Research On Short-term Load Forecasting In Yantai Power Grid Based On Support Vector Machine

Load forecasting is the basis for safe and economic operation of power system.For the current shortcomings of existing short-term load forecasting method, a newmodel of multi-output weighted least square support vector machine is proposed todo the short-term load forecasting efficiently, based on the selection of similar days.The main contents include:The improved transverse and longitudinalcomparison method is proposed torealize the bad data identification and the supplement of missing data. It can solve theproblem of the generation of bad data or the data missing, which can improve the dataquality by preprocessing of the load sample data.Day feature vector of the load is extractedfrom the effect size for daily loadlevels and trends. According to this, a new method based on unsupervised supportvector machine is proposed to select the similar days. And the modeling and solvingsteps are presented subsequently. The method can achieve the validity and rapidity ofthe similar days selection, and it further improves the accuracy of load forecasting.The model of multi-output weighted least square support vector machine ispresented to realize the short-term load forecasting. The research on effect size of theinput samples for the model, analysis and screening of the input elements, and thekernel functions selection is done, and the result is given in the thesis. Finally, acomplete solution for load forecasting is developed, and an example in Yantai powergrid is taken to verify the effectiveness of the proposed method.