Short-term load forecasting method for integrated energy system based on ALIF-LSTM and multi-task learning

The output of renewable sources such as wind power and photovoltaics in the integrated energy system is volatile and intermittent, and accurate short-term load forecasting is beneficial to smooth the impact of renewable energy on system operation. Shortterm load prediction is the basis for the safe and stable operation of the integrated energy system, while the time series of multiple loads of the system is a typical non-stationary signal with strong randomness, which makes it difficult to make an accurate prediction. In order to improve the accuracy of the short-term load prediction model of the integrated energy system from the data level, a historical load data decomposition method based on ALIF is proposed, which decomposes the historical load sequence into multiple components with different frequency band modal functions. In order to solve the problem of long time series processing in predictive model training and coupling information mining among multiple loads in the system, a short-term load forecasting model for integrated energy system based on LSTM multi-task learning was proposed. Experimental results show that, compared with LSTM, ALIF-LSTM single-task learning, random forest, and LGBM, the proposed method can cope with the condition of severe load fluctuation, and has a higher prediction accuracy, which meets the requirements of safe and stable operation and control of the integrated energy system. © 2022, Solar Energy Periodical Office Co., Ltd. All right reserved.