Adaptive Neuro-Fuzzy Integrated Forecasting–Control Architecture for High-Penetration Renewable Microgrids

Abstract

High penetration of renewable energy in microgrids introduces operational challenges due to the stochastic nature of solar and wind generation. This paper presents a unified Adaptive Neuro-Fuzzy Inference System (ANFIS) framework that integrates short-term renewable forecasting, real-time dispatch optimisation, and power-quality enhancement into a single control architecture. The forecasting module leverages environmental and demand data to predict photovoltaic and wind generation, while the dispatch module allocates power flows among renewables, battery storage, and the utility grid under state-of-charge and converter constraints. The power-quality layer regulates voltage and frequency while mitigating harmonic distortion at the point of common coupling. The framework is implemented in MATLAB/Simulink and evaluated using a minute-resolution synthetic dataset representing a 100-kW PV array, 50-kW wind turbine, and 200-kWh battery. Compared with linear regression forecasting and PI-based control, the proposed system reduces PV forecasting RMSE by 37.8%, decreases total harmonic distortion from 19.86% to 10.52%, and achieves faster voltage settling. Results demonstrate the advantage of coupling data-driven prediction with fuzzy rule-based adaptation for stable and efficient renewable microgrid operation.