Research On The Distributed PV All-weather System And Regional Output Forecasting Method

Due to the volatility and intermittent nature of photovoltaic power generation,large-scale PV grid-connected networks have affected the stable operation of source and load of large power grids,and have affected the acceptance of photovoltaic power sources by large power grids.PV output forecasting is one of the key technologies to effectively solve this problem.This paper has carried out research on the development of full meteorological system and regional output forecasting method around PV output forecasting technology.As one of the important basic data of PV forecasting,meteorological data has the problems of low acquisition accuracy and poor scalability for traditional meteorological acquisition devices.This paper develops a distributed detection terminal and cloud platform based on embedded technology and cloud computing technology.A full-scale meteorological detection system for PV power plants with high precision and scalability.With reference to the power industry standards and technical specifications,the whole meteorological acquisition terminal equipment and sensor network were developed.The detection terminal transmits the data collected by the sensor to the RTU through CAN communication,and the RTU sends the data to the cloud for storage via TCP/IP Industrial Ethernet.Calculation and visualization processing.Remote management of the system and remote transmission of data are realized.The terminal equipment of the meteorological inspection system was calibrated and tested at the National Institute of Metrology and the Institute of Metrology of Zhejiang Province,respectively,and obtained a number of calibration certificates.Aiming at the lack of historical data of PV output,based on the analysis of the characteristics of PV output,a method based on the correlation of PV output is proposed.Firstly,the volatility of single-field and regional PV power generation is analyzed from different intervals of output and the relationship between output volatility and installed capacity.The analysis shows that the impact of regional installed capacity on volatility is smaller than that of PV power station.Secondly,Pearson correlation and cosine similarity are compared to represent the correlation of PV output.The analysis shows that the power plant output curve selected by cosine similarity is more similar to the target output curve profile.Finally,based on this,a method of data reconstruction based on the correlation of photovoltaic output is proposed.The simulation data proves that the reconstructed data better restores the trend of the original data,and the deviation is within the acceptable range.Aiming at the way of the sub-region division and selecting the power station of the traditional upscaling forecasting method is imperfection.The sub-region partitioning model based on EOF decomposition and hierarchical clustering and the representative power station selection model based on maximum correlation minimum redundancy criterion(mRMR)are established.The former exerts the EOF decomposition space mining ability,and the result of the division fully considers the spatiotemporal characteristics of PV output.The latter uses the mRMR criterion to select the representative power station,which cover as much sub-region information as possible,and introduce redundant information as little as possible,which is beneficial to improve the forecasting accuracy.Through the example simulation and comparison with the traditional method forecasting results,the forecasting method MAE is reduced by 3.68%,and the RMSE is reduced by 4.74%,which effectively improves the forecasting accuracy.This paper relies on the embedded platform and cloud platform to develop a full meteorological system,which achieve remote management of the system,with high precision,versatility,and obtained a number of calibration certificates.Based on the analysis of PV output characteristics,a missing data reconstruction method based on the correlation of PV output is proposed.In the regional forecasting method based on multi-level spatial upscaling,a sub-region partitioning model based on EOF decomposition and hierarchical clustering and a representative power station selection model based on mRMR criterion are established.The forecasting method can significantly reduce the forecasting error and improve the forecasting accuracy,which has a certain application value.