The fault-prediction platform for power networks

AI fault prediction in electric power systems uses machine learning on the disturbance recordings substations already produce to detect the early signatures of developing faults — and warn before they cause outages. Eneryield’s IntelliView® analyzes every recording with physics-informed AI built on 400 expert-designed waveform features, and reports which asset is at risk, the predicted root cause and the likely source — on average 15 days before the fault. Unlike condition monitoring that needs new sensors, it runs on standard relay, fault-recorder and power-quality data (COMTRADE or IEC 61850) — no proprietary hardware.

Often not — and never anything proprietary. IntelliView® is the analytics layer: it runs on standard digital-substation measurement infrastructure — relays, fault recorders and power-quality meters writing COMTRADE records or IEC 61850 sampled values — so in many deployments nothing new is installed at all. Where a substation isn’t digitalized yet, we supply the necessary measurement hardware together with our partners: ABB substation digitalization with IntelliView® on the CogniEN platform, or EcoPhi QMU 800 merging units streaming at up to 2 MHz. That enables the analysis — and brings the wider benefits of a digitalized substation with it.

Wherever your policy requires. Choose the hosted cloud service — we host, monitor, maintain and update, and your users need only a browser — or run IntelliView® on-premise with Docker and Kubernetes inside your own environment, including fully offline installations with no external connectivity. Either way: TLS in transit, encryption at rest, role-based access, and each organisation’s data in its own schema. Security documentation and penetration-test summaries are available for review.

Overhead lines, underground cables and submarine cables, across transmission, distribution and industrial grids. Reference deployments range from rural distribution feeders in Finland to transmission cable, reactor and transformer banks in the United States and 33 kV and 69/13.8 kV substations in Saudi Arabia.

With a six-week replay on your own data. You send historical disturbance recordings and your fault log; we replay the models across that window and report what we would have flagged, when, and why — measured against success criteria you set before we start. The fee is scoped up front, nothing is installed, and the result arrives six weeks after we receive the data. An evidence-based go/no-go before any deployment decision.

Measured in live service, not in a lab: 64 faults predicted in advance over five years at Vaasan Sähköverkko, nine on NYPA transmission assets, and five within months of go-live at Saudi Electricity Company — more than 100 outages predicted in advance across all deployments, with an average warning of 15 days and horizons up to three weeks. And honestly: a lightning strike or a dug-up cable cannot be predicted by anyone. The faults that matter commercially develop slowly — insulation, connectors, earth contact — and those are the ones we catch. The replay measures accuracy on your grid, against your own fault history, before you commit.

Every warning carries its predicted root cause (insulation degradation, foreign object intrusion, evolving earth faults and more), the likely source, and the time horizon. Behind each risk value sits the list of recordings that produced it, ranked by contribution — and each recording opens in the inspector with RMS traces, harmonic spectrum and impedance, events segmented and labelled. The analysis runs on 400 expert-designed waveform features grounded in electrical physics. The engineer decides — not the model.

Yes. On the input side it reads COMTRADE records and IEC 61850 sampled values from your relays, fault recorders and power-quality meters — and it runs as an application on the ABB CogniEN platform where ABB is your digitalization supplier. On the output side, risk values and alarms are available over a REST API to feed your SCADA/ADMS, historian or work-order system, alongside scheduled reports and threshold alarms by email.

€3,000 per monitored bay — a one-time perpetual licence, with service, support and model updates included for the first three years. A 20-bay substation is €60,000 one-time, hosted or on-premise. Deployment starts with a six-week replay on your historical data, scoped and priced before any installation — so the business case is built on your own numbers before you commit to anything.