IoT UPS Health Monitor with Backup-Time Dashboard

A UPS is the one device everyone ignores until the power fails — and then discovers the battery died months ago. Offices lose unsaved work and routers drop because nobody knew the backup time had degraded to minutes. The failure is not the UPS; it is the absence of visibility. This project adds that visibility as a non-invasive monitor: an ESP32 node senses mains presence, battery terminal voltage and load current, derives charging/discharging state and state-of-charge trend, and estimates remaining backup time from the live load. All of it streams to a purpose-built web dashboard — battery gauge, backup-time readout, mains/battery state timeline, and a log of every power event. The hardware is the sensing node; the software is the dashboard, and both are built for this project.

1. The ESP32 node connects isolated sense leads to the mains input, the battery terminals and a current sensor on the UPS output.
2. Firmware samples voltage and current, classifies the operating state (mains present, charging, on-battery, battery low), and estimates backup time from battery voltage and live load.
3. Readings publish over Wi-Fi (MQTT/HTTP) every few seconds to the cloud backend feeding the web dashboard.
4. The dashboard renders the battery gauge, backup-time readout, state timeline, voltage history chart and the timestamped power-event log.
5. Crossing the low-battery or long-outage thresholds triggers phone/push alerts to the owner.
6. Long-term voltage-under-load curves accumulate in history, letting the owner spot battery degradation trends.
7. The student calibrates the voltage sense channels against a multimeter with the included procedure so dashboard readings match reality.

Tech stack:

• ESP32 development board (Wi-Fi)
• Isolated mains-presence sensing + battery voltage dividers
• Hall/ACS-class current sensor on UPS output
• 0.96-inch OLED (local status at the node)
• Arduino IDE (C/C++ firmware)
• MQTT/HTTP telemetry to cloud backend
• Single-file HTML/CSS/JS dashboard (Chart.js-style gauges and charts)
• Push alerts (Blynk/ThingSpeak class)

• [Mains and battery sensing] Isolated sensing of mains presence, battery terminal voltage and load current gives the node everything needed to judge UPS health without opening the UPS.
• [Backup-time estimation] Firmware combines battery voltage trend with live load current to estimate remaining backup time, shown as a prominent dashboard readout that updates as load changes.
• [Charge/discharge state tracking] The node classifies MAINS / CHARGING / ON BATTERY / BATTERY LOW states and the dashboard renders them as a live status timeline.
• [Power-event log] Every mains failure and restoration is timestamped and logged, so the owner can see outage history and the dashboard can flag unusually long outages.
• [Battery-health trend] Long-term battery voltage curves under load reveal a degrading battery weeks before it fails a real outage — the dashboard keeps this history.
• [Low-battery and long-outage alerts] Phone/push alerts fire when backup time drops under the configured threshold or the mains stay down beyond the set duration.
• [Working web dashboard] A single-file HTML/CSS/JS dashboard with gauges, charts and the event log — built for this project and screenshotted for the listing gallery.

• ESP32 UPS sensing node with isolated sense inputs and OLED
• Current sensor and battery/mains sense wiring kit
• Firmware with state classification and backup-time estimation
• Working single-file web dashboard (gauges, charts, event log)
• Cloud telemetry + alert setup guide
• Buyer-run voltage calibration procedure
• Project report PDF (sensing design, backup-time method, methodology)
• PPT presentation for final review
• Viva Q&A preparation document (battery behaviour, sensing isolation, IoT pipeline)

• The monitor observes the UPS externally; it cannot change UPS charging behaviour or revive a dead battery.
• Backup-time estimation is approximate — it depends on battery age, temperature and load type, and must be validated by the student with a timed discharge test.
• Battery-health trending needs weeks of history to be meaningful; a fresh install shows only instantaneous state.
• Voltage sense channels must be calibrated against a multimeter; uncalibrated readings are indicative only.
• The node needs continuous Wi-Fi for live dashboard data; the local OLED keeps working regardless.