IoT Home Energy Monitor with Appliance-Level Tracking

Most households discover their electricity bill only once a month, as a single number with no breakdown — so nobody knows whether the geyser, the air-conditioner or the old refrigerator is the real culprit. Utility smart meters are rolling out slowly and give the distribution company data, not the homeowner. This project puts the measurement in the student's hands: split-core CT clamps snap around the mains and up to four individual circuit wires without cutting anything, a ZMPT101B-class module senses the mains voltage, and an ESP32 computes real power, power factor and cumulative energy. Readings publish over Wi-Fi to a cloud dashboard with live gauges and history charts, and the student configures daily budgets that trigger phone alerts. The engineering spans analog sensing, RMS math on a microcontroller, and an end-to-end IoT telemetry pipeline.

1. Split-core CT clamps are snapped around the mains incomer and up to four branch circuits; the voltage-sense module connects across the mains.
2. The ESP32 samples the voltage and current waveforms, applies the calibrated burden and phase constants, and computes RMS voltage/current, real power, apparent power and power factor per channel.
3. Energy is accumulated in non-volatile counters so totals survive power cuts and reboots.
4. Every few seconds the node publishes the readings over Wi-Fi to the cloud dashboard; a heartbeat keeps the device-online status fresh.
5. The dashboard renders live gauges, per-circuit power bars, and daily/weekly/monthly energy charts with tariff-based cost estimates.
6. When a configured budget or per-appliance threshold is crossed, the cloud service sends an alert to the student's phone.
7. The student calibrates each channel once against a known load (e.g. a 100 W bulb) using the included procedure, and the constants are stored in firmware.

Tech stack:

• ESP32 development board (Wi-Fi)
• SCT-013-class split-core CT clamps (mains + 4 circuits)
• ZMPT101B-class AC voltage sensing module
• Burden resistors + bias conditioning per channel
• 0.96-inch OLED display
• Cloud dashboard (ThingSpeak/Blynk class, MQTT/HTTP)
• Arduino IDE (C/C++ firmware)
• Tariff-based cost estimation on the dashboard

• [Non-invasive CT sensing] Split-core SCT-013-class clamps snap around live conductors without cutting wires, so the monitor installs on a real home distribution board safely and reversibly.
• [Mains + 4 circuit channels] One CT on the mains plus four on individual appliance circuits enables appliance-level tracking — the student can see the geyser's 2 kW signature separately from the baseload.
• [True power computation] The ESP32 samples voltage and current waveforms together and computes real power, apparent power and power factor — not just current — with burden-resistor and calibration constants documented.
• [Live cloud dashboard] Power, voltage, current and power factor stream over Wi-Fi (MQTT/HTTP) to a dashboard with live gauges and daily/weekly/monthly energy charts.
• [Cost estimation] The student enters their tariff slab and the dashboard converts energy into rupee estimates per circuit and for the whole home.
• [Usage alerts] Daily budget and per-appliance power thresholds trigger phone/push alerts when consumption crosses the set limits.
• [On-device display] A small OLED shows live mains power and today's energy at the distribution board without opening the app.

• ESP32 energy-monitor node with CT inputs and OLED
• Set of split-core CT clamps + voltage sensing module
• Firmware with RMS/power computation and Wi-Fi telemetry
• Cloud dashboard setup guide with charts and tariff configuration
• Buyer-run per-channel calibration procedure (known-load method)
• Project report PDF (CT theory, power math, methodology)
• PPT presentation for final review
• Viva Q&A preparation document (CTs, real vs apparent power, power factor, IoT pipeline)

• Current readings are only as good as the calibration: each channel must be calibrated against a known load by the student; uncalibrated error can reach several percent.
• Split-core CTs measure AC current only and need the conductor centered in the core for best accuracy.
• Appliance-level tracking covers the wired circuits only — it does not identify individual devices on a shared circuit (no NILM disaggregation in this build).
• Power factor is computed from the sampled waveforms; heavily distorted loads (cheap SMPS) reduce its precision.
• The monitor needs continuous Wi-Fi for live data; readings are buffered only briefly during outages.