Distributed Photovoltaic System Anomaly Detection Based On Metering Data

As a renewable energy power generation technology with high technology maturity and low resource cost,photovoltaic power generation is an important means to achieve the "dual carbon" strategic goal.Affected by technological progress and policy support,the development of centralized photovoltaic power generation systems is slow,and the development of distributed photovoltaic power generation systems has begun to accelerate.With the advantages of being free from land use policy constraints and flexible deployment,the proportion of new photovoltaic systems has increased significantly.increase.Distributed photovoltaic power generation systems are scattered outdoors,lack professional operation and maintenance,have a large number of devices,and have various forms of abnormal failures,which may be manifested as array open circuit/short circuit and inverter overcurrent/overvoltage,or power tube open circuit.and hot spots.The probability of failure of photovoltaic systems is significantly higher than that of other power generation forms.In addition,there are no rotating parts,and the failure is highly concealed.If the failure cannot be diagnosed and repaired early,it will significantly reduce the full-life cycle power generation.Therefore,it is urgent to develop a fault anomaly detection method for distributed photovoltaic(PV)systems,so as to improve the equipment integrity rate and give full play to the power generation efficiency.Unlike centralized PV,which can use complete information such as infrared and electrical quantities of grouped strings to diagnose and identify faults,distributed PV systems can only identify and diagnose faults based on limited information such as upstream port metering information and reported installation capacity.In this paper,the fault of PV system is finally reflected in the characteristics of abnormal power generation,and the rainy and cloudy weather will reduce the power generation indefinitely,and the idea of using only the power generation data in clear weather to identify abnormal faults and eliminate the interference of cloudy weather is proposed to diagnose abnormal faults.In addition,in order to exclude persistent abnormal interference and determine whether it is a sudden abnormality,a spatiotemporal correlation analysis is performed on adjacent photovoltaic systems.Combined with the above analysis,this paper first proposes a clear weather screening mechanism based on the difference calculation of solar irradiance;and then combines the Pearson correlation analysis to compare the space of different PV power stations and the time of the same power station,and select the PV output on the same standard data,eliminate the factors that interfere with the training of the abnormal recognition model,and then use the change of the power generation of PV power plants in sunny weather under the same standard,and propose an abnormal detection method based on the isolation forest algorithm.Finally,combined with the data of normal photovoltaic power plants screened by the isolated forest,an anomaly detection and early warning method based on neural network quantile regression is proposed to fit the normal output range of photovoltaics.on this basis,determine the future time of the photovoltaic power station and other abnormal conditions of the photovoltaic power station to be detected.Based on the limited data of the metering and marketing system,this paper realizes the intelligent detection of abnormal user distributed photovoltaic faults,which can give full play to the power generation efficiency and help achieve the goal of dual control of energy consumption.