TY - GEN
T1 - Automating Methods for Validating PV Plant Equipment Labels
AU - Ranalli, Joseph
AU - Hobbs, William B.
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Operators of utility scale photovoltaic plants can face challenges in accurately matching operational data at various levels with their physical locations in the plant. This may have financial implications due to misdirected maintenance efforts. A method to predict equipment locations using time series data analysis was developed, but previously relied on some manual processing. This paper details efforts to automate the technique for improved efficiency and to facilitate operational deployment. The automated methodology relies on selecting ideal cloud motion vectors by targeting time periods based on their variability score metric. Ten vectors with well-spaced angular directions were selected after filtering based on quality control parameters. After the analysis methodology was run for an entire plant, mislabeled equipment was detected by implementing optimization code for solving the well-known Assignment Problem. Because predicted equipment positions have a circular dependence on their original expected positions, the process was applied iteratively as mislabeled entities were identified to converge to a final assignment of predicted position for each plant component. The results from this algorithmic methodology agree with the validated results from the original manual method. These improvements further the goal of offering the method as a ready-to-use tool for validating the physical locations of equipment within a utility-scale photovoltaic plant.
AB - Operators of utility scale photovoltaic plants can face challenges in accurately matching operational data at various levels with their physical locations in the plant. This may have financial implications due to misdirected maintenance efforts. A method to predict equipment locations using time series data analysis was developed, but previously relied on some manual processing. This paper details efforts to automate the technique for improved efficiency and to facilitate operational deployment. The automated methodology relies on selecting ideal cloud motion vectors by targeting time periods based on their variability score metric. Ten vectors with well-spaced angular directions were selected after filtering based on quality control parameters. After the analysis methodology was run for an entire plant, mislabeled equipment was detected by implementing optimization code for solving the well-known Assignment Problem. Because predicted equipment positions have a circular dependence on their original expected positions, the process was applied iteratively as mislabeled entities were identified to converge to a final assignment of predicted position for each plant component. The results from this algorithmic methodology agree with the validated results from the original manual method. These improvements further the goal of offering the method as a ready-to-use tool for validating the physical locations of equipment within a utility-scale photovoltaic plant.
UR - http://www.scopus.com/inward/record.url?scp=85211608786&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85211608786&partnerID=8YFLogxK
U2 - 10.1109/PVSC57443.2024.10749123
DO - 10.1109/PVSC57443.2024.10749123
M3 - Conference contribution
AN - SCOPUS:85211608786
T3 - Conference Record of the IEEE Photovoltaic Specialists Conference
SP - 310
EP - 315
BT - 2024 IEEE 52nd Photovoltaic Specialist Conference, PVSC 2024
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 52nd IEEE Photovoltaic Specialist Conference, PVSC 2024
Y2 - 9 June 2024 through 14 June 2024
ER -