Automatically and real-time identifying malfunctioning pv systems using massive on-line PV yield data
R.C. Nijman (TU Delft - Electrical Engineering, Mathematics and Computer Science)
Arno H.M. Smets – Mentor
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Due to the transition in the energy market from fossil fuel to renewable energy sources it is expected that the total installed amount of photovoltaic (PV) energy will continue to increase. To make sure that all these PV systems work optimally it is essential to have a good way of monitoring these devices. In this project an open generic model that can detect and identify malfunctions of PV systems based on yield data was developed. The model detects malfunctioning systems by comparing the yield ratio of a specific system with the yield ratio of similar neighboring systems. To determine the similarity of these neighboring systems properties such as their orientations were identified. The azimuth of the PV systems could be identified with an accuracy of 8 cardinal points, the standard error for the azimuth approximation was 2%. Finally the model identifies the system malfunction using several different algorithms.
The static error that can be detected by the model is inverter clipping or a limiting inverter. Furthermore the model can identify the following dynamic (system) errors:
· Broken string malfunctions (in module string and/or cell string).
· Data connection malfunctions.
· Data incomplete malfunctions.
· Less than diffuse yield malfunctions.
The significance of the model is that more than 22% of all systems that were analyzed had a limited inverter. Of the 1300 systems that were analyzed during July and August 2017 11% were identified to be malfunctioning with an average of 13.6% below the mean performance that is to be expected based on the performance of similar neighboring systems.