Design Of Photovoltaic Module Fault Diagnosis System Based On Data Platform

Since 2012,China has vigorously developed the photovoltaic industry.The standard of installed capacity of photovoltaic power generation has been included in the 12 th Five-Year Plan,and it has been elevated to the height of a strategic emerging industry.With the vigorous development in the past decade,the installed capacity of photovoltaic power stations has been improved by leaps and leaps,and distributed photovoltaic power stations are becoming more and more popular.However,there is still a large room for improvement in the operation and maintenance efficiency of photovoltaic power stations and the collection and management level of field data and information.Abnormalities are found in the sampling data of photovoltaic equipment,and then the fault diagnosis of photovoltaic modules can help timely maintenance or replacement of module equipment,so as to avoid greater impact.Module fault diagnosis helps to improve the operation and maintenance efficiency of photovoltaic system,improve the quality of photovoltaic power generation,reduce the cost of photovoltaic power generation,and increase the proportion of photovoltaic clean energy in the total power generation,which has become an important thrust to achieve the goal of "carbon neutral".In this paper,a photovoltaic data platform is designed and built.Based on the sampling data of the platform,the fault diagnosis method of components is studied,and the system design of fault diagnosis application is realized.The main work contents of this paper are as follows:Firstly,based on the photovoltaic demonstration power station,the "end-to-cloud" twolevel data acquisition scheme is adopted,and a distributed photovoltaic data platform is designed to realize the reliable collection,storage and management of component-level data.The data platform aims at unified storage and analysis of massive photovoltaic data of photovoltaic power stations,improves the use efficiency of photovoltaic data samples by taking advantage of the large capacity,extensibility and parallel computing of the distributed framework,and provides the data basis and platform basis for the subsequent research on component fault diagnosis algorithm.Secondly,a method to judge the power abnormality of components based on sampling data is proposed.In this paper,photovoltaic data samples collected and processed by the data platform are taken as the support to conduct a visual qualitative analysis of the characteristics of instantaneous power of photovoltaic modules,focusing on the quantitative relationship between environmental parameters and power,and pointing out the influence of sampling equipment asynchronization on sample quality.The ANN network model was selected based on the power characteristics,and the model predicted the due power of the component according to the environmental parameters.The due power was taken as the reference index to judge the abnormal power of the component.Then,in order to calibrate the output power of the module,a method to estimate the annual attenuation of the generation performance of the module is proposed.The proposed method extends the beta distribution probability model to model the skewed distribution of component calibration power with its high freedom.The model estimates the sample population probability distribution based on a small number of long-term monitoring samples.According to the probability distribution,the typical power values of the components under different service periods were selected,and then the annual attenuation of the generation performance of the components was calculated.The standard of abnormal power of components is set by using the annual attenuation and the output power of components.Thirdly,because the power abnormal judgment only uses the single point sample,the description force of the component condition is insufficient,and only the components that may have failure can be preliminarily screened.In this paper,two fault diagnosis methods based on the power sequence of components are proposed.The first method uses DTW method to measure the similarity of component samples to solve the matching problem of components in the same case.The samples to be tested are matched with the samples with existing fault labels to obtain the diagnosis results.The second method gives the index value of the component fault as a diagnostic reference.In order to obtain more information about the working condition of the component,the power sequence was reconstructed in phase space.The system state trajectory of the component in phase space is analyzed,and the calculation methods of the factors related to fault diagnosis,such as environmental stability,power anomaly,degree of deviation and time scale,are given,and then the index value of fault diagnosis is obtained.Finally,with the photovoltaic data platform as the support and algorithm research as the core,the fault diagnosis method is integrated into the software module to realize the system design of photovoltaic module fault diagnosis application.Completed the interaction with other platforms in the photovoltaic operation and maintenance project,was responsible for responding to various detection requests for component fault diagnosis,and reflected the engineering practical value of photovoltaic data and algorithm research.