Choosing representative strings for measuring power loss

Share

The first time Lucie Garreau-Iles visited the location where the solar plant she was to investigate was installed, framed with willowy grasses 1.5 meters high, it probably had the romantic impact that many tourists love so much when traveling in the Mediterranean region.

As enchanting as it probably was, it was also one of the reasons why the analysis of the monitoring data originating from the 3.5 MW solar plant wasn’t leading anywhere.

Garreau-Iles works as Technical Regional Manager at DuPont Photovoltaic Solutions in a team which investigates the health of solar plants. For this mission, she needed to investigate whether the yield of the eight-year-old installation could be improved.

pv magazine Virtual Roundtables Europe 2020

 How can asset managers, O&M-contractors, and project developers ensure high yields? And to what extent can they increase them? These essential questions will be discussed at the 2020 pv magazine Roundtable Asset Management session. Join us on June 9 to meet speakers from Allianz Global Investors, JinkoSolar, DuPont Photovoltaic Solutions, and Ariel Re.  More information and free registration.

From the monitoring data, it was even difficult to judge whether the plant had a yield problem, because the weather sensors were also intermittently failing; meanwhile, there was no data for the first two years of operation, and missing or invalid data thereafter. This may not have been relevant to the investor, however, since high feed-in-tariffs secured enough revenue.

Determining power loss

During her field visit in the summer of 2019, Garreau-Iles found many modules with snail trails and delaminated EVA sheets. “There were high visual impacts,” she said. With snail trails, delamination and microcracks, it is not obvious to what extent a module’s power is affected. This only becomes apparent when microcracks develop into full cracks, parts of the cells are no longer fully connected, and yield declines.

The challenge in such cases is to determine the power loss with the least effort. This is for two main reasons: (i) to avoid spending too much money before it is clear that work is necessary; and (ii) because the stakeholders want to know the exact numbers, even when it is clear the modules have to be exchanged.

As a next step, the experts commissioned drone IR-thermography. On the images, it was clearly visible that hotspots had spread over the whole module field. They were visible on roughly 20% of the panels. This indicated that it is was worth further investigation, since deeper analysis could lead to a better understanding of the losses. It was still unclear how much the power loss and a potential yield gain would be when the installation was revamped.

Results from representative strings

IV-curve-measurements are a good method to determine power loss, but it is labor intensive. Therefore, based on the IR-images, the experts selected a representative string sample for IV-testing. This allowed them to accurately measure the performance by just testing half of the strings.

For some strings, the experts tested before and after cleaning, and determined that there was a power loss of about 1.6% due to soiling. Furthermore, the IR inspections revealed that 3.5% of the panels were disconnected on site. The IV-curve-measurement itself showed that the remaining connected panels had lost 19.3% power on average. The overall power loss was 22.8%.

“We could therefore conclude that a revamping operation would have a positive impact,” said Garreau-Iles. The threshold at which it was more cost efficient to exchange the modules in this installation was about 9% power considering a total panel replacement. With exchanging the modules and revamping this installation the experts calculated that the revenue would increase by 17%.

The revamping process is currently going on. All panels will be exchanged, even those not exhibiting hotspots, as they displayed microcracks which will evolve into cracks with or without hotspots, with an associated power drop. “It is therefore not cost effective to change only part of the panels,”  Garreau-Iles said.

Questions to discuss at the virtual pv magazine Europe 2020 Roundtable asset management session:

  • How important is it to know the exact power loss number?
  • How good are monitoring solutions and data acquisition for new installations? Is the problem discussed in this case “solved”? What should project developers and owners consider?
  • There are developments that allow thermal imaging to measure module performance. How good are they?
  • How should it be decided whether it is worth investing additional funds into a plant? Which “risks” should owners go for? Is it possible to conclude from statistics and “average failure rates of plants” how much effort one should invest in plant improvement?
  • What is the status of machine learning to tackle these problems a) for old plants and b) for new installations?