Research On Power Prediction Model Of PV Module Based On Parameter Identification

With the PV modules long-term outdoor operating, the performance of PV moduleswill be reduced. The performance degradation is a direct result of changes in the staticparameters of PV modules. The model of solar power prediction is a mathematicalmodel which established on the static parameters. The change of static parameter valueswould result in inaccurate forecasts of the prediction model. With the increase of servicetime inaccurate forecasts phenomenon will become more apparent. Therefore, it’snecessary to study the predictive model parameter identification method to identify thechanged of static parameter, so as to solve prediction model problem of inaccurateforecasts.In this paper, we launched the study to overcome the lack of current identificationmethod. From both the data of pretreatment and identification method of parameter, weanalyzed the reasons for the prediction error of the power prediction model based oncurrent identification algorithm. Based on the study of the data mining technologyfeatures, the article proposed the method of extraction characteristic data based onassociation rules analysis as a parameter identification data pre-processing program, anddesigned the parameter identification method based on genetic algorithm. The maincontents are as follows.First, according to the parameter identification requirements, in order to obtain thedata record which is caused by the static parameters changing, the article used theassociation rules analysis to select rule sets of which the output power in the data ofeach period stable changed while in similar weather conditions. Then depending on thevariation of the identification data, we selected the data records which meet certainvariation as input of the identification algorithm.Second, in the paper we designed the objective function based on current equationof component model, and the method of parameter identification using geneticalgorithms, to solve the problem of low recognition accuracy of the currentidentification algorithm.At last, According to the previous program which have been elaborated we lanched some simulation experiments. The results of experiments show that: the data extractionprocess characteristics based on association rules, effectively shielding the adverse dataof low frequency, and successfully extracted feature data which reflect changes in thestatic parameters. In the parameter identification process, we accurate identified thechange of PV modules static parameters in each period. Based on data fitting techniquewe used the historical recognition results to fit the current static parameters. Adjustingthe parameters of power prediction model to predict, we compared the predicted valueunder the conditions of adjustment parameters and the predicted value which underunadjusted parameter conditions. The compareing result show that parameteridentification scheme proposed effectively solved the problem of inaccurate powerprediction.