Research And Implementation Of Photovoltaic Inverter Anomaly Detection System Based On Deep Learning

Photovoltaic power generation technology is the most promising technology in renewable energy power generation technology.In the photovoltaic power generation system,the photovoltaic inverter mainly converts the direct current generated in the photovoltaic array into alternating current and then outputs it.At the same time,it can also protect the circuit and guarantee the performance of solar cells.Therefore,photovoltaic inverter plays a very important role in photovoltaic power generation system.At present,the maintenance mode of photovoltaic inverter usually adopts fixed-point maintenance and maintenance after failure,so the operation and maintenance personnel can not understand the real-time fault situation of photovoltaic inverter.Therefore,it is urgent to study the fast and low-cost abnormal detection technology of photovoltaic inverter in operation and maintenance environment.In this paper,the deep learning method is used to study the abnormal detection of photovoltaic inverterFirstly,by summarizing the research status of anomaly detection at home and abroad,this paper analyzes the shortcomings of this kind of research and finds the starting point of this topic;The generation countermeasure network is selected to study the problem of anomaly detection.Secondly,according to the data characteristics of photovoltaic inverter,the improved random forest algorithm is used to fill in the missing data,and then the Z-score standardization is used to normalize the filled data,and the distance image is used to visualize the inverter data for the imbalance of positive and negative samples.Then,the GANomaly structure is studied and a deep learning model based on it is established for anomaly detection of multi station photovoltaic inverter.According to the characteristics of time series data,the long-term and short-term neural network is introduced,and the model is improved to LSTM-GANomaly and ConvLSTM-GANomaly inverter anomaly detection model.Experiments show that the anomaly detection results are improved to a certain extent.On this basis,the residual connection is introduced,and an encoder is removed from the network structure to form a skipConvLSTM-GANomaly model,which makes the algorithm better integrate multi-scale features.The results show that skip-ConvLSTM-GANomaly algorithm is better than other algorithms.Finally,a photovoltaic inverter anomaly detection system based on B/S structure is designed with Python language.The improved photovoltaic inverter abnormal detection algorithm is built into the operation and maintenance system to realize the abnormal detection of photovoltaic inverter operation and maintenance data.