Study On Short - Term Load Forecasting Of Power System

Short-term load forecasting is an component of energy management system (EMS), and it is also an important part of power system operation scheduling. For power system operation、control and planning has a very important significance, the accuracy of Short-term load forecasting directly affects the security and economy of power system.This paper summarizes the background and status of the power system short-term load forecasting, simple analysis of the the existing traditional methods and intelligent methods, introduces the basic principles,requirements and evaluation method for forecast results of short-term load forecasting, especially deep analysis the inherent law and external characteristics of short-term load, according to the characteristics of short-term load forecasting propose three effective methods.First, according to the characteristics of short-term external loads, proposed the method that Short-term load forecasting for similar days based on PSO-SVM and daily feature vector, This method introduces the human body amenity indicator, considering the coupling effects of weather factors on short-term load,and composed the characteristic vector with the rest of dominant factors, apply the method which is used to calculate similarity degree to select similar day, and built up the PSO-SVM prediction model, this method solves the problem of knowledge redundancy of SVM prediction model, reducing the training sample. simulation results show that the final result is better than predicted results of predicted load forecasting competition winners.Secondly, according to the internal laws of short-term load,proposed the method that short-term load forecasting based on the discrete Frechet distance and LS-SVM,this method according to the characteristics of the load curve by the method that discrete Frechet distance discriminant the curve similarity to selected similar time period, and using historical load data of similar periods of historical load to establish LS-SVM prediction model. simulation results show that the method improve the prediction accuracy compared with the predictions of the standard LS-SVM. Finally, aiming at the problem that from the surface of total load data to establish prediction model can not concerned about the characteristics of the composition of load in the short-term load forecasting method based on discrete Frechet distance and LS-SVM, proposed the short-term load forecasting method based on EMD decomposition similar days date and LS-SVM,in order to obtain the IMF component of representatives load characteristics.using EEMD decomposition to resolve the similar data selected by discrete Frechet distance, according to the characteristics of each IMF to establish LS-SVM prediction model with different kernel functions to forecast each component separately, finally, the prediction result obtained by adding each component of the final result, after simulation, IMF component can better reflect the characteristics of the components of the load,and the method improve the prediction accuracy compared with the method short-term load forecasting based on the discrete Frechet distance and LS-SVM.