Generative Adversarial Network and CNN-LSTM Based Short-Term Power Load Forecasting

In view of the increasing demand for the short-Term load forecasting accuracy of commercial buildings, in order to overcome the data shortage, so that the subsequent forecasting model can learn the characteristics of the original data set more fully, and extract the effective characteristics of the data better, this paper proposes a short-Term load forecasting method based on the hybrid model of TimeGAN generative adversarial network and CNN-LSTM Network. Firstly, the TimeGAN network is used to generate synthetic data to expand the scarce data set. Then, the CNN network is used to filter the input data and extract useful information, and the LSTM network is used to analyze and forecast the time series data. The proposed method calculates the power load data of a company in Shanghai for two months. Compared with CNNLSTM model and LSTM model which do not use synthetic data, the obtained short-Term power load forecasting results improve the forecasting accuracy.