Multi-load short-term prediction of an integrated energy system based on GAN-LSTM

Abstract An integrated energy system’s load consists of diverse forms, including electric load, cooling heat and more. Due to its strong randomness volatility, traditional forecasting methods are not suitable for the system. To address issue short-term multiple loads system, proposed paper utilizes generating adversarial networks (GAN). The method analyzes characteristics comprehensive multi-load combines meteorological factors with historical data form input dataset prediction model. further improve accuracy loads, model employs a generator discriminator constructed based on cyclic neural network long- memory (LSTM). example results demonstrate that is effective in predicting significantly better than methods.