Short Term Prediction Of Power Generation For Photovoltaic Power Station

Photovoltaic power generation is an emerging industry,which power generation technology has been attracted much attention.But because the output of photovoltaic power generation system is affected by the solar radiation intensity and weather factors.In fact the photovoltaic power generation system has a great instability in the output when the photovoltaic power generation is a non-stationary process with a certain immediately.It is this nature that will cause the impact of photovoltaic power generation into the grid after the impact of the entire power grid.When to make what kind of grid scheduling is to reduce the impact of the key.Therefore,accurate prediction of photovoltaic power generation has become a research topic for many scholars inland and abroad.This study makes an in-depth study on the short term prediction of power generation of photovoltaic power station.Firstly,the power generation data of a photovoltaic power plant inverter is obtained and analyzed.It is pointed out that the weather type and temperature influence on the photovoltaic power generation.And,according to the data,making correlation analysis to the weather type and temperature is to get the correlation coefficient.Weather types are positively correlated with power generation,and are highly correlated.This paper puts forward the concept of the weather type mapping to the weather class index.The temperature is negatively related to the power generation,and the maximum temperature,minimum temperature and average temperature are compared,and the maximum temperature and the minimum temperature are taken as the influencing factor.Then,four prediction models of PLS,RF,SVM and SAA optimizing SVM are established.Four models are used to predict all the samples.The PLS prediction model,sunny,cloudy,cloudy,cloudy,rain five weather forecast the average accuracy rate are 94.2%,86.3%,81.6%,81.7%,73.6%.The RF prediction model,sunny,cloudy,cloudy,cloudy,rain five weather forecast the average accuracy rate were 93.5%,86.1%,82.3%,83.2%,75.3%.The SVM prediction model,sunny,cloudy,cloudy,cloudy,rain five weather forecast the average accuracy rate are 94.6%,87.9%,84.3%,85.7%,75.9%.The SAA optimizing SVM prediction model,sunny,cloudy,cloudy,cloudy,rain five weather forecast the average accuracy rate are 94.8%,90.8%,86.7%,87.4%,79.1%?PLS model belongs to multiple regression model,the model structure is relatively simple,and the program is easy to operate.RF model,SVM model,and SAA optimizing SVM model have machine learning ability.No matter what kind of weather types,SAA optimizing SVMmodel is relatively higher than the other two kinds of machine learning models which has higher accuracy,The second is the SVM model.The prediction model with machine learning,in cloudy,cloudy,cloudy to sunny weather fluctuations,has certain anti-interference ability,but it requires a certain program running time.In practice,different forecasting models are chosen according to different demands.