Research On Solar Radiation Forecasting Method Based On Data Mining

Under the background of accelerated global energy consumption and environmental pollution caused by traditional fossil fuels,the focus of energy strategy has gradually shifted to the development and utilization of clean energy.Solar energy is an important clean energy.As one of the utilization ways of solar energy,photovoltaic power generation has been increasing year by year in recent years.The power output of photovoltaic power generation has obvious intermittence and uncertainty.The integration of large-scale photovoltaic power plants into the grid will bring severe challenges to the stability and economic operation of the power system.Solar radiation intensity is the main factor that affects the power output of photovoltaic power generation.The forecasting of solar radiation intensity can indirectly realize the forecasting of photovoltaic power generation,which is helpful for power grid companies to make dispatching plan in advance and ensure the stable and economic operation of power system.This article focuses on the forecasting of solar radiation intensity.This paper analyzes the working principle and output characteristics of photovoltaic cells,establishes a mathematical model based on the equivalent circuit of photovoltaic cells and uses MATLAB/Simulink for simulation experiments to study the influence of solar radiation intensity and temperature on output current and power.A comprehensive analysis of the simulation experiment results shows that the solar radiation intensity is the main factor affecting the output of photovoltaic power generation,which proves the feasibility of forecasting the solar radiation intensity indirectly to realize the photovoltaic power forecasting.At the same time,the data mining technology is introduced,and the applicability of applying the data mining method to the field of solar radiation intensity forecasting is analyzed.Aiming at the short-term forecasting of solar radiation intensity,a short-term forecasting method based on support vector regression is proposed.Firstly,the change trend curve of the clearness index is calculated,and the cluster analysis of the curve is performed through the K-means algorithm to realize the weather classification.Then use the gradient boosting decision tree to judge the weather type of the forecast day.Combined with the corresponding weather types,the support vector regression model is established by using meteorological information to get the clearness index.Finally,the solar radiation intensity is calculated.It avoids the problem that the forecasting accuracy is relatively low due to the use of the same fitting model under different weather types.Simulation results show that the method achieves better forecasting results.Hourly forecasting of solar radiation intensity is very important for ultra-short-term power forecasting of photovoltaic power plants.Existing forecasting models have disadvantages such as high input dimensions and difficulty in optimizing model parameters.To this end,this paper proposes a time series forecasting method based on long and short-term memory networks in deep learning methods.First of all,the clearness index time series data after data preprocessing are transformed by empirical wavelet transform,the long and short-term memory network forecasting models are established for each component,and the mutual information values are calculated to select the input features of the model.Then the whale optimization algorithm is used to optimize the input dimensions and parameters of the model at the same time.The forecasting value of the clearness index is obtained after the empirical wavelet inverse transformation of the predicted results of each component,and then the predicted value of solar radiation intensity is calculated.Finally,the actual observation data of the two places are selected for simulation experiments,and the forecasting performance of the model under different weather conditions is verified to verify the effectiveness of the proposed method.