A Short-Term Forecasting Method Based on Time Series Filtering with Savitzky-Golay for Power Load Curve

The accurate prediction of the load curve is not only vital to the establishment of the generation scheduling, unit start-stop, and maintenance plan, but also has an important influence on ensuring the smooth operation of the generation side and consumption side, reducing the cost of power generation and improving the economic benefits. Based on this, a short-term power load curve forecasting method based on Savitzky-Golay time series filtering is constructed. Firstly, the relevant factors that affect the load are used as input components of the forecasting model, such as holidays, weather, etc. Secondly, the Savitzky-Golay filter is used to smooth the load sequence to weaken the adverse effects of local load fluctuations on the forecast. Finally, combining with the training and testing process of the Deep Neural Network (DNN) model, the training samples were constructed by using the load time series after filtering, holidays, weather and other data, so as to realize the prediction of the short-time power load curve in the next seven days. The prediction effect of the model is verified by the daily load data influencing factor data collected from a certain power plant. The simulation results show that the accuracy of the short-time power load curve prediction model based on the cooperation of Savitzky-Golay and DNN can reach 95%.