ÅF uses AI to streamline major energy customer
Advanced AI technology helps a major energy customer to predict and optimise the maintenance of its hydroelectric power plants. ÅF has developed an innovative Predictive Maintenance solution that streamlines the production chain and reduces the risk of expensive downtime. “This is the future of the industry,” says Azad Noorani, Lead Data Scientist, at ÅF Digital Solutions.
Rapid digitalisation – and the possibilities offered by the interconnected society’s ability to collect, store and process large amounts of data – has revolutionised the industry. Using the right models and powerful analytical tools, it is now possible to identify and swiftly rectify weak points in a production or process – before costly failures occur.
One such tool is ÅF's innovative product Predictive Maintenance.
“This makes it possible to avoid unforeseen downtime," says Azad Noorani of ÅF Digital Solutions.
Predictive Maintenance is based on machine learning and is an application of advanced AI technology that, among other things, provides the user with information about vulnerabilities in the operating chain and flags necessary maintenance.
"It's all about getting the right information at the right time," says Azad
One major energy company, Sweden's leading producer of renewable electricity, was quick to see the benefits of being able to plan and schedule the maintenance of its hydroelectric power plants. A concrete example is the large generators that are cooled by water pumps. If a pump fails, there is a risk that the generator will overheat and break down.
Up until the beginning of 2018, these needed to be manually inspected at least four times a year. With ÅF's Predictive Maintenance solution, the need for these periodic inspections has been significantly reduced.
“Production is more even while at the same time a great deal of money is being saved. When the system is up, there is no human interaction. Everything is automated until an issue needs to be addressed,” says Azad Noorani.
A prerequisite for the optimal application of Predictive Maintenance is the availability of large amounts of data. The larger the amount of available data, the better the quality and accuracy of predictions.
"In this case, we used historical data for parameters such as water flow and energy consumption and machine learning to build unique, customised models for each cooling system. This provides a completely different level of control over what’s happening in the hydroelectric power plants, says Azad Noorani.
The results are visualised in real time on a control panel that indicates if, for example, a generator is at risk of overheating.
“Maintenance is streamlined and any downtime can be scheduled to minimise impact,”, says Azad Noorani.
The number generator breakdowns has been significantly reduced, with a 10 to 15 percent reduction in downtime; something that has repaid the energy company’s investment.
"They are very satisfied with the solution," says Azad Noorani.
He sees enormous opportunities for exploiting this technology, which is based on an idea originally developed at NASA.
“it can be used wherever machines are involved, from cars to all types of industrial processes.”