Smart EV Charger Load Balancer with Dynamic Current Control and Dashboard

Apartment basements and campuses adding EV chargers hit the same wall: the building's sanctioned load cannot feed every charger at full current simultaneously, and upgrading the sanctioned load is expensive. The industry answer is dynamic load balancing — chargers share a site budget, throttling gracefully instead of tripping the main breaker — but students only ever read about it in application notes. This project builds a working two-outlet balancer around an ESP32: per-outlet current sensors feed a control loop that allocates the site's current budget between outlets, rebalancing within seconds when a vehicle plugs in, finishes, or the site limit changes. A companion dashboard visualizes per-outlet power, the site-limit gauge, the allocation history and a session log. The prototype operates at safe low-voltage demo scale with scaled current limits, which keeps the build student-safe while the algorithm — measurement, allocation policy, rebalancing dynamics — runs exactly as it would at full scale.

1. Each outlet's ACS712 sensor is sampled continuously; the firmware converts readings to RMS current with the student's calibration offset.
2. The controller sums both outlets and compares the total against the configured site limit (scaled for the demo rig).
3. When headroom exists, outlets run at their requested current; when the sum would exceed the limit, the allocator throttles per the active policy.
4. Throttling is applied through relay/contactor staging and PWM pilot control on the demo outlets, re-evaluated every control cycle.
5. Plug-in and unplug events are detected from current steps and trigger immediate rebalancing plus a session log entry.
6. All readings, allocations and sessions are published over Wi-Fi via MQTT/HTTP; the dashboard renders the live view and the allocation history chart.

Tech stack:

• ESP32 development board (Wi-Fi + ADC)
• ACS712 current sensors (2x, per outlet)
• Relay/contactor staging + PWM pilot control
• 0.96-inch OLED local display
• Low-voltage demo load rig (scaled outlets)
• MQTT/HTTP telemetry to dashboard
• HTML/CSS/JS power dashboard
• Arduino IDE (C/C++ firmware)

• [Per-outlet current sensing] ACS712 Hall-effect sensors measure each outlet's draw continuously, giving the controller the real load picture it allocates against.
• [Dynamic current allocation] The firmware splits the site current budget between active outlets by policy (equal share or priority), rebalancing within seconds of any plug-in, unplug or limit change.
• [Site-limit enforcement] The combined draw is hard-capped at the configured site limit; the controller throttles outlets via relay/PWM control before the limit is breached, never after.
• [Live power dashboard] Per-outlet power cards, a site-limit gauge, an allocation-over-time chart and a session log with energy totals in one view.
• [Session logging] Every charging session records start/end time, energy delivered and peak current, building the usage history the report analyzes.
• [Priority outlet mode] Outlet A can be configured as priority (e.g. an ambulance or staff vehicle), demonstrating policy-based allocation beyond equal sharing.
• [On-device display] A 0.96-inch OLED shows per-outlet current, total load and the active allocation at the charger itself.

• Working hardware prototype (two-outlet demo rig with sensors, staging relays and controller)
• ESP32 firmware source with sensing, allocation policies and session logging
• Companion dashboard demo (single-file HTML power dashboard)
• ACS712 offset-calibration procedure and record sheet
• Wiring diagram and demo-rig safety notes
• Project report PDF (background, load-balancing principles, allocation algorithms, methodology)
• PPT presentation for final review
• Viva Q&A preparation document (current sensing, control loops, allocation policy, scaling)

• The prototype runs at safe low-voltage demo scale — it demonstrates the balancing algorithm, not a mains-voltage charger installation, which needs certified hardware and a licensed electrician.
• ACS712 readings need per-sensor offset calibration by the student; uncalibrated offsets bias every allocation decision.
• The demo uses resistive demo loads, not real EV onboard chargers, so pilot-signal negotiation with actual vehicles is outside the build.
• Relay staging gives stepped throttling, not smooth continuous control; the report discusses this honestly against contactor-based designs.
• Only two outlets are balanced; the allocation logic extends to N outlets in firmware structure, documented as future scope.
• Site-limit changes mid-session are supported in firmware but demonstrated as a configured scenario, not a utility feed.