Dialing in Data Key to Developing Successful Predictive Maintenance

Data analysis is helping operators in the solar power sector develop more efficient systems that optimize operations and deliver cost savings.

Utilities and other developers of solar power installations know the importance of optimizing operations to reduce costs and improve efficiency. There are several tools available to support operational goals, including the use of data analytics to support predictive maintenance.

1. Owners and operators of solar power installations can improve the efficiency of their systems through the use of predictive maintenance programs that analyze performance data. Source: Shutterstock 

Monitoring the performance of solar panels (Figure 1) and other equipment, and identifying issues before they become problems, supports operations and maintenance (O&M) work, helps avoid unplanned downtime, and provides cost savings. Analyzing sensor data can help predict equipment failures, while machine learning (ML) and artificial intelligence (AI) algorithms can identify patterns and anomalies that may indicate potential failures. This allows operators to take corrective action before issues occur.

“Developing a predictive maintenance program for a utility or solar farm involves strategic planning and resource considerations for owners of this asset class,” said Ryan Bullock, senior asset manager for CAMS, a fully integrated service provider for owners of energy infrastructure assets. “The most-effective approach often involves collaboration with the O&M provider, OEM [original equipment manufacturer], and specialized third-party entities.”

Bullock and other experts who spoke with POWER noted various ways that data analytics can help solar energy businesses improve their asset management strategies. Those include tracking the performance of equipment, finding areas that could be improved, and setting up more efficient maintenance schedules. That can maximize an asset’s lifespan, cut repair and replacement costs, and support a goal of increased profitability. A solid predictive maintenance program can help manage risk, and improve decision-making related to a solar energy facility.

“By leveraging advanced analytics, machine learning, and AI algorithms, asset managers and operators can forecast potential failures by analyzing broader datasets,” said Bullock. “This predictive capability allows for early anomaly detection, enabling proactive maintenance measures to prevent downtime and optimize system performance. The integration of diverse data sources and cutting-edge predictive technologies empowers asset managers to anticipate O&M needs more accurately, ensuring efficient and proactive maintenance strategies.”

Supporting Operational Data

The industry insiders who spoke with POWER said operational data alone is not enough to establish a strong predictive maintenance program. They said several strategies are needed to allow asset managers and operators to anticipate O&M needs and maintain efficient system performance.

Vikhyat Chaudhry, chief technology officer of Buzz Solutions, told POWER, “Utilities need to implement data collection techniques via sensors in order to inspect their solar assets. Utilities are using IoT [Internet of Things] sensors and mobile visual sensors, such as drones, for collecting data around their assets. Sensors [both visual and non-visual] are becoming commoditized and cheaper, hence, we are seeing the trend of utilities owning these sensors and deploying them in-house. Additionally, we are seeing utilities train their workforce to operate drones around their solar assets for regular inspections.”

Chaudhry continued, “Utilities are evaluating the in-house expertise in data analytics, machine learning, and domain-specific knowledge related to the utility or solar farm, as well as assessing the availability and quality of data. But with recent trends and expertise of external organizations with AI/ML technologies, utilities are choosing to use third-party digital and AI tools for analysis and reporting while maintaining the data collection and data retention with their in-house teams and tools.”

“Operational data provides valuable insights into equipment performance, but solely relying on this data has its limitations in establishing a predictive maintenance program. Operational data primarily reflects current or past performance, lacking the predictive capability needed to anticipate future issues proactively,” said Bullock. “To establish a strong predictive maintenance program, a more holistic approach is necessary. This entails integrating additional contextual information beyond operational metrics. Incorporating environmental factors, historical maintenance records, and equipment specifications provides a more comprehensive view of system health.”

Chaudhry agreed. “Operational data is a crucial component of a predictive maintenance program, relying solely on operational data has limitations,” he said. “To establish a robust predictive maintenance program, additional elements are needed to allow asset managers and operators to anticipate O&M needs and maintain efficient system performance. Operational data provides information about the current state of equipment but may not inherently contain predictive patterns of future failures or performance degradation. Additional analysis and modeling are often required to extract meaningful insights.”

Developing Programs

Doug Mackenzie, vice president, Energy Resilience at JLL, which focuses on renewable energy and sustainability solutions, said his company “works with our clients to provide holistic O&M programs for sustainable infrastructure assets, including solar PV. We use a combination of independent technical delivery and third-party providers to develop and implement programs. Having in-house capability to monitor and diagnose equipment degradation and failure allows cost-effective and rapid response to common issues, while working with third-party providers allows us to leverage advanced analytics, diagnostic, and repair capabilities to supplement our in-house capabilities.”

Mackenzie said, “Power electronics failure and degradation are typical problems” of solar arrays. “Detection systems should be able to monitor for both failures and potential root causes such as weather, power quality, and other issues.” Mackenzie told POWER that adding external data such as “weather, utility power quality, and typical equipment trends, creates a more holistic and accurate representation of root causes and vulnerabilities, which in turn increases the reliability and efficiency of that power source.”

Along with the use of external data, bringing more perspectives to O&M situations can also bring value. Bullock told POWER, “While leveraging internal knowledge and resources is crucial for understanding specific operational needs, bringing in external expertise offers diverse perspectives and specialized tools that can complement internal capabilities. Collaborating with third-party vendors, service providers, and OEMs can provide valuable insights into advanced technologies, predictive analytics, and industry best practices. This combination allows for a comprehensive program that merges internal operational understanding with external innovation and specialized knowledge, leading to a more robust predictive maintenance strategy for utility-scale solar facilities.”

Industry analysts have identified several common problems for solar arrays, noting it’s important for any predictive maintenance system to be able to spot those issues. Optimizing the performance of solar arrays can include the use of data analytics to identify areas where panels are not receiving enough sunlight, and the angle and positioning of equipment can then be adjusted to improve efficiency.

Archie Roboostoff, vice president of Software at Tigo Energy, told POWER the company “has a leading solar monitoring and management platform that flags issues within a system to fleet operators. When combined with Tigo predictive modeling technologies, specifically through our EI Predict+ software, advanced AI and ML technologies can flag many issues before they even happen.”

Said Roboostoff, “These capabilities are designed to forecast electricity generation and consumption at individual endpoints and aggregated energy portfolios automatically. The goal is to provide utilities, IPPs [independent power producers], and grid planners with accurate, scalable, and robust insights. Every day, we outline trending diagnostic information to our customers, along with the ability to anticipate and identify equipment failure before a piece of equipment fails. Tigo is currently working on next-generation technologies to help operators spend even less time fixing issues and more time maximizing power generation.”

Bullock said, “An effective anomaly detection system must be able to identify deviations in power output, temperature fluctuations, voltage or current irregularities, and irregular patterns from the anticipated norms within a solar array. Low-quality data is also a problem across the industry. Detection system capability extends to recognizing concerns like diminished output from individual panels, the emergence of hotspots due to shading effects, or disparities in power generation across the array’s panels. By methodically monitoring a spectrum of critical parameters and examining pattern irregularities, anomaly detection systems proactively uncover potential issues. This proactive identification equips maintenance teams with the foresight to make timely decisions related to the equipment, ensuring that solar arrays operate at their peak performance levels.”

Field Work

Successful predictive maintenance programs also benefit workers in the field (Figure 2), those specifically tasked with maintaining and/or repairing solar power equipment. Those in the industry agree that being armed with more information provides time and cost savings for utilities and other operators.

2. Part of a successful predictive maintenance program is the ability to ensure field workers have the most relevant and timely information about the equipment they’re repairing or maintaining. Source: Pexels 

“Predictive maintenance programs empower field workers by providing them with timely and relevant information about the condition of solar power equipment and an ongoing log of preventative measures taken,” said Roboostoof. “The maintenance crews will be familiar with the equipment and understand what maintenance activities have been done. By doing this, each maintenance crew will be more efficient and safer, and activities that focus on maintenance ultimately benefit both the workers and the overall performance of the solar power system.”

Luis Gerardo Guerra, a Germany-based researcher at DNV GL, told POWER several factors impact development of a predictive maintenance program. “It depends on the resources that are available to each utility or operator,” said Guerra. “The development of a predictive maintenance program requires a combination of multi-disciplinary skills, including data engineering, software development, domain knowledge, data science, and others.”

Guerra said that “most, if not all, predictive maintenance systems are based on normal behavior models, that is, a model that can describe how an asset should behave under normal operating conditions. However, defining when an asset is operating under normal conditions can itself be a challenge. Although it is possible to perform a completely data-driven analysis of operational data in order to filter out unwanted conditions, this strategy may not be entirely effective, which is why we believe that having access to high-quality O&M logs where the most important events related to an asset are recorded is essential to developing a successful normal behavior model.”

Guerra told POWER, “The use of high-quality O&M logs is not just restricted to filtering out potentially unwanted conditions from operational data, but it is also the cornerstone of an effective validation strategy of a predictive maintenance system. If we as developers of a predictive maintenance system do not know when an asset has failed [or malfunctioned], then we cannot adequately assess the performance of our system. For this reason, an adequate and thorough recording of the asset’s conditions is also crucial for a successful predictive maintenance program.”

Experience Is Important

The need to utilize all those factors means solar operators will often need an experienced team to build a successful data gathering, analysis, and predictive maintenance system. “Although possible for a large company, it would be difficult for a small- or medium-sized enterprise to gather all the necessary in-house expertise to develop its own predictive maintenance program, especially if it must be built from the ground up,” said Guerra. “On the other hand, third-party solutions bring immediacy, that is, a validated [or near validation] system that has been extensively tested with other clients and benefits from years of development, something that not every utility or operator may be able to afford. From our experience, I can say that most operators we have spoken to have shown preference toward bringing in a third-party solution rather than an internally developed one.”

Guerra noted, though, “Regardless of the choice, I do believe that it is essential for an operator or utility to have some internal talent that understands the concepts behind a successful predictive maintenance program. They are the ones who should direct the internal strategy and establish realistic goals for the program. Another important aspect is obtaining feedback from final users. After all, they are the ones who know the asset best and will ultimately make use of the system. Integrating their comments and suggestions will make sure that the tool under development is successfully deployed and meets their needs.”

Roboostoof told POWER, “The cumulative effect is a workforce that operates with increased efficiency and safety. Armed with a thorough understanding of the equipment and a record of past maintenance activities, maintenance crews can approach their tasks precisely and confidently. Finally, with remote monitoring and deep, predictive analysis, maintenance crews can optimize truck rolls and ensure they have any replacement components with them for site visits, both of which greatly reduce one-off service visits by bundling work tasks.”

Darrell Proctor is a senior associate editor for POWER (@POWERmagazine).

SHARE this article