Classification and hidden neuron count effect on renewable microgrid power management
Ismail Elabbassi, Mohamed Khala, Naima Elyanboiy, Omar Eloutassi, Youssef El Hassouani · 2025
The chapter presents a proposed energy management system using Neural Network Time Series model, with Levenberg-Marquardt (LM) training method, compared with the author's previous works, which incorporated classification methods. The study explores the effect of neuron count in a neural network time series model to enhance energy management, revealing that 10 neurons optimize validation performance, achieving a 96.2196% determination coefficient using the LM algorithm and with an error of 0.0379, outperforming all other simulations. The proposed method excelled in simulating energy systems, achieving 99.864% accuracy and surpassing previous benchmarks (99.747%-99.81%).The research underscores the neural network approach based on neuron count determined value to enhance energy storage management and promoting sustainable energy integration.