Enhancing microgrid performance with AI‐based predictive control: Establishing an intelligent distributed control system
Afshin Hasani,Hossein Heydari,Mohammad Sadegh Golsorkhi
DOI: https://doi.org/10.1049/gtd2.13191
2024-07-26
IET Generation Transmission & Distribution
Abstract:The primary objective of this paper is to present a method that utilizes deep neural networks (DNNs) for effective microgrid control. By training the DNN, the network becomes capable of managing load changes within the microgrid and adjusting the output power of the distributed generation (DG) system. To develop and train the DNN model, it is essential to use information from the secondary controller, along with voltage and current measurements from the output of each distributed energy resource (DER). These data are derived from an AC microgrid that is regulated by P–f and Q–V droop characteristics under steady‐state conditions and various operating modes, with control implemented in a distributed manner. This specific microgrid serves as the dataset for training the DNN. Leveraging this data, the DNN is able to predict voltage and current trends for any DG source through artificial intelligence. As a result, this approach reduces the need for periodic communication and integrates the functions of primary and secondary controllers within microgrids to form an Intelligent distributed control (IDC) system. Additionally, by preprocessing the acquired data to filter out information that causes voltage and current fluctuations during transient states and replacing it with stable data, the DNN can more accurately predict voltage and current trends during these states. Consequently, this aids in achieving a seamless transition from transient to steady state. Microgrids play a pivotal role in modern power distribution systems, necessitating precise control methodologies to tackle challenges such as performance instability, especially during islanding operations. This paper introduces an advanced control strategy that employs artificial intelligence, specifically deep neural network (DNN) predictions, to enhance microgrid performance, particularly in an islanding mode where voltage and frequency (VaF) deviations are critical concerns. By utilizing real‐time data and historical trends, the proposed controller accurately forecasts power demand and generation patterns, enabling proactive planning and optimization of efficiency, reliability, and sustainability in microgrid management. One significant aspect of this approach is to establish an intelligent distributed control system that minimizes reliance on communication devices while ensuring that VaF remains within acceptable limits. Moreover, it consolidates the roles of primary and secondary controllers within the microgrid and facilitates the prediction of load changes and load injection processes. This capability significantly reduces microgrid VaF deviations, enhancing system performance through precise power distribution and balanced coordination among distributed generators. Consequently, it ensures the stability and reliability of the system. In summary, the integration of DNN‐based predictive control represents a significant advancement in microgrid management, providing a solution to address performance challenges and optimize operational efficiency, reliability, and sustainability.
engineering, electrical & electronic