Enhanced electrical and thermal energy storage systems performance in smart building using FLHNN and BWOA approach

Abstract

With the increasing integration of photovoltaic (PV) systems in smart residential buildings, the efficient management of electrical and thermal energy storage is becoming a critical challenge. This study introduces a novel hybrid technique combining Federated Learning on Heterogeneous Neural Networks (FLHNN) with the Binary Whale Optimization Algorithm (BWOA), termed the FLHNN-BWOA approach, to optimize the integration of photovoltaic systems and energy storage in smart residential buildings. The principal aim is to enhance PV self-consumption and lower the price of electricity through the optimal sizing of thermal storage systems (TSS) and battery storage systems (BSS). Unlike existing techniques, the FLHNN-BWOA approach offers a unique combination of accurate energy demand forecasting and optimized energy storage management, improving system performance and cost-efficiency. The proposed technique is implemented on the MATLAB platform and its performance is compared with existing methods like Artificial Neural Network (ANN), Feed forward Neural Network (FNN) and Light Neural Network (LNN). With the proposed approach, the optimal sizing of TSS and BSS reduces the electricity cost by 82.6 % and PV self-consumption is raised to 59.1 %. Furthermore, the study demonstrates the scalability of the approach, highlighting its potential for broader applications in various residential and commercial buildings. These findings highlight the potential of the FLHNN-BWOA technique in optimizing the sizing of energy storage systems, providing a highly efficient solution for enhancing energy self-sufficiency, sustainability and cost-effectiveness in smart residential buildings.