Optimal control strategy based on differential evolution algorithm for seamless transition between islanded and grid-connected operation modes in microgrid clusters

Abstract

The increasing integration of renewable energy sources and distributed energy resources is accelerating the transformation of traditional power systems into smart grids. This transformation relies heavily on the deployment of microgrids (MGs), which offer enhanced flexibility, resilience, and sustainability. However, ensuring the stable and efficient synchronization of MGs, especially during transitions between islanded and grid-connected modes, remains a critical and unresolved challenge. This challenge is further amplified in MG clusters (MGCs), where coordinated operation is essential to maintain power quality and system reliability. Driven by the need to address this challenge, this study proposes a real-time synchronization control method to enhance the dynamic performance and operational reliability of MGCs. The main objective is to design and validate an optimal control strategy capable of minimizing frequency deviations and improving power sharing during the synchronization process. To achieve this, an optimal seamless synchronization control based on a differential evolution (DE) algorithm is developed. This controller optimizes a frequency error control objective function in real time and is tested on an MGC architecture combining grid-forming (GFM) and grid-feeding (GFD) inverters. This work addresses the lack of robust, fast, and quantifiable synchronization methods for hybrid inverter-based MGCs, a gap that this study aims to fill. The proposed method enables accurate optimization of the transition between islanded and grid-connected modes. Simulation results demonstrate substantial performance gains: a 66.33% reduction in the ITSE compared to a synchronization strategy based on the fmincon optimization algorithm, and a 37.91% improvement over a conventional approach that adjusts synchronization by modifying the droop control coefficients. Furthermore, a comparative analysis with the particle swarm optimization (PSO) algorithm demonstrated that the DE-based approach reduces computation time by 20.29%, highlighting its superior efficiency and suitability for real-time embedded implementation. A sensitivity analysis involving 500 different scenarios, including evaluations under fault conditions, confirms the robustness of the approach. The average ITSE for these 500 simulations was 0.2459, with a standard deviation of 0.041, demonstrating consistent and reliable performance under varying load conditions. Moreover, a second sensitivity analysis, conducted over 250 simulations, identified the optimal DE parameters, enabling the selection of an effective combination of population size, mutation factor, and crossover rate. Finally, experimental validation using a Hardware-in-the-Loop setup, with an OPAL-RT4512 unit and a dSPACE MicroLabBox, verifies the effectiveness and real-time performance of the proposed control strategy.