Enhancing Demand Response Scheduling in Smart Grids With Integrated Renewable Energy Sources PV and Wind Systems Using Hybrid Epistemic Neural Networks—Clouded Leopard Optimization Algorithm

Abstract

Demand response (DR) improves grid stability by enabling communication between the grid and consumers. However, managing residential load variability during DR events is challenging, especially in smart grids with renewable energy sources like wind and photovoltaic systems. This study aims to develop an advanced demand response scheduling strategy that optimizes electricity costs, reduces peak loads, and maintains user comfort. The primary goal is to enhance load demand prediction accuracy and optimize cost-efficient energy consumption in residential smart grids. A hybrid approach, the Epistemic Neural Network-Clouded Leopard Optimization Algorithm (ENN-CLOA) technique, is proposed. ENN is used for precise load demand forecasting, while CLOA optimizes electricity costs by dynamically adjusting energy consumption patterns. The method is implemented in MATLAB and compared with existing approaches, including artificial neural networks (ANN), deep neural networks (DNN), and recurrent neural networks (RNN). The ENN-CLOA technique achieves superior cost efficiency, with a minimum electricity cost of ¥10580, outperforming ANN (¥10870), RNN (¥10780), and DNN (¥10670). The proposed method also demonstrates lower error rates in load prediction and improves peak load management. The proposed technique enhances demand response performance by reducing electricity costs, mitigating peak loads, and ensuring better energy efficiency in smart grids.