On Analysis And Prediction Of Power Load Characteristics Of Hotan Xinjiang

Load characteristic analysis and prediction research is an important foundationfor the China’s electric power system planning, production, scheduling, safe operationof the work, and is also one of the important references of corresponding powersupply policy. With the continuous development of our country’s economy, theelectric power market and the regional grid load have a change trend of differentproportion. These different levels of changes bring impact to the electric powersystem, electric power marketing and operating plan. Therefore, in order to achievesecurity, high quality, economics of power grid, guarantee the power balance betweensupply and demand, and provide the high quality service for power users, we mustcontrol and track the power grid load characteristics persistently by combining theelectric power system load characteristic analysis and prediction research work.This thesis took Hotan Xinjiang power grid as an example. Firstly, preliminaryanalysis of the region’s power grid load characteristic was carried on by traditionalload index analysis method. Then, under the environment of SPSS, the paper hadclustering statistics of the grid load characteristics of different regions in Xinjiang byutilizing K-mean clustering and systemic clustering analysis and then came to thespecific conclusions by comparative analysis on clustering results which had made bytwo kinds of previous clustering method. Finally, in the environment of MATLAB,prediction model of historical load data was built by using the relevance vectormachine algorithm (RVM) and model validation was carried out and then madeevaluation analysis on prediction result by using prediction accuracy indicators again.At last, the relevance vector machine (RVM) and support vector machine (SVM) werecomparatively analyzed through examples to make sure whether the relevance vectormachine can be applied to the grid forecasting work of Hotan. The research resultsshow that the precision of the relevance vector machine (RVM) is higher than supportvector machine (SVM) and which reached the demand of our country’s prediction accuracy so the method of the relevance vector machine (RVM) can be applied to thegrid forecasting work of Hotan.The research achievements of this paper can provide some theoretical basis andmethods to the extension work of the xinjiang regional power grid load characteristicsand intelligent load forecasting method in the future.