Projects

AI tools and experiments for power systems engineering. Each one started from a real problem in my work.

Live

Transformer DGA Analysis

Dissolved gas analysis is one of the most powerful diagnostics for oil-filled transformers, but classical interpretation methods often disagree. This tool combines them with a trained model to give a single, explainable diagnosis.

  • Inputs: H₂, CH₄, C₂H₆, C₂H₄, C₂H₂, CO, CO₂ (ppm), plus optional history
  • Classical methods: Duval Triangle & Pentagon, Rogers ratios, IEC 60599, key gas
  • AI fault classification with a confidence score
  • Trend and rate-of-change analysis across samples
  • Plain-language recommendations and an exportable report
Fault classes PD, D1, D2, T1, T2, T3, normal ageing
Standards IEC 60599, IEEE C57.104
For Asset managers, maintenance engineers, oil labs
Beta

Relay Event Analysis

After a trip, engineers can spend hours going through oscillography and sequence-of-events records. This tool reads the event, explains what happened and checks whether the protection behaved as expected.

  • Inputs: COMTRADE (.cfg/.dat) files, relay event and SER logs
  • Fault type and phase identification (AG, BC, ABC…)
  • Fault inception, clearing time and breaker performance
  • Protection element pickup and operation review
  • Auto-generated event summary report
File formats IEEE C37.111 COMTRADE (1999 / 2013)
Analysis Phasors, sequence components, fault location estimate
For Protection engineers, control centers, commissioning teams
Soon

Next in the lab

Ideas I'm exploring. Have a problem you'd like solved? Tell me.

  • Transformer health index from multiple test results
  • Partial discharge pattern recognition
  • An LLM assistant for protection settings review
  • Load forecasting for distribution feeders