Research On Ultra-Short-Term Photovoltaic Power Prediction Considering Cloud Sheltering

Under the impact of external factors such as solar radiation and weather,the output power of the photovoltaic stations has notably randomness and volatility,thus exerting a bad influence on the safety,stability and economic operation of the power grid when executing grid connection.Therefore,accurate prediction of photovoltaic power can provide necessary support for grid scheduling and decision-making to reduce operating cost of power system effectively.At present,the researches on ultra-short-term photovoltaic power prediction at home and abroad are mainly consisted of weather nephogram modeling,fuzzy prediction using cloud amount information and other methods.Among these,modeling methods for nephogram generally describe the cloud features in a stereotyped way with higher costs.Moreover,predicting methods using cloud amount information usually obtain the expected value of cloud blocking through the process,which have a limited accuracy.In order to solve the above problems,in this paper,we present an ultra-short-term photovoltaic power prediction method considering cloud cover coefficient.This indirect method focuses on the influence of cloud sheltering on surface solar radiation:Firstly,we established a sunny-day surface solar radiation model based on the related data in clear weather;Then,we obtain the historical cloud cover coefficient data according to the output of this model and the historical surface solar radiation data,and consider the former as observation value,the latter as status value;And then,we model these data with a weighted averaged hidden Markov model to predict the future cloud cover coefficient and correct the result of surface solar radiation.Finally,we respectively launched prediction models to make prediction of photovoltaic power according to the cluster result of the cloud cover coefficient and made comparative analysis for the prediction result in the primitive characters situation and the compound features situation.The experimental results show that this method is applied for forecasting different types of weather,especially in cloudy conditions where it could reach a more ideal prediction performance.