DNV strives to improve machine learning-based maintenance models by introducing high-quality standardized logs, currently lacking in the industry
DNV, the independent energy expert and assurance provider, has launched a joint industry project (JIP) to prompt a significant leap forward in maintenance practices in the Solar PV industry.
With the remarkable growth of Solar PV, DNV forecasts an annual addition of between 300 and 500 GW from 2030 onwards, resulting in a staggering 9.5 TW of total installed capacity by mid-century. However, running a PV plant profitably is a significant challenge, particularly when it comes to maintenance: identifying parts requiring immediate attention, prioritizing maintenance tasks, and scheduling downtime efficiently are crucial for ensuring healthy margins.
DNV has identified a clear trend towards more professional maintenance in the industry, with improvements made to maintenance processes and procedures. However, there is still a long way to go, and DNV is now calling on the industry to embrace the next step: predictive maintenance. Unlike traditional preventative (visiting sites at regular intervals) or corrective (repair) maintenance approaches, predictive maintenance is carried out only when a fault is expected and before it occurs – thereby optimizing the costs associated with on-site maintenance.
To implement predictive maintenance systems successfully, large amounts of data are required, including production figures, power and status information, and most importantly, maintenance logs that meticulously document all activities at the PV plant. While the PV industry is generating and capturing data from its processes, DNV has identified a critical issue with maintenance logs: their levels of detail and quality are highly variable – when both inconsistency and poor data quality poses a significant challenge in training machine learning models to meet acceptable performance standards.
In response to this challenge, DNV’s newly-initiated JIP focuses on improving machine learning-based predictive maintenance models by introducing high-quality and standardized maintenance logs. The project consists of several work packages, including coordination, definition of maintenance log standards, implementation at operating PV plants, development of predictive maintenance models based on the logs, and analysis of results – in order to establish a final agreement on the format for maintenance logs.
Image Source: DNV
'Through this JIP, we aim to achieve an agreed minimal standard of maintenance logs, develop predictive maintenance models, and evaluate their performance', said Lars Landberg, Vice President and Group Leader for Research and Development on Renewables at DNV. This project has the potential to significantly enhance the efficiency and profitability of PV plants. We believe that this initiative aligns perfectly with our mission: our experience shows that programs that explore the potential for collaboration are a great way to reach and share innovation that benefits whole sectors, and that’s exactly what solar PV needs to sustain its formidable growth'.
'This JIP is a testament to our commitment to driving progress and ensuring the Solar PV industry's long-term success', added Juan Carlos Arévalo, CEO at GreenPowerMonitor, a DNV company and Executive Vice President for Energy Systems at DNV. 'Standardized practices will foster better communication and more effective decision-making, and the proactive use of advanced predictive models will reduce downtime, minimize breakdowns, and optimizes solar PV systems' lifespan. The project's forward-looking vision aims to elevate industry performance, promote sustainability, and ensure success in an ever-changing energy landscape.'
Source: DNV
