Research On Adaptive Neural Network For Solar Power Prediction Under Abnormal Weather

With global warming and depletion of fossil fuels,renewable energy such as solar power has been rapidly developed in recent decades.Solar energy is an inexhaustible and clean source of energy.However,the efficiency of solar power is easily influenced by meteorological factors,so it is random and volatile.The instability will put some pressure on the grid system after the PV power station is connected to the network.Accurately forecasting the power generation of photovoltaic power plants will help to optimize the scheduling of photovoltaic power generation and improve the safety level of power grid.PV power forecasting has high prediction accuracy under clear sky,but low accuracy in abnormal weather.This paper presents a self-adaptive neural network model.According to the different types of weather,the model of photovoltaic power forecasting is selected automatically to predict.In the premise of ensuring the accuracy of clear sky prediction,the prediction accuracy of abnormal weather such as haze and cloudy weather is improved.The main contents of the dissertation are as follows:(1)Analyze PV output characteristics and the influence of meteorological factors and PV output.Determine the temperature,humidity,radiation and AQI on the impact of photovoltaic output.(2)Use PSO-BP neural network to predict the PV power and analyze the prediction results.(3)Establish self-adaptive neural network model.Use SVM to determine the type of weather,and the classification is based on 4 hours of the difference between the theoretical irradiance and the actual irradiance before the current time.Establish corresponding BP neural network model for different types of weather(4)Developed photovoltaic power forecasting system.Integrate the training model into the system and display the forecast results through the interface.