Intelligent demand-side energy management based IoT using multi-attention fusion residual convolutional neural network for day-ahead energy forecasting

This paper, Intelligent Smart Energy Management System with Internet of Things (IoT) Using Multi-Attention Fusion Residual Convolutional Neural Network (SEMS-IOT-MFRCNN) is proposed. Initially, input data is gathered from Solar Radiation and Meteorological Dataset. The Bayesian Boundary Trend Filtering (BBTF) is used to find missing values and normalizing data. Then pre-processed data are given to Lotus Effect Optimization Algorithm (LEOA) for feature selection. Then the selected features are given to Multi-Attention Fusion Residual Convolutional Neural Networks (MFRCNN) to predict the solar energy forecasting. Generally MFRCNN doesn't express adapting optimization methods to define optimal parameters to predict the solar energy forecasting. Hence, the Sea-Horse optimizer Algorithm (SHOA) is used to optimize MFRCNN which accurately predict the solar energy forecasting. The proposed SEMS-IOT-MFRCNN method demonstrates superior performance, achieving an accuracy of 98 %, compared to 85 % for Particle Swarm Optimization Algorithm (PSOA), 75 % for Binary Particle Swarm Optimization Algorithm (BPSO), and 70 % for Iterative Optimization Algorithm (IOA). Also, the proposed method achieves the lowest cost of 1250 cents and the lowest latency of 2 s, significantly outperforming existing techniques. These findings validate the performance of SEMS-IOT-MFRCNN in optimizing energy management and forecasting accuracy for sustainable energy solutions.