Neuron count impact on NNTS-based energy management in hybrid-powered renewable microgrids

Abstract Efficient energy-storage management is critical for enhancing the reliability and sustainability of hybrid microgrid systems. This study examines the influence of neuron number in a Neural Network Time Series (NNTS) model on prediction quality and control performance within a hybrid energy-storage framework. The objective is to evaluate how different network architectures affect forecasting accuracy using MATLAB. Simulations were performed under variable loading and system disturbances. The model trained with the Levenberg–Marquardt algorithm achieved its lowest validation error 0.01112 when configured with 60 neurons and the highest correlation coefficient 98.093%, indicating strong consistency between predicted and actual system responses. These results confirm that neuron selection has a significant impact on NNTS predictive capability and overall energy-management efficiency. The findings,indicates that a network with 60 neurons attains superior performance across all evaluation metrics when training the NNTS model with the Levenberg–Marquardt algorithm, provide practical guidance for optimizing intelligent control strategies in hybrid microgrids and contribute to improving the integration and utilization of renewable energy sources in advanced engineering applications.