Research On Forecasting Model Of Renewable Energy Generations Considering Their Spatial And Temporal Characteristics

Global warming and the depletion of traditional energy have become the focus of attention for all countries.Promoting a high proportion of clean energy access has become an inevitable trend in the development of optimal allocation of global energy resources.The development of clean energy on a global scale requires extensive interconnection,development and allocation on the basis of global energy interconnection(GEI).The assessment of the development potential of clean energy provides support for global energy interconnection and collaborative planning.When only the meteorological information is available for the interconnected regional grid,it is urgent for grid planners to propose a widely applicable renewable energy power forecasting model to reconstruct the historical and future output of the interconnection regional grid with high spatial and temporal resolution as the intermittent power input to the grid-source collaborative planning model.In this thesis,a renewable energy output forecasting algorithm model based on complex principal component analysis(CPCA)is proposed,using wind and photovoltaic power reconstruction as the research object.The main achievements of the thesis are as follows.In this thesis,firstly,the spatial-temporal features of meteorological data are extracted from the gridded multidimensional meteorological information of the regional grid based on the feature extraction method of traditional principal component analysis(PCA).Considering the high correlation between meteorology and output,the complex original data matrix is constructed with meteorological data as the real part and output data as the imaginary part.The traditional principal component analysis is further extended to principal component analysis in the complex domain to extract the complex spatial-temporal feature matrix of meteorology and output.In addition,the time-series and distribution of wind power and photovoltaic generation are analyzed,and the complex spatial-temporal feature matrix is modified.Secondly,based on the correlation between the meteorological information of different regions,the applicability of the renewable energy output reconstruction model is addressed when the spatial scale of the known region is inconsistent with that of the region to be reconstructed.The evaluation system of power reconstruction error for renewable energy generation is established.In addition,considering the differences of renewable energy generation policies and grid planning in different countries or regions,the ARMA model is used to build the regional total installed capacity prediction model.Then the actual value of regional reconstruction output is obtained by multiplying the high spatial-temporal resolution reconstruction output per unit value and the regional total installed capacity.Besides,considering that the abnormal values in the historical measured data of wind farms and photovoltaic power plants will affect the accuracy of spatial-temporal feature extraction and the accuracy of output reconstruction,an anomaly recognition model based on continuous identical value recognition,quartering method model and k-means clustering analysis and the method of handling outliers with mapping mean filling and adjacent mean substitution are proposed.Thirdly,a regional grid in Mongolia as the typical grid in Greater Central Asia and Germany as the typical grid in Europe are selected to verify the accuracy of the renewable energy output reconstruction model proposed in this paper on regional and national nodes respectively.This thesis introduces the source of the original data and deals with the outliers in the measured data.The spatial and temporal characteristics of six wind farms and six photovoltaic power stations in a province of China are extracted,and the spatial correlation matrix is modified.The error analysis of the sub-regional and total regional reconstruction results is carried out using the mean absolute error(MAE)and root mean square error(RMSE)as prediction accuracy evaluation indicators.In this thesis,a method based on complex principal component analysis(CPCA)is proposed to extract the spatial-temporal characteristics of renewable energy generation,which solves the problem of inaccessibility of high spatial-temporal resolution output data of regional grids under the global energy interconnection demand.