Status
Working prototype
2025 · Working prototype
Smart Energy Meter
Per-appliance energy monitoring
Per-appliance energy monitoring system for home appliances.
Ownership
End-to-end
Stack
ESP32MicroPythonFlaskMQTTLightGBM
Repo
Open repositoryProblem
Semester project. Stick a compact meter on each appliance. It measures voltage, current, power, and energy using an ESP32 with ADC sampling. A calibration workflow stores sensor offsets and scaling factors in JSON. OLED display shows readings locally, with touch-based screen navigation.
Implementation
Evidence
Full pipeline working: hardware sensing, MQTT telemetry, live dashboard, relay control, and ML-based consumption prediction and spike detection.
Constraints & next step
Data goes over MQTT to a Flask backend that streams it to a browser dashboard via Server-Sent Events. A bounded deque buffer pushes the latest readings to connected clients. Also supports remote relay control: flip a switch from the dashboard and the appliance turns off.
An ML model (LightGBM) predicts hourly power consumption and flags unusual spikes. Isolation Forest handles anomaly detection with a Z-score fallback for when the model does not have enough history. The service loads pickled models per appliance and runs feature engineering on a sliding window of lag features and rolling stats.