Short-Term Electricity Load Forecasting Based on WOA-LSTM model

Abstract

Electricity load forecasting is a critical component in ensuring the safe, stable, and economical operation of power systems. However, the nonlinear and dynamic characteristics of load data impose limitations on the accuracy and generalization capabilities of traditional forecasting methods. To address this, this paper proposes a hybrid model (WOA-LSTM) based on the Whale Optimization Algorithm (WOA) and Long Short-Term Memory (LSTM) networks to achieve high-precision forecasting of ultra-short-term electricity loads. This model leverages WOA's global search capability to adaptively optimize LSTM's key hyperparameters, thereby enhancing the model's convergence speed and predictive performance. The LSTM network effectively captures the temporal dependency features of load sequences through its memory gate mechanism, enabling precise modeling of complex nonlinear trends. Experiments validate the model using power load data (15-minute sampling interval) from the 2016 Electrical Engineering Mathematical Modeling Competition, with data divided into training and testing sets. The results demonstrate that the WOA-LSTM model outperforms the single LSTM and other comparison models in metrics such as Root Mean Square Error (RMSE) and Coefficient of Determination (R²), achieving an improvement in prediction accuracy of approximately 10% to 12%. This model provides efficient and reliable decision support for power grid load scheduling and intelligent energy management, offering significant reference value for enhancing renewable energy utilization rates and achieving the dual carbon goals.