Comparative Study On Short-term Forecasting Models Of Photovoltaic Module Temperature Based On ANN And SVM

As a key factor in modeling and assessing the performance of photovoltaic(PV)modules,the accuracy forecasting of PV module temperature is important for improving the accuracy of PV power generation prediction.At present,the domestic research on PV module temperature analysis is few,and so basic problems have not been solved,such as thermodynamic analysis of PV module temperature,factor identification of PV module temperature,the statistical forecasting model of PV module temperature.Thus,in depth analysis of the impact factors of PV module temperature and their relationship with PV module temperature,this paper directs at the statistical prediction models of PV module temperature in different weather types based on Artificial Neural Network(ANN)and Support Vector Machine(SVM).And the actual data of PV power station is simulated to compare the performance of different forecasting models,adding the lack of PV module temperature researches and providing a scientific basic for the improvement of PV power generation prediction,with theoretical significance and application value.The identification and confirmation of primary MIFs of PV module temperature are investigated as the first step of this research from the perspective of physical meaning and mathematical analysis about electrical performance and thermal characteristic of PV modules based on PV effect and heat transfer theory.The analysis results point that there are 12 impact factors for PV module temperature,in which the key impact factors are 3 meteorological impact factors: solar irradiance,ambient temperature and wind speed.Furthermore,due to the multi-coupling nonlinear relationship of PV power generation and meteorological impact factors,the quantitative description of the MIFs influence on PV module temperature is mathematically formulated as several indexes using mutual information theory to explore the specific impact degrees under four different typical weather statuses named general weather classes(GWCs).The quantitative research result indicate that the order of influence degree of three meteorological impact factors for PV module temperature is: ambient temperature,solar irradiance and wind speed,and the redundancy exists between them.Finally,based on ANN and SVM,the direct and step-wise PV module temperature prediction models are built under four GWCs.The input variables are the data of three meteorological impact factors after pretreatment: wavelet denoising,normalization and principal component analysis,and the output variable is the predicted values of PV module temperature.The prediction results are measured by Mean Absolute Percentage Error(MAPE)and Root Mean Square Error(RMSE).The results show that in four GWCs,all the forecasting models have a good performance,and step-wise forecasting models are more accurate than direct models,in which the step-wise forecasting models based on SVM have the best performance in the PV module temperature prediction aspect.Thus,the results indicate that the analysis of impact factors is significant to the PV module temperature change rule,and on the basic of that,the step-wise forecasting models based on SVM has been established to accurately predict the PV module temperature,providing a good foundation for further research of PV power prediction.