Output Prediction And Application Of Distributed Photovoltaic Power Station Cluster

Photovoltaic power generation has volatility and intermittentness.With the largescale integration of distributed power generation into the power grid,it has an impact on grid dispatching and power quality,which brings great challenges to the stable operation of the power grid.PV output forecasting is an effective way to effectively solve this problem,and PV cluster forecasting can effectively provide data support for grid management and enhance its management and control capabilities.This paper develops a distributed photovoltaic power station meteorological data detection and transmission system around this research goal,provides a data source for PV power plant output prediction,and then studies the PV power plant cluster prediction method.The output of photovoltaic power plants is closely related to meteorological factors.Many power stations have no conditions to provide high-quality data.In response to this problem,the paper has developed a photovoltaic power station based on national power industry standards and technical indicators,relying on cloud platform technology and embedded technology.A data monitoring transmission system,the system includes a weather detection terminal and a data transmission terminal,and the data detection system can transmit data to the RTU through the Zigbee protocol,and the RTU transmits the data to the cloud storage through the Ethernet,and can be displayed online in the cloud through the webpage.Data,and visualization,enhance system scalability and data management capabilities.It provides online monitoring for the application of the cluster power plant output method,and provides a meteorological data foundation for PV output prediction.Using four statistical correlation parameters,the relationship between different meteorological factors and power station output,the output between adjacent power stations,the output between non-adjacent power stations and the different weather types of the power station were analyzed to verify the performance of a correlation parameter.The output data shows that the cosine similarity indicates the data correlation between the power stations,which can reflect the fluctuation of the data volume between the power stations,and can also reflect the nonlinear correlation between the power station data.For the traditional power station output prediction,only the time domain analysis is considered,the analysis of the frequency domain characteristics is lacked,and the single field power plant combined prediction model based on CEEMDAN and Bayesian method is established.The representative of the cluster is selected by the principle of maximum correlation minimum redundancy.The power station,combined with the correlation coefficient and weight factor for cluster output prediction.The former exerts the characteristics of the combined method,and provides an effective means for the analysis of the characteristics of the power station and the prediction of the output.The latter can cover more geographical IMFormation and reduce the redundant IMFormation.Through the example verification,the predicted value is compared with the superimposed value of all the power stations in the cluster,and the absolute error of the predicted value of the cluster output is 2.1%,and the root mean square error is 3.6%.Compared with the predicted superposition value,this method can The average absolute error is reduced by 3.3%,the root mean square error is reduced by 3.8%,and the predicted value of the cluster power station is obtained.The meteorological data monitoring and transmission system of photovoltaic power station developed in this paper realizes the effective management of photovoltaic power station output data,and has the characteristics of high precision and strong applicability.Based on the correlation of photovoltaic output,a method for identifying and reconstructing bad data of photovoltaic power plants is proposed.Based on the EMD method,a single-field prediction model based on CEEMDAN and Bayesian neural network combination method is established.The correlation criterion is used to select the representative points of the cluster.This prediction method can effectively reduce the error of the predicted superposition value of the cluster power station.and enrich the means of the cluster output prediction.