Medium And Long-term Load Forecasting Base On Support Vector Machine Of Phase Space Reconstructing

In power system, the term load forecasting has very important meaning, it is power planning, production and operation of the important work, is mainly for maximum load in unit time in society, the forecast. However with great power load forecasting accuracy of challenging, because of its history, load less data from countries or regions, economic, social, environmental uncertainty, etc. Yet precise load forecast and improve power system is helpful to improve the economic benefit and social benefit, can effectively guarantee the safety of power grid operation stability, effectively reduce the cost of electricity, fully satisfy demand, enhance the power supply reliability.First to study term load forecasting of power system research background and significance of both at home and abroad are expounded, the term load forecasting the study summarized and reviewed the basic principle of term load forecasting, analyzes the advantages and disadvantages of various methods and load forecasting error analysis, etc.Secondly, analyzed the characteristics of term load forecasting, using support vector regression algorithm for long-term load forecasting. Review the support vector machine (SVM) method, based on the basic principle of support vector machine forecasting model, and the load of regression design MATLAB procedures. And through the actual example analysis and other methods comparing the prediction result show that the forecasting model with the term load forecasting, the feasibility of this method was verified.Followed by using chaos theory is proposed, to support vector regression term load forecasting model of phase space reconstruction initial data processing. Improved load forecasting model of support vector machines, the input based on time sequence of phase space reconstruction will get better effect. Then, through example analysis, and standard support vector regression methods of comparative analysis and prediction results verified based on phase space reconstruction and support vector machine forecasting method is feasible.Finally in use in the theory of phase space reconstruction based on two kernel learning method based on support vector regression of power load forecasting method. Using the solution path algorithm, and example analysis verified based on the phase space reconstruction and two step kernel learning method support vector regression model of the load forecast, have very good prediction accuracy, and achieved satisfactory results, show that this method has its feasibility and validity.