Cables: Leak in underground oil-insulated cables is found faster with machine learning
Repairs and maintenance of old oil-insulated underground cables in Copenhagen are necessary, but both costly and cumbersome with the current method. A digital algorithm is to make it much quicker and less expensive for Energinet to find cable leaks.
Kilometres of old electric cables wind their way under the streets of Copenhagen. They are nearly all insulated with a special type of oil which makes the cables expensive to maintain and repair. The leak detection method is, in fact, fairly cumbersome and slow. When, in January, one of the cables under Papirøen on the east bank of the harbour in Copenhagen developed a leak, this underlined the need to explore the possibility of smarter maintenance and repair to avoid having to dig up half of Copenhagen in the process.
Data analysis and tracers
t was therefore examined whether it is possible, using the machine learning data analysis method, to identify potential cable leaks without having to carry out repairs in accordance with the current procedure. Data from various factors such as pressure and temperature which affect the state of the cable were collected and resulted in an algorithm which can detect, with 90% reliability, that a cable has a leak.
“At the same time, Energinet is running a parallel project which is to help identify exactly where the defect has occurred,” says Steffan Morrison from Energinet. He explains that the leakage is identified by means of a tracer in the cables which is supplied concurrently with the continuous replacement of the oil around the cable:
“If there’s a leak, the particulates from the substance can be traced from the surface. When the data analysis shows that there may be a leak, the tracer can thus tell us where to dig to repair the cable.”
Faster identification is good business
It all leads to much faster identification of defects and thus much quicker repairs. Energinet expects that the method will result in cost reductions of up to DKK 1-2 million per case. In the near future, Energinet will therefore examine how the machine learning method can be used more operationally. This will be done through a prototype process.
So far, the indication is that this method will make it possible to commence troubleshooting and repair much quicker to the benefit of Energinet, consumers and neighbours to the potentially leaky cables.
Digitalisation in Energinet
Digitalisation is one of 4 strategic objectives in Energinet’s strategy “Energy across borders”.
In 2020, we have improved the ways in which Energinet solves its tasks through digitalization and innovation.
SUBSCRIBE TO REPORTS
Receive notification directly by mail.