Short-Term Electrical Load Forecasting Through Optimally Configured Long Short-Term Memory

Short-term electrical load forecasting plays a pivotal role in modern energy systems, addressing the need for accurate predictions of electricity demand within a time frame ranging from a few hours to a few days. The implications of inaccurate predictions extend beyond operational challenges to potential economic and environmental consequences, emphasizing the critical role that short-term electrical load forecasting plays in the modern energy landscape. The purpose of this research is to address the aforementioned consequences by developing an optimally configured Long Short-Term Memory (LSTM) model for predicting short-term electrical load forecasting in Tamil Nadu, specifically focusing on India's Villupuram region. While LSTM models are recognized for their overall effectiveness, their performance in short-term electrical load forecasting necessitates a tailored approach. Hyperparameter optimization is the appropriate choice for configuring the LSTM model for short-term electrical load forecasting. The manual or trial-and-error process in hyperparameter tuning is time-consuming and complex to compute. To address this, the research integrates the Cauchy-distributed Harris Hawks Optimization (Cd-HHO) approach for the optimal configuration of the LSTM model. The optimally configured LSTM through Cd-HHO consistently achieves lower Mean Squared Error (MSE) compared to other state-of-the-art methods, which is ۰.۷۲۲۵ in the ۲۰۱۷ database, ۰.۹۷۴ in the ۲۰۱۸ database, and ۰.۱۱۶ in the ۲۰۱۹ database.

Keywords:

short-term load forecasting , Long Short-Term Memory , Cauchy-distributed Harris Hawks Optimization , hyperparameters tuning , Uncertainties in weather forecasts , Power system management , Villupuram region