitating decision-making processes. In some implementations, in order to preprocess the training dataset, the hardware controller 104 is further configured to remove duplicate data rows. Removal of duplicate rows ensures that the training dataset is free from redundant information, allowing machine learning model 110 to focus on relevant patterns and relationships within the training dataset.
In some implementations, in order to train the machine learning model 110, the hardware controller 104 is further configured to apply data transformations to the derived training feature data to reduce skewness and stabilize variance. The skewness refers to the asymmetry in the distribution of feature data, while variance measures the spread or dispersion of the feature data around the mean. By applying data transformations such as logarithmic transformation, square root transformation, or Box-Cox transformation, the hardware controller 104 adjusts the distribution of the feature data to make it more symmetrical and closer to a normal distribution. The data transformations help in reducing the skewness of the data, which is essential for ensuring that the assumptions of many machine learning models are achieved.
In some implementations, in order to train the machine learning model 110, the hardware controller 104 is further configured to standardize the transformed training feature data. Standardizing the transformed training feature data ensures that all features have a mean of 0 and a standard deviation of 1. Standardization helps to scale and normalize the training data features, making them comparable and preventing any particular feature from dominating the training process. Standardization improves the convergence speed of the machine learning model 110 and prevents numerical instabilities, leading to more stable and reliable training results. In some implementations, training feature data is used without transformation and standardisation.
In some implementations, in order to train the machine learning model 110, the hardware controller 104 is further configured to evaluate one or more candidate machine learning operations using the standardized training feature data and the recorded methane emissions. Evaluating multiple candidate machine learning operations allows for the comparison of different operations and techniques. By assessing the performance of various models using standardized training feature data and recorded methane emissions, the hardware controller 104 may identify the most suitable approach for estimating the methane emissions from the ruminant data. The process ensures that the selected machine learning operation is well-suited to the specific characteristics and patterns present in the training dataset, maximizing prediction accuracy.
In some implementations, in order to train the machine learning model 110, the hardware controller 104 is further configured to select a candidate machine learning operation from the one or more candidate machine learning operations to be utilized in the machine learning model 110 for estimating methane emissions from the ruminant data based on the evaluation. The selection of the operation that demonstrates the performance on the standardized training data, the hardware controller 104 ensures that the final machine learning model 110 is capable of accurately capturing the underlying relationships between the feature data and the methane emissions in ruminant data.
In some implementations, in order to train the machine learning model 110, the hardware controller 104 is further configured to perform hyperparameter tuning on the selected candidate machine learning operation to optimize the performance of the machine learning model 110. Hyperparameter tuning is the process of selecting the optimal values for the parameters that govern the training process of the machine learning model 110. Performing hyperparameter tuning optimizes the configuration of the machine learning model 110, leading to improved performance metrics such as accuracy and generalization. This ensures the model is finely tuned to the training dataset, maximizing its effectiveness in making accurate predictions.
The hardware controller 104 is further configured to generate an alert 126 based on the estimated methane emission from the first ruminant animal. The alert 126 is indicative of an intervention required to reduce the estimated methane emissions if the estimated methane emission exceeds a predetermined threshold value. The system 100 has a predetermined threshold value for methane emissions, which is set based on emission targets, regulatory requirements, or desired levels of environmental sustainability. The predefined threshold may vary depending on the specific region, industry, or farm operations. After estimating the methane emissions for the first ruminant animal, the hardware controller 104 compares the estimated emission value with the predetermined threshold. If the estimated methane emission for the ruminant animal exceeds the predetermined threshold, the hardware controller 104 generates the alert 126.
For example, a dairy farm has set a target to reduce its overall methane emissions by 20% compared to the previous year’s levels. The predetermined threshold for methane emissions per cow per day is set at 300 grams. If the system 100 estimates that a particular dairy cow’s methane emission is 350 grams per day, it will generate the alert, indicating that intervention is required to reduce the emissions from that cow.
In some regions, there may be regulations or guidelines in place for maximum permissible methane emissions from the livestock management. For example, a regulatory body has set a threshold of 250 grams of methane per day per cattle head. If the system 100 estimates that a particular cattle's methane emission is 280 grams per day, it will generate the alert 126, indicating that the farm needs to take measures to comply with the regulatory requirements. Upon receiving the alert 126, the farmer may explore various intervention strategies to reduce the methane emissions from the identified ruminant animal.
In an implementation, the hardware controller 104 is further configured to recommend an alternative feed data for the first ruminant animal based on the estimated methane emissions. The alternative feed data comprises adjusted types and quantities of the feed materials that reduce the methane emissions. The strategies may include adjusting the ruminant animal’s diet by modifying the feed composition, incorporating feed additives, or introducing specialized feed formulations designed to reduce methane emissions. Further, implementing improved manure management practices or anaerobic digestion systems to capture and utilize the methane generated from the manure of the cattle . Considering selective breeding programs or genetic improvement efforts to develop livestock breeds with lower methane emissions and adopt specific management practices, such as rotational grazing or improved forage quality, which can indirectly impact methane emissions.
In an implementation, the hardware controller 104 is further configured to generate a visualization of the estimated methane emissions over time for the first ruminant animal and display the visualization of the estimated methane emission on the user device 120 via the second user interface 124. The visualizations may be in the form of line graphs, showing methane emissions trends over time, or bar charts illustrating daily methane production. For example, the visualized methane emission data may be displayed on the second user interface 124, such as a computer or tablet. Each cow in the dairy farm may be represented individually, allowing the farmer to track methane emissions for each ruminant animal separately or the methane emissions of the whole herd of cattle can be displayed together with trends shown over time for the whole herd.
In an example, a line graph may display the methane emissions for each cow over time, with the x-axis representing time (e.g., days or weeks) and the y-axis representing methane emissions (e.g., grams per day). Each cow's methane emission data may be color-coded for easy identification. For example, different colours represent different cows in the herd. The second user interface 124, allows farmers to interact with the visualized data by zooming in on specific time periods, selecting individual cows for closer inspection, or viewing additional details such as feed composition and environmental conditions. Further, the visualizations, summary statistics such as average daily methane emissions, total methane emissions over a selected time period, and comparisons between individual cows or groups of cows are provided to assist in data interpretation. Farmers can use the visualized methane emission data to make informed decisions about the farm management practices. For example, the farmers can identify cows with high methane emissions and implement targeted interventions, such as adjusting feed compositions or optimizing feeding schedules, to reduce emissions.
In some implementations, the hardware controller 104 is further configured to cause a carbon credit action based on a difference between a first estimated methane emission before implementing the alternative feed data and a second estimated methane emission after implementing the alternative feed data. The hardware controller 104 determines the variance between the initial estimated methane emission and the re-estimated methane emission after implementing the alternative feed data. If the difference between the two estimations indicates a reduction in methane emissions, the hardware controller 104 may initiate a carbon credit action. For example, the carbon credit action may involve documenting the reduction and facilitating the process for the farm to receive carbon credits as a reward for mitigating methane emissions. The hardware controller 104 continuously monitors methane emissions and records the effectiveness of interventions such as adjusting feed data. The hardware controller 104 generates reports and documentation to support claims for carbon credits or other benefits associated with methane emission reduction efforts.
In another implementation, the hardware controller 104 is further configured to interface with a third-party carbon credit device to quantify reduced methane emissions based on implementing the intervention to reduce methane emissions from the first ruminant animal. The term “third-party carbon credit device” refers to an entity that facilitates the quantification, validation, and certification of reductions in greenhouse gas emissions, such as methane, carbon dioxide, or nitrous oxide. The third-party carbon credit device is used by organizations, governments, or carbon market participants to monitor and verify emission reduction projects.
For example, consider a dairy farm that is seeking to reduce the methane emissions from its dairy cattle herd. The farm implements several interventions aimed at achieving a reduction in methane emissions, including changes in feed composition, dietary supplements, and modifications in feeding practices. The hardware controller 104 collects detailed data on the implemented interventions and their effects on methane emissions from the dairy cattle herd. The hardware controller 104 interfaces with the third-party carbon credit device, providing data on the interventions implemented at the dairy farm and their impact on methane emissions reduction. The third-party carbon credit device utilizes established methodologies and standards to quantify the reduction in methane emissions achieved through the implemented interventions. The third-party carbon credit device validates the data provided by the dairy farm and issues certifications or carbon credits based on the verified reduction.
FIG. 2 is a flowchart 200 of a series of operations that illustrates a process of training and working of the system for methane estimation from the ruminants, in accordance with an embodiment of the present disclosure. FIG. 2 is described in conjunction with elements from FIG. 1. With reference to FIG. 2, there is shown a flowchart 200 that illustrates the process of developing the machine learning model 110 and estimating the methane emission from the ruminant animals. The flowchart 200 includes a series of operations 202 to 230. The hardware controller 104 (of FIG. 1) is configured to execute the operations shown in the flowchart 200.
At operation 202, the ruminant data is collected from the farmers directly via a mobile application. The ruminant data enables facilitates a deeper understanding of the factors influencing the methane emissions. The collected ruminant data serves as a rich repository for training the machine learning model 110. By incorporating diverse information from the farmers, the machine learning model 110 may learn patterns and relationships, enhancing its predictive capabilities for estimating the methane emissions.
At operation 204, the cleaning and processing of the ruminant data is performed, and the cleaned ruminant data is further processed for extracting the animal characteristics and feed data regarding the ruminant animal. Any entries in the ruminant data that are duplicated are identified and removed to maintain data integrity and prevent bias in analysis. Further, missing or null values in the ruminant data are addressed by either imputing them with appropriate values or removing them. Any inconsistencies or errors in the ruminant data, such as typos or formatting issues, are corrected to maintain accuracy and reliability. Data formats and units are standardized across the ruminant data to ensure consistency and facilitate analysis. The ruminant data is transformed into a structured format that is suitable for analysis. The transformation may involve feature engineering, encoding categorical variables into numerical representations, and scaling or normalization of numerical features.
At operation 206, the master database 114 is generated. The master database 114 comprises all types and quantities of all the feed materials and the ruminant animal characteristics. The farmer’s input is used as the base for the generation of the master database.
At operation 208, the nutritional database 116 is generated. A first spreadsheet is created, containing nutritional information on the feed materials of the ruminant animal. The nutritional information is collected from credible sources such as research papers, academic journals, and scientific publications where the nutritional information is thoroughly researched, validated, and published by experts in the field of the ruminant animal nutrition and agriculture. The spreadsheet includes details of the nutritional content feed material, for example, various nutritional content percentages like the CP, CF, the NDF, the ADF, lignin, the EE, ash, the GE, and the ME.
At operation 210, independent training variable features and dependent training variable features are prepared using the master database 114 and the nutritional database 116. Weight values of feed materials are converted from a fed basis to a dry matter basis. New feature columns like overall CP, overall CF, etc., are derived from the nutritional content of feed materials. The dependent training variable feature, i.e. methane emission per day, is calculated using a specific methodology by estimating the methane conversion factor (Ym) and thereafter estimating the methane emission. Further, the methane conversion factor (Ym) and estimated methane emission are stored in the master database 114 for further processing and training of the machine learning model 110. Independent training variable features are rounded to reduce noise. Statistical summary and visualization techniques like boxplots and histograms are used to understand data distribution and identify outliers. Skewness is checked and corrected using Yeo-Johnson transformation. Standardization of the independent training variable features is performed using a Standard Scaler application to bring them to a standard scale. Correlation analysis is conducted to understand the relationship between the independent and dependent training variable features. In some other implementations, the independent variable features are not standardized and are used as it for training the machine leaning model 110. In order to address multicollinearity, the Variance Inflation Factor (VIF) is used to identify highly correlated independent variables, which are then excluded from the machine learning model 110. The independent variables are utilized without applying transformations or standardization. For the dependent variable, no transformation or standardization is applied; however, its precision is capped to a specific number of decimal places, and outliers are removed using the Interquartile Range (IQR) method. The approach ensures the stability of the model by mitigating multicollinearity and handling outliers while maintaining the integrity of the dependent variable.
At operation 212, the derived independent training variable features and the derived dependent training variable features are passed through different AI algorithms in order to obtain the best regression model. The independent training variable features and dependent training variable features are extracted from the master database 114 and the nutritional database 116. The best random state for the train-test split and the accuracy of a linear regression algorithm is determined iteratively. Performance metrics, including, but not limited to, R² score, an adjusted R² score, MAE, RMSE, and MBD, are used to evaluate various regression algorithms. Algorithms like the Lasso Regression, Ridge Regression, Linear Multiple Regression, Polynomial Regression, etc., are compared, and the Linear Multiple Regression is identified as the best performer. Separate code blocks are used to fine-tune the model and obtain the best random state parameters.
At operation 214, the best regression model is obtained. linear multiple regression model is identified as the best-performing model. The model allows for a more complex analysis that can examine how multiple independent variables influence the target variable simultaneously.
At operation 216, hyperparameter tuning is performed. By performing hyperparameter tuning, the hardware controller 104 may systematically explore different combinations of hyperparameters and evaluate their performance using techniques like GridSearchCV. The hyperparameter tuning process allows the machine learning model 110 to find the hyperparameters that result in the best performance on the validation dataset, thus improving the ability of machine learning model 110 to generalize to unseen data.
At operation 218, the machine learning model 110 is deployed on cloud server so that it can be used for the methane emission estimation. Cloud servers can easily handle large-scale computations, allowing the model to process vast amounts of data efficiently and effectively. By deploying the machine learning model 110 on a cloud server, the users can access it remotely from anywhere with an internet connection, enabling widespread use and collaboration. The cloud server minimizes the risk of downtime and ensures the availability of the machine learning model 110. Further, the cloud server platforms implement advanced security measures to protect data and resources, helping to safeguard the machine learning model 110 and the data it processes from unauthorized access and cyber threats.
At operation 220, the hardware controller 104 is configured to receive the input data 108. The input data 108 is provided by the user directly, allowing them to enter specific details about their ruminant animals and the feed combinations given to them. The personalized input data 108 is crucial for the hardware controller 104 to accurately estimate methane emissions for each individual ruminant animal, considering the ruminant animal’s unique characteristics and dietary intake. By receiving this detailed input data 108 from the user via the first user interface 122, the hardware controller 104 can then process it further, access relevant databases, and use the trained machine learning model 110 to provide a personalized estimate of methane emissions per day for that specific ruminant animal.
At operation 222, the input data 108 underwent the data cleaning. The data cleaning involves removing any irrelevant or unnecessary data from the input data 108 based on predefined criteria. The input data 108 may contain missing or null values for certain animal characteristics or feed data. The hardware controller 104 may handle these missing values by either removing the corresponding rows or imputing the missing values using appropriate techniques. The input data 108 may contain different variations or spellings for the same ruminant animal breed. The hardware controller 104 may combine similar breed names into representative categories to ensure consistency in the input data 108. The hardware controller 104 may perform data type conversions on the input data 108 to ensure consistency and compatibility with the processing steps. The hardware controller 104 may analyse the input data 108 to identify and remove any outliers or extreme values that may adversely affect the accuracy of the methane emission estimation.
At operation 224, cleaned input data 108 is selected. By selecting cleaned input data 108, any inconsistencies, errors, or irrelevant information are removed, resulting in a more accurate representation of the ruminant animal's characteristics and the feed data. Ensures that the subsequent analysis is based on reliable information, leading to more accurate methane emissions estimation.
At operation 226, the feature data is generated using the nutritional database 116. The feature data captures the specific composition of the feed materials in detail. The precision of the feature data enables more informed analysis and predictions regarding the ruminant animal's dietary intake. The feature data derived using the nutritional database 116 is directly relevant to the dietary factors influencing the methane emissions in the ruminant animals. The relevance of the features enhances the effectiveness of subsequent analyses and interventions aimed at reducing the methane emissions.
At operation 228, the hardware controller 104 is further configured to access the derived feature data from the master database 114 containing methane conversion factors associated with different ruminant breeds. The process involves using the methane conversion factor (Ym) corresponding to different feed compositions for each individual ruminant.
At operation 230, the derived feature data is utilized for methane estimation by using trained machine learning model 110. The derived feature data is processed and passed to the pre-trained machine learning model 110 in the cloud server, which gives the prediction of the output in grams of methane emission per day on a per-head basis.
FIG. 3 is a flowchart of a method for efficiently estimating methane emissions from ruminants, in accordance with an embodiment of the present disclosure. FIG.3 is explained in conjunction with elements from FIGs. 1 and 2. With reference to FIG. 3, there is shown a flowchart of a method 300. The method 300 is executed at the server 102 (of FIG. 1). The method 300 may include steps 302 to 312.
At step 302, the method 300 includes receiving, by the hardware controller 104, the input data 108 related to the first ruminant animal. The input data 108 is stored in the master database 114 and comprises the animal characteristics, and the feed data comprises types and quantities of the feed materials consumed by the first ruminant animal. The input data 108 is received by the hardware controller 104 via the first user interface 122. The input data 108 specifies details regarding the first ruminant animal. The input data 108 encompasses information about the characteristics and feed data of the first ruminant animal. The feed provided to the ruminant is essential and directly influences its relationship with the enteric fermentation.
At step 304, the method 300 includes accessing, by the hardware controller 104, the nutritional database 116 comprising the nutritional data for the feed materials. At step 304, the hardware controller 104 accesses the nutritional database 116 by sending a request to retrieve nutritional data for the feed materials. This request is processed by the database management system, which searches for and retrieves the relevant nutritional data based on the provided criteria, such as the type or name of the feed materials. Once retrieved, the hardware controller 104 receives the nutritional data from the nutritional database 116 for further processing in the estimation of the overall nutritional content of the feed materials consumed by the ruminant animal.
At step 306, method 300 includes processing by the hardware controller 104 the input data 108 using the nutritional data from the nutritional database 116 to derive feature data representing the overall nutritional content of the feed materials consumed by the first ruminant animal. The processing involves analyzing the nutritional information of the feed materials based on the nutritional database 116, such as protein, fiber, carbohydrates, fats, vitamins, and minerals, and combining it with the quantity of each feed consumed by the animal. By synthesizing the input data 108, the hardware controller 104 generates the feature data that encapsulates the nutritional profile of the first ruminant's diet. The feature data is crucial for accurately estimating methane emissions, as the nutritional content of the feed directly influences the digestive processes and methane production of the first ruminant animal.

At step 308, the method 300 includes estimating, by the hardware controller 104, the methane emissions from the first ruminant animal based on the derived feature data using the machine learning model 110 trained on the ruminant data. The machine learning model 110 outputs a personalized estimate of the methane emissions per day for the first ruminant animal. The hardware controller 104 uses the empirical relationship in order to calculate the methane conversion factor (Ym). Further, after obtaining the methane conversion factor (Ym) the hardware controller estimates methane emission by using IPCC energy equation. The estimated methane conversion factor (Ym) and corresponding methane emission are stored back into the master database 114. The use of a machine learning model 110 allows for personalized estimates of the methane emissions based on the specific characteristics and dietary habits of the ruminant animal, leading to more accurate predictions. By considering a wide range of factors and leveraging the training dataset, the machine learning model 110 can provide more precise estimates of methane emissions compared to traditional methods. The machine learning model 110 may adapt to changes in the input data and continuously improve its predictions over time, ensuring it remains effective in estimating methane emissions for different scenarios.
At step 310, the method 300 includes generating, by the hardware controller 104, an alert 126 based on the estimated methane emission from the first ruminant animal, wherein the alert is indicative of an intervention required to reduce the estimated methane emissions if the estimated methane emission exceeds a predetermined threshold value. The alert 126 is obtained on the second user interface 124. By flagging high emitters, the system 100 brings attention to ruminant animals that may be having an outsized environmental impact, enabling timely interventions to reduce emissions. The alert 126 signals that changes may be needed for that specific animal, such as dietary adjustments or feed additives proven to reduce enteric methane production. By setting emission thresholds aligned with environmental goals, the alert system enables active management and reduction of the overall methane footprint from the ruminant animals over time.
Modifications to embodiments of the present disclosure described in the foregoing are possible without departing from the scope of the present disclosure as defined by the accompanying claims. Expressions such as "including", "comprising", "incorporating", "have", "is" used to describe and claim the present disclosure are intended to be construed in a non-exclusive manner, namely allowing for items, components or elements not explicitly described also to be present. Reference to the singular is also to be construed to relate to the plural. The word "exemplary" is used herein to mean "serving as an example, instance or illustration". Any embodiment described as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments and/or to exclude the incorporation of features from other embodiments. The word "optionally" is used herein to mean "is provided in some embodiments and not provided in other embodiments". It is appreciated that certain features of the present disclosure, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the present disclosure, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable combination or as suitable in any other described embodiment of the disclosure.
, Claims:We Claim:
1. An electronic system (100) for estimating methane emissions from ruminants, the electronic system comprising a hardware controller (104) configured to:
receive input data (108) related to a first ruminant animal from a user, via a first user interface (122), wherein the input data (108) is stored in a master database (114) and comprising animal characteristics and feed data comprising types and quantities of feed materials consumed by the first ruminant animal;
access a nutritional database (116) comprising nutritional data for the feed materials;
process the input data (108) using the nutritional data from the nutritional database (116) to derive feature data representing overall nutritional content of the feed materials consumed by the first ruminant animal;
estimate methane emissions from the first ruminant animal based on the derived feature data using a machine learning model (110) trained on ruminant data, wherein the machine learning model (110) outputs a personalized estimate of methane emissions per day for the first ruminant animal;
generate an alert (126) based on the estimated methane emission from the first ruminant animal, wherein the alert (126) is indicative of an intervention required to reduce the estimated methane emissions if the estimated methane emission exceeds a predetermined threshold value.
2. The electronic system (100) as claimed in claim 1, wherein the hardware controller (104) is further configured to form a methane emission database for the estimated methane emissions from the first ruminant animal.
3. The electronic system (100) as claimed in claim 1, wherein the hardware controller (104) is further configured to recommend an alternative feed data for the first ruminant animal based on the estimated methane emissions, and wherein the alternative feed data comprises adjusted types and quantities of the feed materials that reduce methane emissions.
4. The electronic system (100) as claimed in claim 3, wherein the hardware controller (104) is further configured to cause a carbon credit action based on a difference between a first estimated methane emission before implementing the alternative feed data and a second estimated methane emission after implementing the alternative feed data.
5. The electronic system (100) as claimed in claim 1, wherein the hardware controller (104) is further configured to interface with a third party carbon credit device to quantify reduced methane emissions based on implementing the intervention to reduce methane emissions from the first ruminant animal.
6. The electronic system (100) as claimed in claim 1, wherein the machine learning model (110) comprises a linear multiple regression model trained to map the derived feature data and the methane emission factors associated with different ruminant breeds.
7. The electronic system (100) as claimed in claim 6, wherein the machine learning model (110) comprises the linear multiple regression operation.
8. The electronic system (100) as claimed in claim 1, wherein in order to derive the feature data, the hardware controller (104) is further configured to convert the feed data to dry matter weights using the nutritional data; and calculate overall percentages of crude protein, crude fiber, neutral detergent fiber, ether extract and ash for the feed material consumed by the first ruminant animal.
9. The electronic system (100) as claimed in claim 1, wherein the hardware controller (104) is further configured to generate a visualization of the estimated methane emissions over time for the first ruminant animal and display the visualization of the estimated methane emission on a user device, via a second user interface (124).
10. The electronic system (100) as claimed in claim 1, wherein the hardware controller (104) is further configured to train the machine learning model (110) by:
obtaining a training dataset comprising the ruminant data including the animal characteristics, the feed data, and corresponding recorded methane emissions;
preprocessing the training dataset to handle missing data and outliers;
deriving training feature data and the methane emission factors associated with different ruminant breeds from the training dataset; and
mapping the derived training feature data and the methane emission factors associated with each individual ruminant to the recorded methane emissions.
11. The electronic system (100) as claimed in claim 10, wherein in order to preprocess the training dataset, the hardware controller (104) is further configured to:
remove unnecessary data columns and rows based on predefined criteria;
handle null values in the animal characteristics and the feed data;
combine similar breed names into representative categories; and
remove duplicate data rows.
12. The electronic system (100) as claimed in claim 10, wherein in order to train the machine learning model (110), the hardware controller (104) is further configured to:
apply data transformations to the derived training feature data to reduce skewness and stabilize variance;
standardize the transformed training feature data;
evaluate one or more candidate machine learning operations using the standardized training feature data and the recorded methane emissions; and
select a candidate machine learning operation from the one or more candidate machine learning operations to be utilized in the machine learning model for estimating methane emissions from the ruminant data based on the evaluation.
13. The electronic system (100) as claimed in claim 12, wherein in order to train the machine learning model (110), the hardware controller (104) is further configured to perform hyperparameter tuning on the selected candidate machine learning operation to optimize performance of the machine learning model.
14. The electronic system (100) as claimed in claim 1, wherein the hardware controller (104) is further configured to periodically retrain the machine learning model (110) using continuously collected ruminant data and the estimated methane emissions to improve accuracy of the machine learning model (110) over time.
15. A method (300) for estimating methane emissions from ruminants, the method (300) comprising:
receiving, by a hardware controller (104), input data related to a first ruminant animal, wherein the input data (108) is stored in a master database (114) and comprising animal characteristics and feed data comprising types and quantities of feed materials consumed by the first ruminant animal;
accessing, by the hardware controller (104), a nutritional database (116) comprising nutritional data for the feed materials;
processing, by the hardware controller (104), the input data (108) using the nutritional data from the nutritional database (116) to derive feature data representing overall nutritional content of the feed materials consumed by the first ruminant animal;
estimating, by the hardware controller (104), methane emissions from the first ruminant animal based on the derived feature data using a machine learning model (110) trained on ruminant data, wherein the machine learning model (110) outputs a personalized estimate of methane emissions per day for the first ruminant animal;
generating, by the hardware controller (104), an alert (126) based on the estimated methane emission from the first ruminant animal, wherein the alert (126) is indicative of an intervention required to reduce the estimated methane emissions if the estimated methane emission exceeds a predetermined threshold value.