Photovoltaic Power Prediction And System Implementation Based On Stepwise Cluster Analysis

With the development of social economy,the renewable energy demand of various countries in the world is increasing day by day.As the critical part of green energy,the solar energy resource has the clean and efficient characteristics,leading to the rapid development of photovoltaic power generation in recent years.However,the electricity generation of photovoltaic plant is affected by various factors such as meteorological condition and device performance,which has shown the volatility and intermittency.The accurate prediction of photovoltaic power generation is of great significance to the safe operation and dispatching planning of the power grid.Currently,the intelligent algorithms such as neural network and support vector machine have the certain limitations,where they are incapable of meeting in-depth analysis requirement of various influencing factors of photovoltaic power prediction.Moreover,the classification of weather types is mainly based on a single weather type,and the function design of photovoltaic power prediction system is imperfect for the integration of weather and power forecast,decision support and other functions.The stepwise clustering analysis method(SCA)can present the complex relationship between the input layer and the output layer through the clustering tree,and conduct in-depth analysis of the influencing factors to solve the random and discrete problems among variables.Therefore,in this dissertation,short-term output prediction model of photovoltaic plant based on the stepwise cluster analysis(SCA)method was established and the design and development of prediction system was carried out.The main research contents are as follows:(1)The current situation of photovoltaic power forecasting method and forecasting system at worldwide is summarized and analyzed.The key research methods are determined,including the photovoltaic output forecasting method based on SCA technology and the photovoltaic power forecasting system development by aid of Java programming language.(2)The main process of SCA method,the principle of photovoltaic power generation and the composition of photovoltaic power generation system are introduced firstly.Then,the correlation analysis between the electricity generation and meteorological factors such as solar irradiance,temperature and humidity are accomplished.Finally,the factors with the high correlation,including total radiation irradiance,wind speed,temperature and humidity,are selected as the input variables of the prediction model.(3)The prediction model of photovoltaic power generation in different seasons and weather types was established.The results demonstrated that the prediction effect of spring was the best,with the lowest mean absolute percentage error(MAPE)and mean square error(MSE)and the highest R2 value,being 7.97%,3.32 and 0.95,respectively;followed by summer and autumn,and the worst in winter.As for the weather type,the effect of single weather type is better than that of compound weather type,where the mean value of MAPE and MSE is 3.67%and 3.82 lower than those of compound weather type,respectively.With regard to single weather type,the fitting degree of sunny weather is the best.Among the compound weather types,the fit degree of other weather types is the best.Compared with LSTM and RF methods,the SCA model has the highest prediction accuracy in four seasons,single weather type and compound weather type,with the prediction accuracy increased by 11.13%,9.51%and 8.26%,respectively.(4)The photovoltaic power prediction system was designed and developed by aid of Java programming language,where the system architecture,database and function modules were designed.Moreover,the prediction algorithm code written in R language was successfully packaged and invoked in this system,which realized the functions such as the integration of meteorological factors and power forecast,statistical analysis,automatic operation and decision support.To sum up,in this dissertation,model of photovoltaic plant based on the stepwise cluster analysis(SCA)method was established for in-depth analysis of the influencing factors of PV power prediction,besides the traditional weather classification,complex conversion weather is also considered.,and combines the actual design of the power station to develop a prediction system that integrates meteorological and power prediction with decision support function.