A data interpolation method for missing irradiance data of photovoltaic power station

As the installed capacity of photovoltaic systems increases year by year in the world, PV energy has become an important energy source. For the output power of photovoltaic power station, irradiance is the most important factor. In order to solve the problem of lack of irradiance data in some power plants, the data are classified according to weather by decision tree classification method. Under different weather categories, according to current time, power and ambient temperature, different neural network parameters are used to fit irradiance sequence and interpolate the irradiance value. The results show that the average absolute error and root-mean-square error of irradiance data interpolation are 9.12% and 11.57% respectively, which are better than ANN, ARMA, SVM and Weighted KNN.