Research On Fault Diagnosis Method Of Photovoltaic Power Plant Based On Electrical Data Characteristics

With the development of society and the increasing demand for energy,fossil energy is gradually exhausted.Solar energy,as the renewable energy with the largest amount of resources,plays an increasingly important role in the energy field.Photovoltaic power generation has gradually developed into the backbone of the new energy power generation field.With the rapid development of the photovoltaic industry,the fault diagnosis technology for photovoltaic power plant is also constantly developing.It is particularly important to ensure the normal operation of photovoltaic power plant and improve the generation efficiency.This paper analyzes the performance characteristics and change rules of multiple faults in the process of electrical data monitoring of photovoltaic power plant in power generation,transmission and distribution,and grid connection.The extraction method of abnormal data sequence based on data mining algorithm is studied.The fault diagnosis method based on deep learning algorithm is studied.And then "distributed photovoltaic power plant data monitoring and fault diagnosis platform" is built.Firstly,the framework and logic of distributed photovoltaic power plant data monitoring and fault diagnosis platform are proposed.The influence of internal and external factors on photovoltaic modules and the influence of typical faults on the output characteristics of photovoltaic modules are emphatically analyzed.Then fault classification and data feature description of distributed photovoltaic power plant are summarized.Secondly,a relatively objective anomaly/normal-deviation rate sequence sample set is established for support vector machine model construction through input feature quantity construction and sample identification construction methods.Then the abnormal data sequence extraction method based on the data mining algorithm is proposed.Method verification based on power plant actual operation data and case analysis of abnormal data sequence extraction are carried out.The results show that the accuracy rate for the test set is more than 96%.Finally,the curve shape,distribution and numerical change of different fault data sequences are analyzed to obtain unique and identifiable data variation characteristics that can be oriented to multiple types of faults.Then a multi-fault diagnosis algorithm based on deep learning is proposed.Method verification and overall case analysis of fault diagnosis are carried out.The results show that the fault comprehensive diagnosis rate is more than 95%.