Research Of TSI Images Based Short-term Solar Irradiance Forecasting

Solar energy,as a kind of green energy,has huge advantages over the traditional fossil fuel resources.With the consumption of the traditional fossil fuel,it has been playing a more and more important role in supplying electricity resources for mankind.Due to the variability of the solar energy resource at the ground level,there is a need for the solar farms to forecast the solar irradiance in order to better manage and integrate the power output into the power grid.Some relative research of TSI images based short-term solar irradiance forecasting has been studied in this thesis.Based on the work of cloud detection,cloud motion estimation and irradiance prediction,a TSI images based system for short-term solar irradiance forecasting is built on MATLAB platform.This system includes various algorithm modules which can forecast the irradiance with the TSI images,which can forecast the online solar irradiance,and the system has convenient and intuitive interfaces for easier using.In this thesis,the modules of the system and the corresponding algorithms are described in detail and the performance of each module is demonstrated.In this thesis,we propose a method to predict the coverage ratio in circumsolar region using the conditional random field to overcome the shortcomings of the coverage prediction method using the cloud coverage ratio in the virtual circumsolar region.The method based on conditional random field can consider the information of multiple TSI images,which enriches the features to make the coverage ratio prediction in circumsolar region.The transfer functions in the conditional stochastic production can make a relation between the coverage ratios in the circumsolar region at two adjacent moments.It integrates the context information of the coverage ratios in the prediction of the coverage ratio in circumsolar region,thus avoiding predicting the coverage ratio at each moment independently.The experimental results show that the coverage ratio prediction method based on the conditional random field has higher prediction accuracy than the method that uses the coverage ratio virtual circumsolar region.A novel method of irradiance prediction based on historical cloud pattern matching is proposed in this thesis to overcome the shortcomings of existing methods of irradiance prediction.Through the matching of cloud historical patterns,the algorithm can select those measured irradiance for forecasting the short-term future irradiance.Thus it can establish a relationship between the predicted irradiance and the recent historical measured irradiance.This method can reduce the prediction error caused by the influence of the other atmospheric particles on the irradiance which can not be considered in the existing methods.Experiments show that the prediction method based on cloud historical pattern matching has better forecast accuracy than the existing methods of irradiance prediction.