EnergyWatch — industrial energy monitoring
Multi-protocol energy meter reading with Z-score anomaly detection. Know when consumption deviates before it becomes a cost problem.
What EnergyWatch does
Multi-protocol meter reading
Modbus TCP (EASTRON SDM register layout), M-Bus stub ready for real hardware, and a built-in simulator with daily/weekly seasonality for demos and development.
Z-score anomaly detection
Each reading is compared against its hour-of-week baseline (168 buckets, Mon 00:00 → Sun 23:00). Severity: LOW (3σ), MEDIUM (4σ), HIGH (5σ+). Baselines auto-recomputed weekly.
Correlation analysis
When GridSense reports a health score drop > 10 points while energy simultaneously rises > 15% above baseline, EnergyWatch raises a CORRELATION anomaly — a failing asset driving up consumption.
KPI dashboard & heatmap
Total kWh, cost in PLN, peak demand, anomaly counts. 7×24 consumption heatmap (hour × weekday) for pattern recognition at a glance.
Platform integration
EnergyWatch consumes GridSense health updates for correlation analysis. Detected anomalies are published to the Kafka energywatch.anomalies topic for downstream alerting and reporting.
MVP status
EnergyWatch is at MVP stage. Forecasting (SARIMA/Prophet), automatic tariff zone optimisation, and carbon footprint reporting are on the 2026 roadmap.