Detecting Abnormal Profiles in the Database of a PV plant Through Programmatic Comparison of Per Unit Profiles

This paper proposes a new concept to detect the abnormal profiles of a grid-tied PV plant. Continuous measurement of many parameters and storage in the database is a routine task performed by any commercial Grid-tied PV plant. Detecting the abnormalities in the operation of PV plant is an import part of the reliable operation of the PV plant. Operational factors like electrical and non-electrical faults result in `abnormal' profiles. However, non-operational factors like interruptions in the data transfer protocols found filling the database with `corrupt' data leading to the `abnormal' profiles. Detecting `abnormal' profiles is crucial to the operation of PV plant. A new concept of comparing the daily profiles of a set of parameters and labeling a day normal or abnormal is proposed in this paper. The methods are programmed to detect the data on a given day is `normal' or `abnormal.' Recorded data from a 6.2 kW grid-tied PV plant is used for validating the proposed methods. A sample dataset and proposed methods can be downloaded at https://ieee-dataport.org/authors/manjunathmatam.