Research On The Health Condition Evaluation And Output Prediction Of Distributed Photovoltaic Generation System Of Hebei Southern Power Grid

The advantages of solar power generation and the introduction of various national incentive policies such as subsidy have made the construction of distributed photovoltaic power generation systems a trend.the number of connected users is increasing,and photovoltaic equipment has covered a vast scope.At the same time,problems such as the failure of photovoltaic equipment,the unauthorized users’ capacity expansion for subsidies and the continuous increase in the waste of solar power,have affected the enforcement of relevant policies,the use of new energy sources,and the stable operation of power systems adversely.Therefore,the health assessment and short-term output prediction of distributed photovoltaic power generation systems have great theoretical and practical significance:This paper takes 209,677 sets of distributed photovoltaic power generation system of Hebei southern power grid as the research object,and collects the power generation data of corresponding users from Power Consumption Information Acquisition System as the main data source,and fully utilizes the high-frequency full collection characteristics of HPLC technology,and uses several mainstream big data algorithms such as random forest and wavelet transform as theoretical support,to conduct research in the following 2 major issues.1)Establish the model to evaluate the health condition of the distributed photovoltaic power generation system.First,a typical power generation curve library of the power supply station is constructed through data cleaning,outlier data processing,null value filling,and data modeling.Second,according to users’ generating capacity,power value and typical data in the user’s region,take the deviation of monthly availability hour,the deviation of power generation in days that extreme value occurs and the correlation coefficient between user power generation curve and typical curve as 3 evaluation indicators to establish a comprehensive evaluation model of the operating status of photovoltaic power generation systems in order to evaluate their health and give results and output abnormal lists.2)Build the model to predict the short-term output of distributed photovoltaic.By combining the historical daily power generation amount,generation power data and corresponding historical weather data of users,construct characteristic factors that affect output.Then establish the model to predict photovoltaic output in 96 time points in the whole day based on machine learning algorithms and output results.Finally,the accuracy of the two models is verified with actual results and operating data in some county in Shijiazhuang,Hebei Province,proving that this research can effectively support distributed photovoltaic abnormal user screening,dynamic adjustment to power generation plan for power grid companies to help the implementation of photovoltaic poverty alleviation and new energy consumption policies.