Description:FIELD OF THE INVENTION
This invention relates to intelligent control systems for power electronics. More specifically, it concerns a decentralised artificial intelligence (AI)-augmented droop and impedance control system and method for balanced power distribution among parallel-connected inverters operating under resistive–inductive (RL) load conditions in microgrids.
BACKGROUND OF THE INVENTION
The increasing integration of distributed energy resources, particularly renewable energy sources such as solar and wind, has driven the widespread adoption of microgrids. Within these microgrids, multiple voltage source inverters (VSIs) are often connected in parallel to ensure system reliability, redundancy, and scalability. These inverters must cooperate to share power proportionally and maintain voltage and frequency stability under varying load conditions.
Conventional control strategies for parallel inverters include droop control and virtual impedance-based methods. While widely used due to their simplicity and decentralized nature, these approaches suffer from several limitations, especially under resistive-inductive (RL) load conditions. Fixed droop parameters and static virtual impedances cannot effectively adapt to dynamic load changes, often leading to circulating currents, inaccurate power sharing, voltage distortion, and slow transient response. These issues are particularly critical in mission-critical or renewable-heavy microgrids, where fast and accurate control is essential.
Efforts to improve performance have explored centralized optimization and adaptive droop techniques. However, centralized solutions introduce communication bottlenecks, reduced system resilience, and a single point of failure. Moreover, they often rely on full data sharing, which can compromise data privacy and increase network load.
Given these challenges, there is a pressing need for an intelligent, decentralized control framework that can adapt in real time, enhance power quality, and maintain stable operation under dynamic RL load conditions without relying on centralized coordination. The present invention addresses this need by integrating machine learning techniques such as reinforcement learning, predictive neural networks, and federated learning into a novel AI-based inverter control system.
OBJECTS OF INVENTION:
1. To provide an intelligent, adaptive control framework for parallel inverters in microgrids operating under resistive-inductive (RL) load conditions.
2. To minimize circulating currents between inverters by dynamically adjusting virtual impedance using reinforcement learning agents.
3. To improve accuracy in active and reactive power sharing among inverters under varying load conditions.
4. To implement predictive voltage and frequency control through neural network-based models (e.g., LSTM or Transformer) that forecast short-term power demand.
5. To enable decentralized coordination among inverters using federated learning, thereby reducing communication overhead and preserving data privacy.
6. To incorporate unsupervised learning models for real-time anomaly detection and automatic corrective actions in case of faults, disturbances, or load changes.
7. To enhance system scalability, resilience, and power quality, ensuring compliance with IEEE standards in renewable-integrated and mission-critical microgrid environments.
Parallel-connected inverters feeding RL loads face significant power sharing challenges that existing techniques cannot adequately address. Current droop control methods achieve only 85-90% power sharing accuracy with RL loads compared to 95-98% with resistive loads due to complex reactive power interactions. These systems suffer from excessive circulating currents causing 5-15% efficiency losses, slow transient response with settling times 3-5 times longer than required, and poor reactive power coordination. Additionally, existing methods are highly sensitive to parameter variations, require complex communication infrastructure, and fail to meet IEEE power quality standards with THD levels exceeding 8-12%. These limitations severely impact industrial applications including motor drives, UPS systems, and microgrids, necessitating innovative control strategies specifically designed for RL load characteristics.
US11121540B2: A multi-port power converter includes a housing that includes a plurality of ports structured to electrically interface to a plurality of loads, the plurality of loads having distinct electrical characteristics. The multi-port power converter also includes a plurality of solid state components configured to provide selected electrical power outputs and to accept selected electrical power inputs and a plurality of solid state switches configured to provide selected connectivity between the plurality of solid state components and the plurality of ports.
US11052784B2: A system include a vehicle having a motive electrical power path and a power distribution unit (PDU). The PDU includes a current protection circuit disposed in the motive electrical power path, the current protection circuit including a thermal fuse and a contactor in a series arrangement with the thermal fuse, and a high voltage power input coupling having a first electrical interface for a high voltage power source. The high voltage power output coupling includes a second electrical interface for a motive power load. The current protection circuit electrically couples the high voltage power input to the high voltage power output, and the current protection circuit is at least partially disposed in a laminated layer of the PDU. The laminated layer includes an electrically conductive flow path disposed between two electrically insulating layers.
Parallel-connected inverters feeding RL loads in microgrids often experience poor power sharing, excessive circulating currents, and slow transient response. Traditional droop and virtual impedance methods use fixed parameters and lack adaptability, resulting in inaccuracies, higher harmonic distortion, and failure to meet power quality standards. Centralised control solutions introduce communication bottlenecks and single points of failure. The present invention solves these problems by embedding reinforcement learning agents, predictive neural networks, federated learning coordination, and unsupervised anomaly detection into each inverter to dynamically adjust virtual impedance, predict demand, and coordinate without centralised control, thereby improving accuracy, stability, and resilience.
SUMMARY OF THE INVENTION
This summary is provided to introduce a selection of concepts, in a simplified format, that are further described in the detailed description of the invention.
This summary is neither intended to identify key or essential inventive concepts of the invention and nor is it intended for determining the scope of the invention.
The invention provides an AI-augmented control system for parallel inverters supplying RL loads in a microgrid. Each inverter incorporates a reinforcement learning agent that continuously adjusts virtual impedance in real time to minimise circulating currents and improve power sharing.
A neural network-based predictive controller forecasts short-term active and reactive power demand using local electrical measurements to enable anticipatory voltage and frequency regulation.
Inverters coordinate via a federated learning protocol that exchanges only model updates, reducing communication bandwidth and preserving data privacy. An unsupervised learning model monitors inverter behaviour to detect anomalies such as faults or load changes and trigger corrective actions automatically.
This decentralised and adaptive control framework enhances scalability, resilience, and power quality compliance with IEEE standards, making it suitable for renewable-integrated and mission-critical microgrid applications.
To further clarify advantages and features of the present invention, a more particular description of the invention will be rendered by reference to specific embodiments thereof, which is illustrated in the appended drawings. It is appreciated that these drawings depict only typical embodiments of the invention and are therefore not to be considered limiting of its scope. The invention will be described and explained with additional specificity and detail with the accompanying drawings.
The invention presents an AI-based control framework for parallel inverters operating in microgrids under resistive-inductive (RL) load conditions. Conventional droop and impedance-based methods exhibit poor adaptability to RL load dynamics, resulting in circulating currents, inaccurate power sharing, and harmonic distortion. This invention introduces a decentralized, intelligent system to address these limitations. Each inverter is equipped with a reinforcement learning (RL) agent that continuously adjusts virtual impedance in real time to minimize circulating currents and improve proportional power sharing. The system also includes a neural network-based predictive controller (e.g., LSTM or transformer) that forecasts short-term active and reactive power demand using local electrical measurements, enabling anticipatory adjustments to voltage and frequency references. To maintain coordination without centralized control, inverters exchange only model updates using a federated learning protocol, allowing for collaborative intelligence while preserving data privacy and minimizing communication bandwidth. Additionally, unsupervised learning models monitor inverter behavior to detect anomalies such as faults or load changes and trigger corrective actions. This architecture provides a scalable, adaptive, and communication-resilient power sharing strategy tailored for microgrids. It ensures higher accuracy, faster dynamic response, and compliance with IEEE power quality standards, making it ideal for renewable-integrated and mission-critical microgrid environments.
BRIEF DESCRIPTION OF THE DRAWINGS
The illustrated embodiments of the subject matter will be understood by reference to the drawings, wherein like parts are designated by like numerals throughout. The following description is intended only by way of example, and simply illustrates certain selected embodiments of devices, systems, and methods that are consistent with the subject matter as claimed herein, wherein:
Fig.1: Operational Workflow diagram of parallel connected inverter
Fig.2: Diagram of parallel connected inverter
The figures depict embodiments of the present subject matter for the purposes of illustration only. A person skilled in the art will easily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the disclosure described herein.
DETAILED DESCRIPTION OF THE INVENTION
The detailed description of various exemplary embodiments of the disclosure is described herein with reference to the accompanying drawings. It should be noted that the embodiments are described herein in such details as to clearly communicate the disclosure. However, the amount of details provided herein is not intended to limit the anticipated variations of embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the scope of the present disclosure as defined by the appended claims.
It is also to be understood that various arrangements may be devised that, although not explicitly described or shown herein, embody the principles of the present disclosure. Moreover, all statements herein reciting principles, aspects, and embodiments of the present disclosure, as well as specific examples, are intended to encompass equivalents thereof.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the singular forms “a",” “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises,” “comprising,” “includes” and/or “including,” when used herein, specify the presence of stated features, integers, steps, operations, elements and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and/or groups thereof.
It should also be noted that in some alternative implementations, the functions/acts noted may occur out of the order noted in the figures. For example, two figures shown in succession may, in fact, be executed concurrently or may sometimes be executed in the reverse order, depending upon the functionality/acts involved.
In addition, the descriptions of "first", "second", “third”, and the like in the present invention are used for the purpose of description only, and are not to be construed as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Thus, features defining "first" and "second" may include at least one of the features, either explicitly or implicitly.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which example embodiments belong. It will be further understood that terms, e.g., those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The invention presents an AI-based control framework for parallel inverters operating in microgrids under resistive-inductive (RL) load conditions. Conventional droop and impedance-based methods exhibit poor adaptability to RL load dynamics, resulting in circulating currents, inaccurate power sharing, and harmonic distortion. This invention introduces a decentralized, intelligent system to address these limitations. Each inverter is equipped with a reinforcement learning (RL) agent that continuously adjusts virtual impedance in real time to minimize circulating currents and improve proportional power sharing. The system also includes a neural network-based predictive controller (e.g., LSTM or transformer) that forecasts short-term active and reactive power demand using local electrical measurements, enabling anticipatory adjustments to voltage and frequency references. To maintain coordination without centralized control, inverters exchange only model updates using a federated learning protocol, allowing for collaborative intelligence while preserving data privacy and minimizing communication bandwidth. Additionally, unsupervised learning models monitor inverter behavior to detect anomalies such as faults or load changes and trigger corrective actions. This architecture provides a scalable, adaptive, and communication-resilient power sharing strategy tailored for microgrids. It ensures higher accuracy, faster dynamic response, and compliance with IEEE power quality standards, making it ideal for renewable-integrated and mission-critical microgrid environments.
This invention introduces a novel AI-based control framework for parallel inverters in microgrids operating under resistive-inductive (RL) loads. Unlike conventional droop or impedance methods, it employs reinforcement learning to adapt virtual impedance in real time, minimizing circulating currents and enhancing power sharing. A neural network controller (e.g., LSTM or transformer) predicts short-term power demand for anticipatory voltage and frequency adjustments. Coordination is decentralized using federated learning, ensuring privacy and low bandwidth use. Unsupervised models detect anomalies and trigger corrective actions. This integrated, intelligent system provides scalable, accurate, and communication-resilient power control for modern renewable-based and mission-critical microgrids.
The invention comprises a network of parallel-connected voltage source inverters supplying RL loads in a microgrid environment.
Each inverter includes an AI-based control module with four main subcomponents: a reinforcement learning agent, a predictive neural network controller, a federated learning communication interface, and an unsupervised anomaly detection unit.
The reinforcement learning agent dynamically tunes the inverter’s virtual impedance parameters in response to real-time voltage and current measurements to minimise circulating currents and achieve proportional active and reactive power sharing.
The predictive neural network controller, implemented for example as an LSTM or transformer model, forecasts short-term active and reactive power demand using historical and instantaneous local measurements. These predictions allow anticipatory adjustment of the inverter’s voltage and frequency references, improving transient response and reducing power quality deviations during rapid load changes.
The federated learning interface enables each inverter to exchange only model parameter updates with peers rather than raw operational data. This decentralised coordination reduces communication overhead, preserves data privacy, and eliminates the single point of failure inherent in centralised control schemes.
The unsupervised anomaly detection unit continuously analyses inverter behaviour to identify unusual patterns indicative of faults, disturbances, or load changes and automatically triggers corrective actions such as parameter resets, protective shutdowns, or isolation of faulty units.
The control module is implemented in software running on a digital signal processor or microcontroller integrated within each inverter. It interfaces with standard voltage and current sensors already present in inverter hardware.
The system supports hot-swapping of inverters without requiring retuning of the overall network, improving scalability and reliability.
Adaptive tuning allows the system to maintain compliance with IEEE power quality standards under dynamic RL load conditions, including harmonic distortion limits.
The invention may be applied in renewable-integrated microgrids, uninterruptible power supply (UPS) systems, and industrial motor drive systems where parallel inverters supply inductive loads.
By integrating reinforcement learning, predictive modelling, federated coordination, and anomaly detection, the system achieves faster transient response, more accurate power sharing, and reduced circulating currents compared to conventional droop and impedance methods.
The architecture is modular, allowing different machine learning algorithms to be substituted or upgraded without altering the basic control structure.
Security and privacy safeguards ensure that model updates exchanged via federated learning are encrypted and authenticated.
This approach provides a robust, decentralised, and adaptive solution for balanced power distribution in inverter-based microgrids.
BEST METHOD OF WORKING
The preferred embodiment deploys the invention in a microgrid with multiple parallel inverters. Each inverter is equipped with a digital controller running the reinforcement learning agent, predictive neural network controller, federated learning interface, and anomaly detection unit. Local voltage and current sensors feed data to the controller. Reinforcement learning dynamically adjusts virtual impedance; the predictive controller anticipates power demand; federated learning synchronises model updates across inverters; and anomaly detection triggers protective measures when needed. This configuration minimises circulating currents, improves power sharing accuracy, reduces communication bandwidth, and maintains compliance with IEEE power quality standards.
, Claims:1. A decentralised AI-based control system for balanced power distribution among parallel-connected inverters comprising:
a reinforcement learning agent in each inverter configured to dynamically adjust virtual impedance in real time to minimise circulating currents and improve proportional power sharing;
a predictive neural network controller in each inverter configured to forecast short-term active and reactive power demand using local electrical measurements and to enable anticipatory adjustment of voltage and frequency references;
a federated learning interface configured to coordinate inverters by exchanging model updates without sharing raw operational data;
an unsupervised anomaly detection unit configured to monitor inverter behaviour and trigger corrective actions upon detecting faults or load changes; and
an output module configured to provide power quality metrics and system status to a user.
2. The system as claimed in claim 1, wherein the reinforcement learning agent tunes virtual impedance parameters based on real-time voltage and current measurements to reduce circulating currents.
3. The system as claimed in claim 1, wherein the predictive neural network controller is implemented as a long short-term memory or transformer model to anticipate load variations.
4. The system as claimed in claim 1, wherein the federated learning interface preserves data privacy and reduces communication bandwidth by transmitting only model parameter updates.
5. The system as claimed in claim 1, wherein the unsupervised anomaly detection unit autonomously isolates faulty inverters or initiates corrective measures.
6. A method for decentralised AI-based control of balanced power distribution among parallel-connected inverters comprising:
embedding a reinforcement learning agent in each inverter to dynamically adjust virtual impedance in real time based on local voltage and current measurements;
forecasting short-term active and reactive power demand using a predictive neural network controller in each inverter to enable anticipatory adjustment of voltage and frequency references;
coordinating inverters via a federated learning protocol that exchanges only model updates without sharing raw data;
monitoring inverter behaviour using an unsupervised anomaly detection unit to identify faults or load changes and trigger corrective actions; and
outputting power quality metrics and system status to a user.
7. The method as claimed in claim 6, wherein virtual impedance parameters are tuned continuously to minimise circulating currents and improve power sharing accuracy.
8. The method as claimed in claim 6, wherein the predictive neural network controller anticipates rapid load changes to improve transient response and reduce power quality deviations.
9. The method as claimed in claim 6, wherein federated learning reduces communication overhead and eliminates a central control point.
10. The method as claimed in claim 6, wherein anomaly detection autonomously initiates protective shutdowns or parameter resets when faults or disturbances are detected.