Power Load Forecasting Based on Refined Meteorological Factors

Due to the problem of low resolution of current numerical weather forecasting, this paper proposes a power load forecasting model that combines support vector regression(SVR) and Bi-directional long short-term memory network(Bi-LSTM). SVR reduces data processing by focusing on key samples, while Bi-LSTM captures information from past and future feature data for efficient and accurate load forecasting. The proposed model was tested using numerical forecast data from European Centre or Medium-Range Weather Forecasts (ECMWF) in THORPEX Interactive Grand Global Ensemble(TIGGE) and historical load data from Shanghai, demonstrating its effectiveness through simulation and comparative analysis.