An Explainable Artificial Intelligence Model for Energy Flow Management in EV-Dominated Microgrids

Abstract

Microgrids are being changed rapidly as Electric Vehicles (EVs) become more popular and many new variable and unpredictable energy loads are being added, along with an abundance of new ways to store energy using electric vehicles (EVs) through vehicle-to-grid interactions. To achieve optimal energy flow control in microgrids characterized by power generation and energy storage facilities with an overwhelming percentage of power flowing from the electric vehicle (EV), intelligent decision support systems that provide suitable solutions based on accurate and credible information must be used. This paper describes a model utilizing Explainable Artificial Intelligence (XAI) for managing the flow of energy between various types of energy generation resources, stationary energy storage resources, EVs, and the grid in order to produce a minimum cost/performance and maximum reliability. Specifically, the proposed model integrates SHAP (SHapley Additive exPlanations) and rule extraction with gradient-boosted trees to balance performance improvements, cost reductions, and grid stability. Unlike conventional black-box or purely optimization-based methods like Model Predictive Control (MPC), this approach provides explicit, quantifiable feature attributions for every operational decision. Simulation results from various types of operation demonstrate that the proposed model improves the utilization of energy, reduces peak demand, and increases transparency of decision making. The inclusion of the explainable AI aspect will enhance the level of confidence operators will place on the model and improve regulatory compliance, and therefore, this approach is a good candidate for use in future smart microgrid deployments.