 This paper proposes a new deep learning algorithm for short-term load forecasting, SCLF. The algorithm outperformed other AI algorithms commonly used in load forecasting, achieving a mean absolute percentage error, MAPE, of 9.77% in cumulative variation of root mean square error, CVR MSI, of 11.66%. This shows that the proposed algorithm has excellent forecasting accuracy. This article was authored by Ping Huankua and Chu Jihuang.