Font Size: a A A

Study On Short-term Load Forecasting Based On Weighted LS-SVM

Posted on:2012-03-29Degree:MasterType:Thesis
Country:ChinaCandidate:B BaiFull Text:PDF
GTID:2132330332986451Subject:Power system and its automation
Abstract/Summary:PDF Full Text Request
Short-term load forecasting is so random and unpredictable that the prediction accuracy is difficult to be significantly improved. Consequently, it is a very complex task. The traditional forecasting methods often consider it too simple, which makes many significant impacts on short-term load factor that does not play its due role. In recent years, with the development of artificial intelligence, machine learning language has become the focus of research personnel of load forecasting. Scholars have tried all kinds of intelligent language for short-term load forecasting, least squares support vector machine (LS-SVM) is one bright spot. With the deepening of its research, LS-SVM is more and more favored by experts.In the article, based on comprehensive understanding for the work of short-term load forecasting, chooses the method of least squares support vector machine to research. Taking into account the modern short-term load forecasting will involve a variety of external factors, such as weather, this paper considers the weather type, maximum and minimum temperature and other factors for the specificity of models. Due to various considerations, resulting in increased complexity of the model, and the LS-SVM model is defective itself - loss of robustness, so the paper will take the weight coefficient by use of robust estimation of error of the model, and then construct a weighted least squares support vector machine (WLS-SVM) model. In addition, the model also chose to use the Bayesian evidence theory optimization and radial basis function (RBF) as the kernel function of model.In this paper, through collect and analyze a large number of measured load data, we found a particular phenomenon of load - "rest day delay phenomenon", and establish an adaptive prediction model for it. Taking into account "near the far smaller" principle of load forecast, and in order to take full advantage of Monday loads, the load of abnormal Monday will be modified. Finally, we use the WLS-SVM method under the Bayesian evidence researched in this paper to predict the actual load. Through the simulation of Matlab programming, it can be seen that the method used in this paper has good predictive results, to meet the accuracy of short-term load prediction.
Keywords/Search Tags:Short-term Load, LS-SVM, Robustness, Bayesian Theory of Evidence, Prediction Accuracy
PDF Full Text Request
Related items