Research on Photovoltaic Power Generation Output Forecasting Model Based on Wavelet SVM and Micrometeorological Influencing Factor Analysis

Abstract

With the continuous development of photovoltaic power generation technology, it occupies a more and more important position in the power grid. However, the fluctuation of photovoltaic power generation will impact the power system. How to predict the photovoltaic output efficiently and accurately has become the key to maintain the stability of the power system. The existing photovoltaic power prediction methods can achieve more accurate prediction results in sunny days, but there are large errors for the prediction under the weather conditions with strong randomness such as cloudy days, rainy and snowy days. In this paper, micro meteorological data has the characteristics of strong randomness, large amount of data and strong ability to describe the details of weather changes. Although these characteristics can reflect the impact of weather changes at different time points of a day on the output of photovoltaic panels, they can improve the prediction accuracy according to the characteristics of frequent weather changes and large volatility in cloudy days, rainy and snowy days But at the same time, along with the strong random variation, the meteorological factors affect each other and bring errors. In order to reduce the influence of the randomness and correlation of micro meteorological factors, this paper uses Pearson correlation coefficient and principal component analysis to find out the most influential micro meteorological factors according to Pearson correlation coefficient, and then combines the principal component analysis to reduce the dimension of micro meteorological factors, remove the correlation between variables, and bring the processed variables into the wavelet SVM prediction model. Experiments show that the prediction accuracy is improved successfully.