Application And Improvement Of Support Vector Machines In Power System Short-Term Load Forecasting

Power system load forecasting is an important factor of energy management system. According to the predicted loads, electricity purchase and the operation mode could be arranged reasonably. Short-term load forecasting, which aims at predicting electric loads for a period of minutes, hours, days, or weeks, plays a significant role in the power network's economic and stable running. With the rise of deregulation and free competition of the electric power industry all around the world, short-term load forecasting becomes more important than ever before.Generally, there are two kinds of methodologies for load forecasting. One is in a traditional way, represented by time series, another one is termed as new artificial intelligence method, represented by the artificial neural network. Basically, the traditional way may involve method of time series, multi-linear regression and the Fourier expansion, etc., while the artificial intelligence method mainly involves expert system method fuzzy logic method fuzzy neural model and artificial neural network method, etc. The forecasting effect of non-linear regression and time series is in effect when electric network works well and little change occurred in the production and weather, however, without considering some factors affecting the load, such as holidays and weather. Hence, the forecasting precision will be affected by sudden change of these above factors, neural network and fuzzy theory take into account of the indefinite factors affecting the load, but still have not solve the designing problem of network structure completely yet and need a long time of training as well.Support vector machine is a machine learning algorithm of the new era based on statistics theory, which equals to solving a quadratic programming problem in the principle of minimum structural risk. This algorithm features strong forecasting ability, global optimization and fast speed of approaching, etc. After the introduction of support vector machine algorithm, simulations was carried out compared with short-term load forecasting based on BP neural network. Considering the effect of parameters on forecast precision and the generalization ability, Grid-search method was proposed for parameter optimization. In order to lower the complexity of the computing and enhance the speed of solving, this paper further adopts the least squares support vector machine to build up a model...