Photovoltaic Model-Based Solar Irradiance Estimators: Performance Comparison and Application to Maximum Power Forecasting

TL;DR

This paper compares three PV model-based irradiance estimators for maximum power prediction, demonstrating their accuracy and robustness, and applying them to improve solar power forecasting during curtailment periods.

Contribution

It introduces and evaluates three PV-based irradiance estimators, showing their effectiveness in maximum power estimation and enhancing forecasting accuracy during curtailments.

Findings

Estimators accurately reconstruct maximum PV power during curtailment.

Methods outperform pyranometer-based estimations with simpler sensors.

Application improves training of solar power forecasting models.

Abstract

Due to the increasing proportion of distributed photovoltaic (PV) production in the generation mix, the knowledge of the PV generation capacity has become a key factor. In this work, we propose to compute the PV plant maximum power starting from the indirectly-estimated irradiance. Three estimators are compared in terms of i) ability to compute the PV plant maximum power, ii) bandwidth and iii) robustness against measurements noise. The approaches rely on measurements of the DC voltage, current, and cell temperature and on a model of the PV array. We show that the considered methods can accurately reconstruct the PV maximum generation even during curtailment periods, i.e. when the measured PV power is not representative of the maximum potential of the PV array. Performance evaluation is carried out by using a dedicated experimental setup on a 14.3 kWp rooftop PV installation. Results also proved that the analyzed methods can outperform pyranometer-based estimations, with a less complex sensing system. We show how the obtained PV maximum power values can be applied to train time series-based solar maximum power forecasting techniques. This is beneficial when the measured power values, commonly used as training, are not representative of the maximum PV potential.

Figures