Predictive maintenance (PM) is one of the much-touted features of the smart grid. Connected sensors collect data which is then used to analyze the state of the grid in real-time. PM can not only detect existing problems but also forecast equipment failure. Machine learning algorithms compile historical performance data and identify patterns which enable the prediction of future pressure points. The idea is to help grid managers schedule maintenance only when necessary, which in an ideal world should help to reduce downtime and costs.
European Union: a slow transition
Many nations around the world are gradually adopting smart grid technologies to better support the switch towards renewable energies and increasingly distributed energy resources. European utilities are no exception to the rule. The European Commission adopted the Digitalization of Energy Action Plan in 2022, which contained more than 20 key actions to support smart grid deployment and favor investment in the area.
However, the International Energy Agency (IEA) warns in its 2025 report on world energy investment that in Europe “grid upgrades need to keep pace with the rapid expansion of low-emissions electricity generation. Annual spending on grids is set to exceed USD 70 billion in 2025, doubling the amount spent a decade ago.”
However, the agency continued, “Investment has not yet matched the pace of clean energy deployment, leading to inefficiencies such as long connection queues and difficulties in transmitting cheap renewable electricity from southern parts of the European Union to high-demand areas.” One of the main problems is therefore the rapid expansion of renewable energy generation sources without corresponding upgrades in storage and grid infrastructure.
A cautionary note is also sounded by Chisimdiri Anyaogu, who is an Engineer at Irish utility ESB, on the way PM has been implemented across Europe. “There appears to be more extensive installation and use of PM in the Middle East, while European and American counterparts are slowly following. The diverging understanding and priorities between stakeholders at European level, governmental and regulatory authorities, utilities and operators as well as manufacturers and vendors result in varying interpretation of requirements and expectations, different implementations and complexity of systems, and lack of clarity on recoverable costs and market penetration.”
That’s why, he says, “The industry would benefit from standards to dispel misunderstandings and provide a common platform to move from theory to practice.”
Use cases in Finland and Ireland
To get a better picture of the European state of play, we have asked ESB and another utility Helen Electricity Network, based in Finland, about their use of PM tools.
Helen Electricity Network is a forward-looking distribution system operator (DSO) in the city of Helsinki. According to Helen expert Mika Loukkalahti, the company has 400 000 customers and is the third largest DSO in Finland. “The electricity network in Helsinki is heavily meshed and includes a lot of redundancies and automation. In addition, medium voltage (MV) and low voltage (LV) networks are totally cabled,” he describes.
He thinks that’s why Helen’s supply reliability levels are probably amongst the best in Europe. “The system interruption duration index (SAIDI) per customer per year is typically only two to three minutes,” Loukkalahti says. It also one of the reasons why his company has not fully implemented PM across the whole network “because there is no need for it.”
PM used for cables, transformers and substations
PM is nevertheless used to monitor specific systems and equipment. The newest 110 kilovolt (kV) high voltage cable systems, which are being introduced, have been equipped with partial discharge (PD) and temperature sensors. Helen has also invested in over 10 smart cable guard systems for the MV systems, which measure PD in the cables online. PD is a localized dielectric breakdown which is not visible in solid insulation situations. This can eventually lead to a breakdown of insulation, which can cascade into power outages. The online PD measurement systems enable Helen to foresee evolving faults and, if required, transfer power supply to other systems.
“We have also introduced PM for power transformers which operate between high voltage and medium voltage networks. Some of our older transformers are now equipped with gas analyzers as part of the maintenance process,” according to Loukkalahti. Gases in transformers may occur from the external environment or from chemical reactions internally. And according to this electrical testing equipment supplier, “these gases signal potential issues like overheating, arcing, or insulation breakdown. They may be invisible to the naked eye, but their presence is a tell-tale sign of underlying problems.”
Helen has also automated substations, using intelligent electronic devices as well as serial or ethernet communications. “One hundred per cent of our primary substations are automated and one third of our secondary ones. The systems can partially predict their internal faults and they can send alarm signals if a problem is detected which enables it to be fixed rapidly,” he adds.
The pros and cons of PM
Loukkalahti thinks that PM has been quite beneficial on the whole. “PM and its self-supervision functionality have brought many cost savings. There is no longer a need to carry out periodical maintenance of substations as faults are immediately flagged up. We have also found that the need for periodical relay protection tests has decreased significantly. We used to perform these tests every three years but with PM, these only have to be done every eight to 12 years.”
In addition, Loukkalahti says, “It has also enabled us to be alerted to evolving faults in our transformers before these faults become more severe. We also have an automated meter reading (AMR) fleet which has helped us better target our maintenance for the right meters as well as predict their lifetime and when the next gen meters should be rolled out. We have used AI-based tools to analyze from the AMR-signals the lifetime and condition of the meter fleet. AMR is used for energy metering and also power quality alarms, all remotely read.”
There have been a few drawbacks, however – mostly to do with the different lifespans of smart grid tech and legacy transmission equipment. “An example of this is the gas analyzers we use for our transformers. Their lifespan has been around ten years – albeit fast improving to 15 to 20 years – when transformers work for around 50 years. We have to therefore carefully assess the cost of using these analyzers which has to be lower than the savings they enable. This cost benefit analysis is something we do or each new system,” he acknowledges.
PM used for transformers and switchgear
ESB is the Irish electricity distribution system operator (DSO), distribution asset owner (DAO), and onshore transmission asset owner (TAO). It comprises 3 000 employees who build and run medium and low voltage electricity systems, including power stations, overhead lines, poles, and underground cables in Ireland.
The company is gradually increasing its use of predictive maintenance for critical assets, particularly transformers and switchgear. “Rather than relying solely on time-based inspections or responding reactively to faults and alarms, we use sensors and online monitoring systems to continuously collect real-time data from these assets,” Chisimdiri Anyaogu explains. “Additional data such as criticality, utilization, and profiling of assets arising from failure investigations, helps to identify assets to prioritize for retrofitting of sensors and data collection systems”, he adds.
As a result, ESB “has been able to extend the operational lifespan of critical assets by catching several issues in their early stages already, allowing us to avoid unplanned outages that would otherwise have disrupted supply.” PM has also allowed ESB to avoid significant costs arising from the early replacement of electrical assets.
Reasons for adopting PM
The company decided to move to predictive maintenance because “routine-time-based maintenance schedules, while familiar, did not adequately identify problems early enough, which created a need for a smarter way to monitor asset conditions and provide for data driven decision making.”
Another factor which prompted the move was that other competing companies in the field were also adopting PM tools. And just as important was increasingly widely distributed electrical assets – as a result of the increasing reliance on renewable energy sources like solar PV or wind particularly as these sources expand into more remote locations.
“Getting teams and equipment to remote areas is costly and time-consuming, straining resources and efficiency. Predictive maintenance, enabled by remote monitoring capabilities, provides a way to manage these geographically dispersed assets more effectively, reducing the need for frequent on-site visits and allowing us to prioritize interventions based on actual condition data rather than proximity or scheduling convenience,” adds Stewart Flood, the Lifecycle Specialist at ESB.
What are the drawbacks?
ESB met a number of implementation challenges, some of which are similar to Helen’s, and which have to do with the combination of legacy assets with digital AI-based tools. “Our grid relies on infrastructure that predates modern digital systems, and retrofitting IoT sensors and PM platforms onto legacy assets is rarely straightforward. In most cases, we are also required to work within platforms mandated by equipment vendors, or specific protocols which limit our ability to integrate data into our own systems on our own terms; think of DVD vs Blu-Ray or a more modern example could be Amazon vs Microsoft vs Google,” Anyaogu says.
The uneven performance of sensors can also be a problem. “Sensors in field environments can sometimes produce inconsistent readings, reducing confidence in the data. Knowing when to trust an alert or indicated fault can take time. Gas monitoring systems utilized on gas insulated switchgear initially provided false alarms due to ambient temperature changes not compensated for in the early configurations, for example,” Flood explains.
The cost of implementation is also something to think about, as well as the cultural aspect of moving to a new way of working, which must not be underestimated.
“The upfront investment in PM is considerable, covering not just hardware alone, but the system and expertise needed to make sense of the data. Sensors require ongoing maintenance and periodic replacement, which adds to operational costs that aren’t always anticipated. Internally, bringing non-technical stakeholders along the journey requires deliberate effort. PM involves a shift in mindset and making the case to those outside engineering,” Flood says.
The absolute need for IEC Standards and system thinking
Both utilities agree that standards are indispensable to pave the way for a more seamless adoption of these new AI-based tools. Helen and ESB identify several standards that are already used widely, including some in the IEC 61850 suite of standards, which cover core standards for the smart grid, notably for substation automation. Additionally, ESB points to the relevance of IEC 60422, the standard used for the supervision and maintenance of insulating oil and electrical equipment, and IEC 60599, which provides a guide to the interpretation of dissolved and free gas analysis. Helen points to the IEC 61869 Standard for instrument transformers.
“IEC 62443 is a good backbone relating to the cyber security measures for the process automation systems. Of course, the cyber security area is vast, and the co-operation of IT and OT sectors is mandatory, “Loukkalahti says.
He also mentions a CENELEC standard, EN 50549, which defines the technical requirements for connecting distributed energy resources (like solar PV inverters and battery storage) to the electricity grid.
But both also say that new standards are required. For Loukkalahti, a new battery storage standard referring to EN 50549 is on his wish list. For Anyaogu and Flood, the list is more detailed. “While there are practical examples such as IEC 61968 which fosters the information exchange among electrical power systems supporting business functions, the changing data-based landscape could benefit from a standard focusing on a general architectural template for transmission and distribution systems, for both operators and utilities,” says Flood.
“There are standards such as IEC 61850, IEC TR 23188, a technical report which specifies the edge computing landscape of cloud technology, and IEC 62443, but there does not appear to be an all-encompassing standard that ties these together for the purpose of monitoring HV station assets,” Flood continues.
Some standards help to monitor the performance of smart sensors: IEC 60747-19-1, which deals with the control scheme of smart sensors (bidirectional communication/control scheme for smart sensor devices). Another is IEC TS 60747-19-2, which specifies sensors and power supplies to drive smart sensors for low power operations. But these are not integrated in a general architecture template for the smart grid.
The IEC Quality Assessment System, IECQ, one of the four IEC Conformity Assessment (CA) Systems, proposes an approved component certification, which can be applied to various electronic components, including sensors that adhere to technical standards or client specifications accepted within the IECQ System.. But this is not specific to the grid.
More work needs to be done by the different technical committees and CA Systems standardizing and assessing the various techs used in electricity networks to interoperate on matters relevant to the grid. Some of this system level standardization has been taken on board by the Systems Committee for Smart Energy – SyC Smart Energy – which is a starting point.
Author: Catherine Bischofberger
The International Electrotechnical Commission (IEC) is a global, not-for-profit membership organization that brings together 174 countries and coordinates the work of 30.000 experts globally. IEC International Standards and conformity assessment underpin international trade in electrical and electronic goods. They facilitate electricity access and verify the safety, performance and interoperability of electric and electronic devices and systems, including for example, consumer devices such as mobile phones or refrigerators, office and medical equipment, information technology, electricity generation, and much more.
The post AI for the electricity grid appeared first on pv magazine Global.
