Reasearch On Time Series Modeling Method For New Energy Power Stations Output Considering Spatio-temporal Correlation

The large-scale development and integration of new energy can effectively solve the energy crisis and environmental pollution problems.But its randomness and volatility have also brought tremendous challenges to the traditional power system.New energy power stations are mostly concentrated development models in China.There are many new power stations are close in space.Their output changes are similar,which intensifies the volatility of new energy.In order to comprehensively analyze and evaluate the impacts of large-scale new energy grid-connected and the acceptance capacity of power grid,a large amount of mid-long term new energy output data is often required as the basis for analysis and calculation.However,for most new energy power stations that have not been built or put into operation for a long time,the existing output data has problems such as limited data volume and insufficient record time,which is difficult to meet the needs of different time scale research issues in power system.Therefore,in order to improve the stability and acceptance of power system with large-scale new energy,it is necessary to consider how to use limited measured data to generate a large number of new energy output time series which are similar to actual data.This thesis first studies how to accurately describe the spatial correlation of new energy power stations.We consider using the Copula function to characterize this property.It is found that the single Copula function is difficult to fit the multi-dimensional distribution of the actual output of new energy power stations.Therefore,a hybrid Copula function fitting model is proposed to characterize this property.Simulation tests show that the hybrid Copula function model proposed in this thesis can accurately describe the complex spatial correlation structure of the output of new energy power stations.Then this thesis improves the problem of difficulty in selecting the number of states for existing wind power timing modeling method based on Markov process.An optimizing states number Markov Chain Monte Carlo(MCMC)method is proposed in combination with the preference principle of state number and the random power generation method based on cumulative distribution function.On this basis,a time series modeling method for the output of multiple wind farms combined with the high-dimensinal Markov process is proposed.The wind power data measured in different regions are used for testing.The results show that the proposed time series modeling method for wind farms can not only simulate statistical characteristics of the output of single wind farm well,but also maintain correlation between the output of adjacent wind farms.Finally,for the problem that most existing photovoltaic(PV)time series models have difficulty in meeting the needs of generating mid-long term data,a time series modeling method for the output of PV power plants is proposed in this thesis.In order to maintain the overall regularity of PV output,the time series modeling method uses decomposition technology to divide the measured PV power series into three parts: the normalized ideal output curve,the amplitude parameter series and the fluctuating component series.By analyzing the characteristics of each part and modeling to generate the corresponding simulation series,and then recombine simulation series to obtain the PV output time series with desired length.On this basis,a time series modeling method for multiple PV power plants combined with the high-dimensinal Markov process is proposed.The measured PV data is used for testing.The results show that the time series modeling method for PV power plants proposed in this thesis can not only simulate statistical characteristics of the output of single PV power plant well,but also maintain correlation between the output of neighboring PV power plants.