Description:A MACHINE LEARNING-BASED APPROACH FOR SENSOR FAULT DETECTION IN BATTERY MANAGEMENT SYSTEMS
Technical Field
[0001] The embodiments herein generally relate to a machine learning-based approach for sensor fault detection in battery management systems.
Description of the Related Art
[0002] The industry may benefit from the advancement of battery-powered energy storage technologies, such as those used in hybrid trains, electric cars, and other e-mobility applications. Energy storage systems play a crucial role in electric vehicle and smart grid technologies by facilitating the effective distribution and transmission of energy. There are many possibilities for energy storage because there is a wide range of batteries on the market. Furthermore, hundreds or even thousands of single battery cells are needed for high power LIB applications like electric vehicles and grid-tied energy storage systems because of the limitations of a single LIB cell's cell voltage and storage capacity. Because cell inconsistencies in a LIB pack are a typical problem, a suitable BMS is also essential for the LIB pack's safe and dependable operation, as well as the operation of each and every cell in the battery pack. One of the most important parts of the powertrain in an EA/HEA to maintain from an operational and cost perspective is the battery pack. A battery management system (BMS) regulates and keeps an eye on the performance and functionality of every battery cell as well as the pack as a whole.
[0003] The study in examined battery fires in battery electric cars, hybrid EVs, and electric buses in order to obtain a qualitative understanding of the fire risks and dangers associated with battery-powered EVs. Key characteristics of battery fires that were found during testing and were present in various EV fire scenarios were also examined. A modular battery management system is recommended in for an electric vehicle. Various abusive operating conditions, such as over discharge or overcharge events, low-temperature startup, vibration, and increased heat generation leading to metallic lithium plating, solid electrolyte interphase layer formation, and lithium dendritic formation, can impact LIB performance and hasten its aging process, potentially resulting in catastrophic failure while in operation. Applications involving electric vehicles and hybrid electric airplanes (EV/HEV) prioritize safety above all else. Power system safety might be ensured by a BMS outfitted with a precise and clever problem diagnosis system. Data- and model-based diagnosis is the two general categories into which fault diagnosis techniques fall.
[0004] In order to determine when to charge an EV during a connection session, the paper [16] proposed an intelligent charging technique based on machine learning (ML). By making real-time charging decisions based on a range of auxiliary data, including driving, environment, price, and demand time series, the overall cost of vehicle energy was reduced. However, the failures related to the temperature, voltage, and current sensors were categorized under sensor faults, while the problems related to the cooling system, terminal connector, controller area network (CAN) bus, high voltage contactor, and fuse were categorized under actuator faults. Conversely, in model-based techniques, the battery model is developed using battery electrical equations. Subsequently, the defect diagnosis system decides by contrasting the output of the model with the output that was measured. A possible issue is identified if there is a discernible discrepancy between these two outputs. Model-based fault detection techniques are becoming more and more popular because of their advantages in terms of flexibility, affordability, and accuracy.
SUMMARY
[0005] In view of the foregoing, an embodiment herein provides a machine learning-based approach for sensor fault detection in battery management systems. In some embodiments, wherein a diagnostic technique for voltage problems that was demonstrated under real-world operating settings on an electric vehicle (EV) equipped with a multiple-cell battery system. This study shows that, following the collection and preparation of the customary data periods from the Operation Service and Management Center for Electric Vehicles (OSMC-EV), the overvoltage issue for Li-ion batteries cell may be observed from the voltage curves. Recognizing the significance of fault detection and diagnosis for the safe and dependable functioning of LIBs, numerous research investigations were carried out with the goal of creating a fault diagnostic technique that is precise, dependable, robust, and simple to use. The significance of developing an efficient fault diagnostic system for the progress of LIB-powered systems was briefly outlined by Lu et al. Machine learning-based parameter estimation is becoming more and more popular. Two stages in the identifying process can be eliminated using machine learning: Accurate physical information about the battery is required; one must understand the nonlinear link between battery characteristics and observed data, such as operating temperature and terminal voltage.
[0006] In some embodiments, on the interclass correlation coefficient (ICC) technique (EVs) for defect detection. The voltages were obtained from the Operation Service and Management Center for Electric Vehicles, and the off-trend voltage drop was recorded using the recommended method to calculate ICC values. The precise equivalent circuit model (ECM) of LIB is a major determinant of the precision and dependability of model-based fault diagnosis techniques. It is difficult to obtain a very accurate model since the extremely nonlinear LIB's internal features are still little understood. The introduction of ML-based methods removes this restriction. Furthermore, the use of ML-based techniques significantly lessens the effects of measurement noises that restrict the application of signal processing-based methods. Numerous widely used characterization techniques, including galvanostatic intermittent titration technique (GITT), electro-chemical impedance spectroscopy (EIS), and potentiostatic intermittent titration technique (PITT), offer comprehensive battery data for transportation applications.3.
[0007] In some embodiments, wherein the peculiarities of BMS for electric vehicles and stationary, large-scale energy storage. A wide range of BMS-related subjects are covered in the exam, including testing, component functionality, topology, operation, architecture, and BMS safety concerns. This review study's main goal is to give researchers quick access to references and guidance when they're creating an accurate, effective, dependable, and simple-to-implement machine learning (ML)-based data-driven fault diagnostic approach for the LIB system. a thorough categorization of the particular ML-based techniques now employed in the LIB BMS. The hybrid pulse power characterization (HPPC) approach can be used to characterize batteries in order to address these problems. In this work, a machine learning parameter estimator (MLPE) for the battery model is developed using the data gathered from HPPC testing.
[0008] These and other aspects of the embodiments herein will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings. It should be understood, however, that the following descriptions, while indicating preferred embodiments and numerous specific details thereof, are given by way of illustration and not of limitation. Many changes and modifications may be made within the scope of the embodiments herein without departing from the spirit thereof, and the embodiments herein include all such modifications.
BRIEF DESCRIPTION OF THE DRAWINGS
[0009] The embodiments herein will be better understood from the following detailed description with reference to the drawings, in which:
[0010] FIG. 1 illustrates a machine learning-based approach for sensor fault detection in battery management systems according to an embodiment herein; and
[0011] FIG. 2 illustrates the overview of research process according to an embodiment herein.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
[0012] The embodiments herein and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments that are illustrated in the accompanying drawings and detailed in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.
[0013] FIG. 1 illustrates a machine learning-based approach for sensor fault detection in battery management systems according to an embodiment herein. In some embodiments, the initial step involves collecting sensor data that reflect the battery’s outward features. New sensors, such as built-in pressure sensors and acoustic sensors, can be utilized to obtain parameters that characterize battery internal state information, which helps to achieve specific fault isolations accurately. They lists out the sensor sample data utilized for the research. Artificial Neural Network (ANN) is one of the most widely used frameworks of ML algorithms to perform a wide variety of tasks. It is inspired by the biological neural networks that constitute animal brains. ANN uses supervised learning approaches during model training. Features like self-adaptability and learning abilities of the animal brain enable ANN to perform tasks by considering examples, generally without being programmed with task-specific rules. Moreover, ANN is capable of effectively capturing the dynamics of a highly nonlinear system. Among various available rechargeable batteries, LIBs are the most desirable one for different applications. Here, LIB cells with LTO anode and LMO cathode are considered as the power supply of EA/HEA application due to its higher energy efficiency, fast charging, and higher cycle life than other types of LIBs.
[0014] In some embodiments, the data gathered by the sensors only reflect the battery’s outward features not its inside conditions. Additionally, there is a connection between several faults, and the characteristics of each fault are not always clear. Therefore, achieving reliable fault detection and separation from unidentified fault data remains exceedingly difficult. New sensors such as built-in pressure sensors, acoustic sensors, etc. can be used to obtain parameters characterizing the battery’s internal state information in order to accurately achieve specific fault isolation. Collaboration among trees in RF makes the model more robust compared to any single classifier typically used in other statistical classification problems. RF is a linear classifier with reduced computational complexity when compared to some other popular classifiers, making it suitable for lightweight algorithms for real-time operation. Considering the details and adding components in the cell model is a trade-off between accuracy and complexity. Different researches have shown the second order of ECM can be a good candidate for the cell modeling due to its accuracy and low complexity. The elements of this model shown in Figure are a resistance (R0), and two parallel Resistance-Capacitor.
[0015] In some embodiments, the statistical analysis or the analysis for multivariate datasets for feature extractions is performed on the principle of sparse principal component analysis (SPCA). Data extractions are principally carried out by introducing sparsity structures to input variables, where SPCA enhances the effectiveness of the traditional technique for dimensionality and reduction, commonly known as principal component analysis (PCA). The use of SVM in regression is also termed Support Vector Regression (SVR). SVR uses different kernel functions and regression algorithms to transfigure a nonlinear model into a linear model for ease of analysis. There is also another variant of SVM, namely, kernel space vector machine (KSVM). Further details of SVM can be found in reference. The tested battery is placed in the chamber to conduct the experiments in various controlled temperatures. The measurement errors of current and voltage are below 0.1% and the temperature accuracy is ±0.3C. Data exchange between the setup and the computer was through a serial connection. For later analysis, the measured data were saved as an Excel file on the computer.
[0016] FIG. 2 illustrates the overview of research process according to an embodiment herein. In some embodiments, an optimization problem can be formulated for SPCA by introducing an elastic net constraint for a fixed a. To address this issue, an alternate minimization procedure can be utilized to minimize the SPCA criteria. Detailed mathematical formulas are available to implement SPCA. Due to its sparsity, SPCA has been widely used in gene expression analysis as it facilitates the understanding of data and the identification of key genes. In addition, sparsity can aid in the generalization of a learned model and prevent over-fitting. The two primary goals of this technique are clustering the data into groups by similarity and dimensionality reduction to compress the data while maintaining its structure and usefulness of data. GPR also uses kernel-based ML approaches which can discover prognostics by leveraging prior knowledge based on the Bayesian model. Thereafter, it utilizes the variance around its mean prediction to provide information about the associated uncertainty in the system. In phase one which is data collection, after data cleansing, the data are split into training and validation data. In phase two, these two datasets are applied to choose a suitable machine learning model. In phase three, using the MLPE generated in the second phase, the model parameters are estimated based on inputs, which include SOC and operating temperature.
[0017] In some embodiments, an innovative population-based optimization technique that has seen extensive usage in practical optimization applications is the improved marine predator’s algorithm (EMPA). However, due to a lack of population variety in the late stages of optimization, EMPA may quickly enter a local optimum. The three stages of the marine predator’s algorithm simulates the interactions between marine predators and their prey as shown in Table 3 of enhanced MPA algorithm matrix. Comparative analysis indicated that the fusion of GA and BPNN improved the fault diagnosis performance compared to conventional BPNN. However, no insight into the practical application of the fault diagnosis method was presented in this study. On-line diagnosis and fault handling were not covered as well. The reported prediction error of GA-BPNN is around 2% which may not be acceptable to some of the real-world applications. Each pulse has duration of 20 seconds, and the pulses are separated by 30 seconds pauses.41 this process is repeated in time, until the battery voltage reaches its cut-off. Since the model parameters are dependent to temperature and SOC of the cell, HPPC experiments were conducted under different conditions.
[0018] In some embodiments, the existing approaches are the artificial neural network (ANN), support vector machine (SVM), linear regression (LR), and Gaussian process regression (GPR), whereas the suggested technique is the incipient bat-optimized deep residual network (IB-DRN). The proposed incipient bat-optimized deep residual network technique achieves a high level of accuracy (98%), especially when compared to the existing methods of artificial neural network (53%), support vector machine (73%), linear regression (82%), and Gaussian process regression (66%) that are currently in use. The concept of the deep-learning-enabled fault prognosis method was introduced by Hong et al. [56] where long short-term memory (LSTM) recurrent neural network (LSTM-RNN) was used for multi-forward-step voltage prediction to determine the advent of battery faults and mitigate runaway risk. To ensure the prediction accuracy and model robustness, a high volume of real-world operational data of an electric taxi was used. Alongside the influence of the weather and driver’s behavior on the LIBs, the performance was also considered. SVM follows a technique called the kernel trick for transforming the data and determines an optimal boundary between potential outputs based on those transformations. The SVM with cubic, quadratic, and linear kernels have been used in this analysis.

, Claims:1. A method for the a machine learning-based approach for sensor fault detection in battery management systems, wherein the sensor fault detection method comprises;
the corporate managers must continue to introduce CSR principles related to employee management to improve social acceptance and to increase their commitment while improving organizational performance and development;
the safety in electric vehicles depends heavily on BMSs, which control the electronics of the rechargeable battery pack or individual cells;
to improve the safety and reliability of battery management systems, this article proposes a false battery data detection and classification system based on incipient bat-optimized deep residual networks (IB-DRN);
in general, ML algorithms can be classified based on learning approaches, namely, supervised learning, unsupervised learning and reinforcement learning; and
despite high accuracy and adaptability to uncertainties the major limitations of ANN-based methods are the requirement of a large training data set, a large memory size, a time-consuming training process and poor generalization capability.