Using AI and machine learning to reduce faults on the LV network, working with the Energy Innovation Centre and SSE.
Faults on the LV network reported by customers account for 80% of LV costs and 50% of CMLs. This project will trial & develop the VoltNet tool, which aims to detect, classify and locate developing LV faults before they are apparent to customers. The tool monitors voltage and current at 100kHz and an AI algorithm does analysis and classification.
The SYNAPS (synchronous analysis and protection system) solution will be deployed in substations and feeder link-boxes or feeder pillars. It applies innovative algorithms to power waveforms in order to detect and classify fault events.
SYNAPS uses state-of-the-art advanced statistical signal processing and machine learning algorithms to identify unique features of LV feeder cable faults (including early transient so called ‘pecking’ faults). A high sample rate detector is then employed to identify faults, when a manifesting fault is detected the sensor records the fault waveform and transmits to the server software for further processing. The server software classifies fault type and location (target accuracy 3m) utilising Powerline Technologies (PLT) proprietary algorithms.
The technology works by analysing the waveforms of the electricity voltage 100,000 times per second, and identifying when anything slightly unusual is happening in the cables.
Artificial intelligence then compares each new measurement to the vast library of data to spot any trends, patterns or irregularities. Over time, the machine will build up its knowledge and ‘learn’ to recognise conditions that can cause a fault on the electricity network.
Synaps will be the first time that leading-edge AI technology and ‘big data’ have been used to improve network reliability. If the trial is successful and rolled out across the network it could potentially help halve the number of power cuts.
SYNAPS could enable DNOs to make significant reductions in the cost of LV network operation, replacing expensive and manpower intensive operations with automated procedures. The detection and location of faults at an early stage, before they become permanent, will facilitate proactive planned maintenance, rather than expensive reactive emergency action. Faults can be detected, classified and located before a fuse failure, giving the DNO the opportunity to repair the fault before or immediately after the first fuse failure.
The learning from this project will be about a new method to enable proactive LV fault management. This will be of value to all Network Licensees, as LV fault management is an issue affecting each one. If the SYNAPS system is able to accurately detect, characterise and locate faults, it will feed into a potential wider scale trial that will seek to deliver customer and network benefits.