Know the design
A deep understanding of renewable energy systems and a multi-physics simulation pipeline for the key components — power modules, capacitors, DC busbars and more.
Technology
AI + physics, a decade deep
Our founding team has spent more than a decade studying the reliability of power electronics, accumulating extensive testing data, algorithm models and know-how. Our engine uses physics-informed machine learning to evaluate multi-physics performance and reliability — efficiently and accurately.
Why it matters
The hard problem behind reliable renewables
Power electronics is the key component in every renewable energy system — and the electrical converter has the highest failure rate in a wind turbine. Characterising its physical behaviour and reliability under realistic conditions is the problem everyone else has skipped.
Know the problems
Years of reliability testing built a knowledge base of failure mechanisms and lifetime models — understanding how problems happen, not just when.
Know the data
A library of physics-informed machine learning algorithms that recognise potential faults even from poor-quality raw data, and solve multi-objective maintenance optimisation.
Capabilities
One engine, three disciplines
Design
- Multiple simulation algorithms, integrated
- Hundreds of times faster multi-physics simulation
- Multi-objective algorithms for global optimal design
- Covers power modules, capacitors, DC busbars and more
Test
- A large accumulated base of experimental testing data
- Comprehensive assessment with failure and lifetime models
- Reliability evaluation under realistic conditions, based on failure physics
- Sample database from measured, simulated and synthetic data
O&M
- Works with existing sensor signals — no additional sensors required
- Real-time data streaming and online monitoring
- Unsupervised, adaptive online learning
- Data generation and few-shot learning for sparse data
Measured results
What the engine delivers
100×+
faster multi-physics computing than conventional simulation
90%
reduction in testing time, enabling rapid iterative design
3×
more potential faults detected, up to one week in advance
+2%
total power production from fewer unplanned shutdowns
Field validation
Working with Offshore Renewable Energy Catapult, our prototype — the world’s first fault-detection solution for electrical converters — was validated on a commercial offshore wind farm. It found three times more potential faults than the incumbent system and prevented one catastrophic failure.
The difference
A different starting point from conventional tools
Most monitoring products visualise sensor signals and raise threshold alarms. PowerSense starts from the failure physics of the equipment itself.
| Conventional approach | PowerSense | |
|---|---|---|
| Fault handling | Post-fault alarms on preset thresholds | Faults predicted up to a week in advance, precisely located |
| Coverage | Mostly mechanical components | Electrical and mechanical — including the converter, the most failure-prone part |
| Deployment | Extra sensors and retrofits | Existing sensor signals only |
| Design & test | Batch prototypes, long and costly lab campaigns | Virtual testing under realistic conditions, 90% faster |
| Lifetime | Not managed | Remaining-useful-life prediction and lifetime extension |
Defensibility
Protected, and pulling ahead
Patents
Three patents in progress with university technology-transfer support — covering converter condition monitoring, low-data training methods and operational data generation.
Data & know-how
Our datasets, source code and algorithms are protected by copyright and trade secret, built on testing data that took a decade to accumulate.
Continuous R&D
Ongoing research with Cambridge labs and industrial partners keeps the algorithm library ahead of the market.