Research On Short-Term Electric Load Forecasting Based On Least Squares Support Vector Machine

Load forecasting is one of the key tasks for scheduling and planning departments of power system and the accurate load forecasting is very important to guarantee the power system security, stability and economic operation, as well as increase the electricity sector under the economic and social benefits in the electricity market environment. In this paper, the relevant basis theories of load forecasting are introduced and existing short-term load forecasting methods are summed up. some important concepts and the main idea of statistical learning theory(SLT) and a new generic learning algorithm - support vector machine(SVM) in the theoretical framework of statistical learning theory are introduced, and a short-term load forecasting model is established based on least squares support vector machine(LS-SVM), which is an expansion of the standard SVM.LS-SVM's model parameters have an important impact on learning ability, which is not yet determined by an unified and effective choice method. In this paper, a choice method of model parameters based on similar days and ant colony algorithm is proposed, that is, according to the features of different effect for load with different effect factors, some similar days are separated into training samples and test samples with an improved grey relational degree, and then a new global search algorithm -ant colony algorithm is implemented to optimize parameters. The practical example shows that it's more reasonable and enhances the generalization ability of the model using the method to choose parameters.An identification method of abnormal datas based on statistical probability is introduced, and vertical and horizontal amendments are also given for the abnormal datas. An analysis is given on the effect factors of short-term load forecasting, and in particular weather conditions and the accumulated effects of weather conditions with the actual load datas and weather datas of a local power network. By using "accumulated threshold function" of different weather conditions to quantify the effect, the accumulated effects of weather conditions can be described with the changes of numerical value.On the basis of related effect factors, a practical example is given using the LS-SVM's model which is constructed by the former choice method of model parameters, and the results show the effectiveness of the model in this paper.