Fast Iterative Filtering-Based Deep Belief Network for Accurate Short-term Electric Load Forecasting

Electric load forecasting plays a pivotal role in the electric power industry and smart grid management systems, offering valuable support to government policy formulations. However, this task is highly challenging due to the non-linearity of load series, its complexity, weather dependency, and dynamic consumer behavior. Hence, there is a pressing need for a more robust and accurate technique to facilitate efficient electric load forecasting. In this study, we propose a fast iterative filtering-based deep belief network methodology for precise short-term electric load forecasting. Our specific focus is on one-day ahead predictions, utilizing the historical power demand data available. The efficacy of the approach presented in this work is assessed on diverse datasets through a comprehensive comparative analysis, highlighting its superior performance compared to existing state-of-the-art techniques. The proposed approach demonstrated remarkable efficacy by achieving the lowest error percentage, outperforming all existing methods. Therefore, the proposed methodology shows promise for accurate short-term electric load forecasting.