Entropy-fused Dual-weight Neural Ensemble for Shortterm Load Forecasting

With the growing demand for accurate short-Term load forecasting (STLF), efficient prediction of power system load is crucial for optimal scheduling and resource management. This paper proposes a multi-model ensemble approach combining RNN, LSTM, and MLP for STLF. By incorporating variance-covariance weighting and grey-relational analysis, the model leverages the strengths of each base model and addresses their individual limitations. The method uses dual-weight fusion, where variance-covariance helps control high-variance periods, and grey-relational analysis improves phase alignment between predicted and actual load profiles. The experimental results show that the proposed ensemble model outperforms individual models in terms of MAE, RMSE, and MAPE, demonstrating its ability to consistently improve forecasting accuracy. Moreover, the ensemble model maintains robust performance under various conditions, such as holidays and weather anomalies, with improved adaptability to different load patterns. The method also proves computationally efficient, making it suitable for real-Time forecasting and online updates. The proposed method provides a significant advancement in achieving accurate and stable short-Term load forecasting for modern power systems.