An efficient safety prediction and control strategy using fuzzy neural network architecture search in islanded microgrids

Abstract

In islanded microgrids, voltage and frequency deviations are critical indicators of system safety. Effectively managing these deviations to achieve a dynamic balance between generation and consumption under load and power fluctuations is essential for maintaining stable system operation. This paper proposes a voltage-frequency control scheme based on an accelerated proportional-integral-derivative (PID) controller, with an adaptive neuro-fuzzy inference system (ANFIS) optimized via neural architecture search (NAS) as a preventive control strategy, tailored to the data distribution characteristics observed in microgrid applications. This approach leverages a hybrid differential evolution and particle swarm optimization (DEPSO) algorithm combined with a novel supernet weight sharing mechanism, enhancing the efficiency of ANFIS architecture search and reducing dependency on expert knowledge. Through dynamic evaluation and control, this method aims to improve the safety and stability of islanded microgrids under load and power disturbances. Simulation results demonstrate that the proposed method shows significant advantages in microgrid safety prediction and control, quickly restoring system stability and ensuring the secure operation of the microgrid in the face of load and power fluctuations.