Research On Fault Diagnosis Methods Of Photovoltaic Power Systems Based On Deep Learning

As global environment and energy issues become more and more prominen-t,many countries have raised the development of new energy industries to the national strategic level.As a renewable,non-polluting new energy source,solar energy has excellent characteristics,such as environmental protection and easy obtaining,so it has received extensive attention and research in recent years.Photovoltaic power generation system is the most important device for converting solar energy into electrical energy.It,generates direct current in the photovolta-ic array by photoelectric effect,converts direct,current,into usable alternating current through inverter,and then integrates the electric energy into the grid or directly supplies it to the load.Most of the photovoltaic power stations are built in areas with harsh environments and sparsely populated areas.Thousands of faults per month not only cause great difficulties for manual inspection,but also cause a large increase costs in operation and maintenance and asignificant reduction in power generation efficiency.Therefore,it is particularly important to diagnose faults in photovoltaic power generation systems.With the advancement of technology,the level and complexity of the photo-voltaic power generation system have been improved,and the traditional fault diagnosis methods have been unable to meet the actual needs.At the same time,with the development of sensor technology,computer technology and the coming of big data era,people continue to increase the ability of obtaining,storageing and computing data for photovoltaic power generation systems,it lays the foun-dation for the application of emerging intelligent algorithms based on data in the intelligent fault diagnosis of photovoltaic power generation systems.Deep learning algorithm is an emerging intelligent algorithm based on neural network.It has great advantages in processing high-dimensional,complex and severe non-linear data.Deep neural network is a representative of deep learning algorithm.By increasing the number of layers of the network,the network can learn deeper features of data.It is widely used in image processing,speech recognition and natural language processing.In addition,as another representative of the deep learning algorithm,the convolutional neural network greatly enhances the ability of network feature learning by adding convolutional layers and pool layers,which makes it have strong performance in classification,so it is often used in the fields of image classification,face recognition,etc.Therefore,the applications of deep learning algorithms in the field of fault diagnosis and fault classification also have a very broad prospect.This paper uses Simulink to build aphotovoltaic power generation system sim-ulation model,simulates the operation of photovoltaic arrays and photovoltaic inverters under different fault conditions.In the fault diagnosis research of photo-voltaic arrays,the variation trends of current,voltage and power under different temperat,ure and irradiance conditions are analyzed,and DNN algorithm is used to diagnose the faults under different conditions.After a lot of experiments,it is proved that the DNN algorithm has high reliability in the fault diagnosis of photovoltaic arrays.In the fault diagnosis research of photovoltaic inverter,six common types of inverter faults are selected to analyze the variation of three-phase current under different faults.By using the Empirical Mode Decomposition(EMD)method,the initial feature of the original time domain data is extracted to generate the eigenvalue matrix.Combining with the powerful feature extrac-tion and learning ability of CNN for fault diagnosis,a new method based on EMD-CNN is proposed.A lot of experiments have shown that the EMD-CNN method is effective in the fault diagnosis of photovoltaic inverters and can bear the fault diagnosis task well.After the performance evaluation experiment of fault diagnosis in the simu-lation system by the deep learning method,the reliability of the deep learning method in the fault diagnosis scenario of the real photovoltaic power generation system is verified.We analyze and transform the acquired real data,and then we use statistical methods to diagnosis communication faults,finally we use EMD-CNN method for inverter fault diagnosis.The experimental results show that the statistical method can accurately determine the common communication faults in the photovoltaic power generation system.Using EMD-CNN,we can efficiently and accurately complete the fault diagnosis of photovoltaic inverters in the real environment,the results prove that the method is effective and applicable in real scene.