光伏电站发电效率及异常发电状态诊断数据

本成果利用供电公司光伏发电数据和台账数据实现对全浙江省约30万台各类型光伏电站发电状态的自动监测和精细诊断，主要包括两个方面：一、通过所域指数完成各类小型光伏电站发电状态的精细诊断，此部分算法能够填补小型光伏电站无法全量监控的行业空白（由于光伏监控硬件设备投资费用大，故小型光伏电站基本都没有监控装置），二、自动测算大型光伏电站故障导致出力下降的数值，替代大型光伏电站的硬件监测手段（大型光伏电站虽有硬件监测设备，但是投资费用极其高昂），降低运维费用。本数据产品利用国网数据中台和用数环境实现对浙江30万台各类型光伏电站的每日自动诊断并更新结果，具体详情需要在国网浙江公司内网环境中访问查看，对国家电网内部使用，外部人员查看需经过相关部门审批。本算法主要分为三个部分：1. 所域级光伏发电指数的建立：以供电所所辖区域内的光伏电站为研究单元，自适应计算各个供电所所辖区域光伏电站发电指数：① 计算每个光伏电站的单位装机容量发电量；② 以供电所为单位分组并剔除异常数据，然后确定标杆电站；③ 通过标杆电站计算各光伏电站的光伏发电指数，实现同一供电所同一基准比较。2. 基于所域光伏发电效率指数的小型光伏电站发电状态精准诊断，方法包括：① 空间维度上根据所域级光伏发电指数通过支持向量机算法初步判断异常电站；② 通过时间维度确定异常电站，诊断出故障或异常情况，分为工单类、关注类和提示类，共12小类。3.基于所域光伏发电效率指数的大型光伏电站发电出力精准测算，方法包括：① 根据标杆电站的发电情况确定相似发电日；②利用随机森林算法，以标杆电站单位装机发电量、相似负荷日单位装机发电量、大型光伏电站相似负荷日单位装机容量发电量为输入，以大型光伏电站单位装机容量发电量为随机森林的输出，实现发电出力精准测算；③通过实际电量与测算发电量的比较，诊断大型光伏电站出力是否正常。

This work realizes automatic monitoring and precise diagnosis of the power generation status of approximately 300,000 photovoltaic (PV) power stations of various types across Zhejiang Province, by leveraging PV power generation data and ledger data from power supply companies. It mainly includes two aspects:

1. Precise diagnosis of power generation status for small-scale PV power stations via the regional-level index. This algorithm fills the industry gap that small-scale PV power stations cannot be fully monitored, as the investment cost of PV monitoring hardware is high, and almost no monitoring devices are installed on small-scale PV power stations.

2. Automatic calculation of the output reduction value caused by faults in large-scale PV power stations, replacing the hardware monitoring methods for large-scale PV power stations. Although large-scale PV power stations are equipped with hardware monitoring devices, their investment cost is extremely high, thereby reducing operation and maintenance costs.

This data product enables daily automatic diagnosis and result update for 300,000 PV power stations of various types in Zhejiang by utilizing the State Grid Data Middle Platform and data usage environment. Specific details can only be accessed and viewed in the intranet environment of State Grid Zhejiang Electric Power Co., Ltd. This product is for internal use by State Grid, and external personnel need to obtain approval from relevant departments to access the details.

The proposed algorithm consists of three main parts:

1. Establishment of regional-level PV power generation index: Taking the PV power stations under the jurisdiction of each power supply branch as the research unit, adaptively calculate the regional-level PV power generation index for the jurisdiction of each power supply branch:
① Calculate the power generation per unit installed capacity of each PV power station;
② Group by power supply branch, eliminate abnormal data, and then determine the benchmark power stations;
③ Calculate the PV power generation index of each PV power station using the benchmark power stations, achieving unified benchmark comparison within the same power supply branch.

2. Precise diagnosis of power generation status for small-scale PV power stations based on the regional-level PV power generation efficiency index. The specific methods are as follows:
① Initially identify abnormal power stations via the Support Vector Machine (SVM) algorithm based on the regional-level PV power generation index in the spatial dimension;
② Confirm abnormal power stations in the temporal dimension to diagnose faults or abnormal conditions, which are divided into 12 sub-categories: work order type, concern type and prompt type.

3. Precise calculation of power generation output for large-scale PV power stations based on the regional-level PV power generation efficiency index. The specific methods are as follows:
① Determine similar power generation days based on the power generation status of benchmark power stations;
② Adopt the random forest algorithm, taking the power generation per unit installed capacity of benchmark power stations, power generation per unit installed capacity of similar load days, and power generation per unit installed capacity of large-scale PV power stations on similar load days as inputs, and the power generation per unit installed capacity of large-scale PV power stations as the output of the random forest, to realize precise calculation of power generation output;
③ Compare the actual power generation with the calculated power generation to diagnose whether the output of large-scale PV power stations is normal.

搜集汇总

数据集介绍

特点

该数据集提供了浙江省光伏电站的发电效率及异常发电状态诊断数据，包含1001条记录，每日更新。数据用于自动监测和诊断光伏电站的发电状态，特别适用于小型和大型光伏电站的精细化管理，降低运维成本。