LSTM-RBF Short-Term Load Forecasting Method Based on Gaussian Similar Days

With the development of social economy and power system, the demand for load forecasting accuracy in the whole power system is increasing. Accurate power load forecasting is directly related to the sound, stable and healthy development of social economy. At the same time, with the improvement of people’s living conditions and the transformation of the production and use mode of electric power energy, the proportion of refrigeration and heating equipment in the load has steadily increased. Various factors such as the rapid development of distributed photovoltaic for spontaneous self use at the user side have comprehensively affected the change of daily load curve, increasing the error of load forecasting. In short-term load forecasting methods, similar day algorithm is widely used in short-term power load forecasting due to its simplicity and ease of operation. Due to the characteristics of the load itself, it can be directly used as load forecasting through similar days. Therefore, this paper starts from the similar day method, uses the Gaussian index to improve the selection method of similar days, and uses the combined neural network to forecast the day ahead electric quantity. Through the difference between the predicted electric quantity and the similar daily electric quantity, the results of the similar day method are corrected, and a more accurate day-ahead load forecast is obtained.