AUTOMATED SOLAR POWER PERFORMANCE ANALYSIS

摘要

This paper outlines an algorithmic approach to classifying and quantifying drivers of underperformance in real-world photovoltaic (PV) systems, leveraging full time series data collected in the field. Solar power performance is often assessed by comparing measured energy output against expected energy output on a monthly or yearly basis. Finegrained (e.g., 5-minute) data is often only used for troubleshooting specific issues, because interpreting this large volume of data can be challenging. The approach described in this paper is the first to utilize a history of finegrained time series data from meters, inverter, and weather sensors in a fully automated PV analysis and issue-attribution system. Drivers of underperformance are identified so that issues can be addressed and design models can be improved to more accurately predict future output. Case studies are provided to illustrate the real-world effectiveness of this approach.