Comparative Multi-Objective Optimization Strategies for Intelligent Energy Management of Grid-Connected Microgrids Using BILSTM Forecasting

Abstract

This paper presents a comparative multi-objective optimization framework for intelligent energy management of a grid-connected microgrid integrating photovoltaic generation, battery energy storage, and bidirectional grid interaction. A Bidirectional Long Short-Term Memory (BiLSTM) neural network is employed to accurately forecast load demand and photovoltaic power generation using realistic meteorological and operational data. The predicted profiles are subsequently integrated into a multi-objective Energy Management System (EMS) formulated over a 24-hour scheduling horizon. The EMS optimization problem simultaneously minimizes electricity operating cost and battery utilization while satisfying battery State of Charge (SOC) limits and grid operational constraints. Three multi-objective optimization algorithms, namely Non-dominated Sorting Genetic Algorithm II (NSGA-II), Multi-Objective Particle Swarm Optimization (MOPSO), and Adaptive Multi-Objective Particle Swarm Optimization (AMOPSO), are comparatively evaluated for optimal battery scheduling. Simulation results demonstrate that the considered optimization algorithms exhibit different trade-offs between economic performance and battery utilization. MOPSO achieves the lowest electricity operating cost due to its strong exploitation capability, whereas NSGA-II provides improved Pareto front diversity and exploration performance. AMOPSO offers stable convergence behavior and balanced operating solutions. The comparative analysis highlights the strengths and limitations of each optimization strategy in the context of intelligent microgrid energy management. The obtained results confirm that combining BiLSTM forecasting with multi-objective optimization provides an effective framework for improving energy scheduling, battery management, and operational flexibility in grid-connected microgrids.