Multiple Results in Six Months
“After six months, the number of unnecessary smart meter replacements was reduced by 60,000, amounting to cost savings of approximately six million euros and saving more than fifty man-years”, says Dennis van de Vorst, Machine Learning Engineer at Aurai, discussing one of the most significant outcomes of his recent work for Dutch network operator Enexis. In close collaboration with several Enexis departments, Dennis, along with David Steenmeijer, Analytics Translator at Aurai, recently completed a project to optimize the network operator’s smart meter maintenance.
“The cost savings were achieved during the six-month project and are expected to continue to accrue over time”, Dennis adds. “Furthermore, the project demonstrated that the notice period for meter defects can be extended from non-existent to weeks or even months. Additionally, the predictive maintenance strategy enables Enexis to become compliant with Dutch legislation regarding the timely replacement of malfunctioning smart meters.”
The Challenges of Smart Meter Maintenance
In 2015, Enexis initiated the large-scale deployment of smart meters, successfully replacing 2.6 million electricity meters and 2 million gas meters. Dennis explains: “One of the current challenges relates to the sudden, large-scale occurrence of defects in smart meters. While the number of functional defects typically remains steady and manageable, it can sporadically increase a hundredfold within days, putting an immense strain on repair staff.” Moreover, Dutch regulatory requirements mandate that eighty percent of meters exhibiting measurement inaccuracies must be replaced within ten days. With sudden increases that far exceed the norm, this becomes impossible, posing a formidable challenge to legal compliance.
Making Predictive Maintenance Work
In 2023, Enexis partnered with Aurai to redevelop the maintenance strategy for smart meters, aiming to redistribute the staff workload over longer periods of time. The fact that, during these situations, less than forty percent of meters were being replaced in a timely manner served as an additional incentive to improve the maintenance strategy.
“Aurai supported Enexis in unlocking the full potential of predictive maintenance”, Dennis reflects. “Predictive maintenance uses data, analytics, and machine learning to predict when equipment or machinery is likely to fail. The proactive approach allows an organization to perform maintenance just-in-time to prevent failure. It tends to be more cost-effective than traditional or reactive maintenance, where equipment is serviced or repaired only after it has already failed.”
Enexis has been using predictive maintenance since the introduction of smart meters, employing a set of rules to indicate when a meter might be faulty. The rules are developed based on manufacturer specifications dating back from 2015, and are regularly updated. “The vast amount of data currently available provides an excellent opportunity to refine and enhance the predictive maintenance strategy, using real-world results”, Dennis says. “We identified different valuable data sources, analyzed the existing business processes to determine areas for improvement, and utilized machine learning and statistics to develop an early detection system.”
Smart Meters as Data Assets
Smart meters log a range of events, from critical errors to minor changes such as modified clock settings. Furthermore, Enexis investigates returned defective meters. Together, the meter logs and research results create a massive data set with valuable information.
“We identified relationships between events that can be used to develop an early detection system. With these findings, it is now possible to predict weeks or even months in advance whether an individual meter or group of meters will fail. This notice period offers two major benefits. Firstly, it provides a forecast of the workload, allowing Enexis enough time to bring in external contractors during peak demand, something that was previously unfeasible. Secondly, knowing in advance which meters need to be replaced allows for better route optimization for the electricians that will replace the meters.”
The Aurai team also used data to determine if a meter is truly defective, or if it can remain in operation for longer. In collaboration with Enexis’ meter experts, we discovered large populations of meters that were replaced prematurely. “The correlations identified in this data alone account for 60,000 meters that were scheduled for replacement, even though a vast majority of them are likely to remain in good condition for over a year.”
From Predictive Maintenance to Grid Loss Predictions
Over the course of six months, Dennis and David worked closely with Enexis to make its predictive maintenance strategy for smart meters fit for the future. “I really enjoyed the collaboration with Enexis, appreciating the open atmosphere, the workplace dynamics, and the frequent rounds of table tennis during lunch breaks. Most importantly, of course, the project has been successful and productive, leaving us eager for more. Therefore, we feel fortunate to have already started a follow-up project, which concerns grid loss predictions that can be used to stabilize the electricity network. I’m confident that, once again, this will be a challenging and unforgettable experience.”
At Aurai, we specialize in bridging the gap between data and business, developing effective data strategies to maximize business value. Are you interested in learning more about our solutions for your projects? Please feel free to contact us!