Photovoltaic System Performance Index

The synthetic data is generated loosely following the concepts developed by Skomedal and Deceglie (2020) 

"https://doi.org/10.1109/JPHOTOV.2020.3018219" , 

"https://github.com/asmunds/simulate_pv_time_series"

Instead of producing the performance index through a normalization step, like with the real data, the synthetic data performance index is generated as a product of five synthetically generated components: system change, seasonality, soiling, performance loss rate, and the residuals.

When using real data, the performance index is calculated during a normalization step. This normalization step occurs after training a data-driven model on early-life inverter performance with the measured weather data. The weather data is generated using either the specific site’s weather station data or a satellite source’s weather data. The performance index is then calculated at the inverter-level and is a comparison between the actual energy generated and the model-estimated amount of energy produced under a given set of weather conditions.