Research On Short-term Forecast Of PV Output Based On Cluster Analysis Of Meteorological Factors

At present,the deepening industrialization process in our country and increasing fossil energy consumption make the use of renewable energy become a trend.As a kind of energy from the sun,solar energy is recognized as the most competitive renewable energy for its characteristics of being inexhaustible,green and non-polluting.In recent years,photovoltaic grid-connected capacity is gradually increasing.However,the output of photovoltaic power plants is affected by a variety of meteorological factors,which makes the photovoltaic output appear obvious periodicity and uncertainty.The meteorological factors in the atmosphere are rapidly changing,resulting in instantaneous or short-time volatility and randomness of PV output,which poses great challenges to the day-ahead optimal dispatching and safe operation of power system.Therefore,it is extremely important for the accurate short-term prediction of PV output.In this paper,the characteristics of periodic,intermittent,stochastic and volatility of photovoltaic output are analyzed qualitatively and quantitatively.Then,the relationship between meteorological factors(solar irradiance,temperature,atmospheric aerosol,relative humidity)and photovoltaic output was discussed separately,and the correlation coefficient between photovoltaic output and numerous meteorological factors was calculated quantitatively by Pearson correlation coefficient.Through the analysis of the photovoltaic output daily curve line chart,it is found that the photovoltaic output has obvious typical daily characteristics,that is,there are great differences between the photovoltaic output daily curves of different weather types,and this difference will impact on short-term prediction of photovoltaic output negatively.Therefore,according to the calculation results of correlation coefficient,strong meteorological factors are selected as the input factors of the improved Fuzzy C mean clustering(FCM)algorithm,and the original data of PV power output are clustered according to weather types and clustered into three data sets: sunny,cloudy,and overcast,which provides a data foundation for establishing a short-term forecast model for photovoltaic output.Through the clustering of meteorological factors,the photovoltaic output daily data is divided into three types of data sets: sunny,cloudy,and rainy.For each type of data set,this paper establishes a short-term point prediction and probability prediction model of photovoltaic output.In this paper,the related theories of BP neural network algorithm,genetic membrane optimization algorithm,neural network quantile regression model and kernel density estimation method are expounded in detail,and the parameters of the model and the modeling process are discussed in detail;For different types of data sets,a combined prediction model based on genetic membrane optimization BP neural network and a probability prediction model based on neural network quantile regression and kernel density estimation were constructed.Finally,two kinds of prediction models are simulated through the historical data and meteorological data of the photovoltaic power plant,and the results are evaluated to verify the validity and rationality of the proposed prediction method.