Python program for Detecting corrupt Data in a PV plant Database - Validation Results from a 273kW NIST PV plant dataset

This folder contains two csv files and one .py file. One csv file contains NIST ground PV plant data imported from https://pvdata.nist.gov/. This csv file has 902 days raw data consisting PV plant POA irradiance, ambient temperature, Inverter DC current, DC voltage, AC current and AC voltage. Second csv file contains user created data. The Python file imports two csv files. The Python program executes four proposed corrupt data detection methods to detect corrupt data in NIST ground PV plant data. First and fourth methods are statistical approaches performing a direct comparison of the parameters. These two statistical methods can be applied from the first day of installing a PV plant. Second and third methods are machine learning based approaches involving training and testing procedures. These two machine learning approaches need some days of historical data prior to applying them. This program is useful to PV plant users, researchers, PV plant monitors, third party service providers to clean their PV plant datasets. By replacing the existing dataset set with their own dataset, one can use the program for filtering their data. This program requires the PV data set to have six parameters: POA irradiance, ambient temperature, Inverter DC current, DC voltage, AC current and AC voltage.

本文件夹包含2个CSV文件与1个Python脚本文件。其中一份CSV文件为从https://pvdata.nist.gov/ 导入的NIST（美国国家标准与技术研究院，National Institute of Standards and Technology）地面光伏电站数据集，该数据集包含902天的原始数据，涵盖平面阵列（POA, Plane of Array）辐照度、环境温度、逆变器直流电流、直流电压、交流电流及交流电压共六项参数。另一份CSV文件为用户自建数据集。
该Python脚本可导入两份CSV文件，并执行四种预设的损坏数据检测方法，以识别NIST地面光伏电站数据中的损坏样本。第一种与第四种方法为统计类检测方案，可直接对各项参数进行比对，此类方法可在光伏电站投运首日即可启用。第二种与第三种方法为基于机器学习的检测流程，需先完成训练与测试环节，因此在启用前需要提前积累数天的历史数据。
本程序可助力光伏电站运维人员、研究人员、电站监控方及第三方服务供应商完成光伏电站数据集的清洗工作。用户仅需将示例数据集替换为自有数据集，即可使用该程序完成自身光伏数据的过滤工作。需注意，本程序要求待处理的光伏数据集需包含以下六项参数：POA辐照度、环境温度、逆变器直流电流、直流电压、交流电流及交流电压。