Improved PV fault diagnosis tools through self-learning algorithms, IoT sensors and automatic aerial infrared image analysis

30-03-2020

<

Today’s photovoltaic (PV) technologies are highly efficient and productive, and can last up to more than 25 years. However, continuous maintenance is needed to keep them operating at top capacity. As shown in the chart below, an estimated 30% of PV plants underperform, and current maintenance practices seem to fall short in reliably identifying faults in PV equipment.

In the framework of the imec.icon research program, a collaborative project called ANALYST PV was set up to solve this problem. The project aims to optimize PV asset management and performance by further improving fault diagnosis techniques and complementing these with Internet of Things (IoT) sensors, AI-enabled root cause analysis, and automatic aerial infrared (IR) image analysis. 3E is leading the ANALYST PV in cooperation with imec – PVMS, imec – IPI – Ugent, Allthingstalk, Sitemark and Laborelec.

<
Graph: Actual vs. expected Performance Ratio (PR) of over 600 PV plants


Current basic PV monitoring systems falling short

Most basic PV monitoring and maintenance systems continuously measure the performance of PV systems, but their resolution is too low to identify exactly where the power losses originate from. These systems usually rely on manually entered data, which is time-consuming and prone to misconfiguration errors. Aerial infrared scans performed by drones are also used to gather additional information about PV cells, but current image analysis algorithms generate many false positives.

The power of automation and machine learning
To answer to these shortcomings the ANALYST PV consortium will further improve self-learning and continuous monitoring solutions by enhancing their spatial resolution with data from IoT sensors and the fusion of regular and infrared images.

The four innovation goals of the consortium are:

  1. Using machine learning to develop an automated digital twin of the PV assets that will serve as a basis for a rapid fault detection system.
  2. Creating an automated, efficient and reliable PV fault detection algorithm.
  3. Validating and measuring the benefits of the new framework.
  4. Linking those observations to decision-making by providing insights into PV equipment lifespan, power and revenue loss, and possible repair/replacement actions.

The framework developed by the ANALYST PV consortium will significantly reduce manual intervention and man-hour spent on PV asset management. It will also help maximize PV power output by automating the processes involved in fault detection and diagnosis.

Mauricio Richter, Head of Innovation and Product Manager at 3E, explains: “Within the scope of the ANALYST PV project, 3E will continue improving its failure detection techniques by largely automating the analysis of key PV failure modes. This will ensure optimal response time to underperformances in PV plants monitored by SynaptiQ. This cloud-based software is 3E’s PV monitoring and reporting platform and is currently connected to more than 6 GW PV plants worldwide. The ANALYST PV project will help us to enhance our data analytics tools by, for example, adding the IR based data input, which will provide more insight into the location of failures in PV assets.”

He concludes: “The ANALYST PV project helps 3E sustain its position as leading-edge data and technology company in a fast-evolving sector, bringing best-in-class products and services on the market.”

Get in touch through the form below for more information.

This work was executed within the imec.icon project ANALYST PV, a research project bringing together academic researchers (imec-PV and imec-IPI-Ghent) and industry partners (3E, Sitemark, AllThingsTalk and Engie Laborelec). The ANALYST PV project was co-financed by imec and received project support from Flanders Innovation & Entrepreneurship (project nr. HBC.2019.0050) and Innoviris.