Data Classification And Prediction Technology For Smart Grid Research

Power load classification and prediction of power system has been the power sector which can adjust the time-sharing electricity pricing,substation,the total load prediction,peak management and so on the basis of plan formulation.Good classification and prediction method can provide the correct data for electric power planning management,the change of effective grasp user load characteristic,for short,medium and long term power prediction has better precision,assisted the development of smart grid.Classification and medium-term power prediction for electric power load data,this paper carried out the following work:1)Classification and prediction of power load are reviewed in the first place,summarizes the present situation of classification and prediction of the current load.And detailed analysis on the characteristics of the classification and prediction,pointed out the advantages and problems of various methods,and gives the effectiveness index and the evaluation standard and so on.2)In view of the traditional k-means classification method are the defects existed in the initialization based on load characteristic difference method is proposed.Based on the improved method,combining the reality power grid electricity analyzed experiment data,proved that the method has better classification results and the feasibility.3)On the current situation of the load forecasting and Grey forecasting model(GM)in predicting the developing present situation analysis.Sequence generated from its own deficiencies,introduced to cyclical index,to improve.And accuracy test method is given,in the experiment with the original method is analyzed.4)On the ordinary Grey model,and the paper part,on the basis of improved sequence generation method,this paper puts forward the prediction model of hybridintelligent optimization background value.In view of the traditional methods of GM first generate the sequence to remove the seasonal periodicity;Then aiming at the shortcomings of the background value choice modeling,rebuild the adjustable parameter of background value formula;Then use Particle Swarm Optimization(Particle Swarm Optimization,PSO)of adjustable parameter Optimization.Under the improved to periodic sequence generation method,through the iterative optimization formula of background value of adjustable parameters to obtain the global optimal value,so as to get the optimized development coefficient and Grey model parameters,prediction model is established.We apply the model in actual data of the power grid,through comparing with the original method and other methods,the results show that the model can effectively improve the prediction accuracy.