Artificial Intelligence (AI) technology has delivered a wide range of compelling benefits to industry, arguably the most impactful of which has been the ability to predict the failure of expensive assets and prescribe actions that can mitigate, or even eliminate, these failures. One of the primary beneficiaries of this technology to date has been renewable energy. The various applications of AI to renewables—whether wind, solar, or battery systems—are known collectively as Asset Performance Management (APM) systems. In the APM space, AI provides invaluable insights that contribute directly to more efficient operation of wind (onshore and offshore), solar, and battery storage systems.
“Our goal is to understand and quantify performance, and then determine what value we can contribute using that information. How can we understand which assets are underperforming and provide recommendations about what can be improved? What problems can we proactively remedy?” says Clara Midgley, SparkCognition’s Renewable Suite Product Manager. “We envision a world where you can create a renewable generation plant that’s self-optimized using tools that enable operators to make better decisions on how to maximize asset utilization.”
AI-enabled APM tackles critical societal challenges
The United Nation Climate Action home page spells out several important statistics that quantify in stark terms the global impact that renewable energy is delivering now and is expected to contribute in the coming few years.
- About 29% of global electricity generation currently comes from renewable sources.
- Renewable energy is the cheapest form of generation in most areas of the world today. For example, the cost of solar fell by 85% from 2010 to 2020, while costs of onshore and offshore wind fell by 56% and 48% respectively.
- $4.5 trillion/year needs to be invested in renewables to achieve the UN’s net-zero emissions goal by 2050.
With daunting goals and achievements like these, can AI truly deliver the sort of impact that’s required? The short answer is yes, with the primary caveats being those of commitment and money. But the first requirement for renewable assets is that they operate continuously, reliably, and efficiently.
“There are lots of different use cases,” Midgley says. “One critical area where AI is used is in backcasting and forecasting. You have an expectation of what sort of weather you’re anticipating, what market conditions you expect, and you can quickly learn how a particular asset is going to react to those conditions. You can then create a production forecast, or maybe a price forecast for a market, or develop ways to optimize a battery performance forecast. You can also quantify energy loss under downtime conditions. Those energy loss calculations are really useful for understanding your biggest hits against performance. That, in turn, allows you to assess where you can get the most bang for your maintenance buck, where you should focus first to improve asset performance through prescriptive actions.”
Tracking and reacting to individual performance metrics does not, per se, require artificial intelligence. This has been done for decades using approaches like statistical process control and rules-based techniques. Where AI makes an immense difference is when the goal becomes predicting and prescribing performance-optimizing actions on large, complex collections of components as a holistic system, a requirement vastly beyond the capabilities of earlier decades.
AI and humans working together to optimize renewable energy production
“It’s all about scalability,” Midgley notes. “The main benefit of these models is their ability to enable users to scale their performance management activities across a large, diverse group of assets, including the many OEMs that make up large energy systems. When you have a lot of alarms coming in, you need to know which assets have gone down (or are likely to do so), why they’ve failed, and what are your potential courses of action. Asset managers need to have all this information across these different data sets at their fingertips, so that they have the information they need to make a maintenance or repair decision.”
At the end of the day, though, humans are still going to be in the loop on these actions,” Midgley adds. “We are, after all, dealing with critical infrastructure here. You don’t want everything to be fully automated, and there are also different regulatory and compliance requirements if, for example, you’re going to send a control set point to a piece of equipment. There are a lot of guardrails and different ways to limit what the system can do on its own.
“Another important element of control and credibility is that we always want to be able to show our work, explain how we arrived at a conclusion. If somebody sees a summary of the performance of a solar project, how do they know it’s correct? If I tell a plant manager that the availability of a system is 95%, I need to be able to show them where the other 5% went based on actual data. Our tools help managers understand and believe in the applications. But ultimately they just care about getting the job done. They want a tool that lets them work efficiently and accurately.”
To learn the whole story about SparkCognition’s role in advancing renewable energy technology using the power of AI, read our renewable energy eBook and our solar energy use case.
