Short-term Photovoltaic Power Prediction Research Based On GA-BP-AdaBoost Model

With the continuous improvement of energy demand in countries around the world,countries are urgently looking for new alternative energy sources,photovoltaic power generation due to renewable,pollution-free,diverse applications,unlimited scale and capacity and other advantages in recent years has been rapidly developed,has become an important way to develop and utilize solar energy.However,because photovoltaic power has strong intermittent and random fluctuations,which will have a certain impact on the safe and stable operation of the power grid,it is of great significance to accurately predict the photovoltaic power generation power.At present,BP neural networks are widely used for output power prediction,but there are problems such as poor convergence and easy to fall into local minima,in view of this problem,this paper is committed to optimizing BP neural networks to achieve better prediction results.The main research contents are as follows:(1)This paper introduces the basic principles and components of photovoltaic cell power generation and the output power and output characteristics of the photovoltaic power generation system are then analyzed,choosing solar radiation rate,ambient temperature,relative humidity and pressure as characteristic variables,and classifies the existing weather types into sunny,cloudy and rainy weather types and different season types by K-means clustering.(2)the historical data of the photovoltaic power station is pre-processed,and then the data is monitored,corrected and standardized by the data.Through the comparative analysis of gray correlation degree and regression R~2value,the irradiance,humidity and temperature are selected as the main meteorological characteristics affecting photovoltaic power generation power for modeling,and the first 80%of the historical dataset is divided into training sets and the second 20%into test sets for the following model construction and prediction.(3)The structure and principle of BP neural network(Back Propagation Neural Network,BPNN)are elaborated,summarized and explained,and in the process of using BP model for prediction,it is found that BP neural network has shortcomings such as too many iterations and slow convergence speed,in order to solve this problem,it is proposed to optimize the weights and threshold size of BP neural network through genetic algorithm(GA),The prediction results of BP and GA-BP models and the average absolute error rate and root mean square error are compared and analyzed under different weather types and seasonal types.The results show that the GA-BP model can predict the photovoltaic power more accurately under different weather types and seasonal types.(4)In order to further improve the prediction accuracy and stability of the model on photovoltaic power generation,the BP neural network optimized by GA algorithm is selected as a weak predictor,and multiple GA-BP neural networks are optimized and fused by AdaBoost algorithm,and a GA-BP-AdaBoost strong predictor model composed of many GA-BP neural networks is established.In order to verify the model performance of AdaBoost after optimizing the GA-BP neural network,comparing it with the single BP model,GA-BP model and BP-AdaBoost model,the measured data of a photovoltaic power station in Jiangsu was selected for example analysis,and the results showed that GA-BP-AdaBoost had a better prediction effect in photovoltaic power prediction.