Research on Virtual Power Plant Load Forecasting Based on RNN-LSTM Hybrid Model

As a crucial constituent of the future energy system, the load forecasting capacity of virtual power plants exerts a decisive influence on their economic efficiency, reliability, and sustainability. Through precise load forecasting, virtual power plants can not only maximize their benefit but also facilitate the accommodation of renewable energy and expedite the transformation of the energy structure. Hence, this paper employs RNN-LSTM hybrid model to predict the load of virtual power plants for the subsequent seven days. In the MATLAB software, an RNN-LSTM load forecasting model is constructed with time, maximum temperature, minimum temperature, average temperature, rainfall, and humidity as input variables and load as the output variable. The accuracy of the prediction results is assessed using RMSE, MAE, and MAPE indicators. Simultaneously, traditional BP and RNN algorithms are adopted for prediction comparison. The experimental results demonstrate that the RNN-LSTM hybrid model enhances the prediction accuracy by 7.68% and 0.98% respectively compared with the BP and RNN algorithms.