🎓 Lesson 10 D5

Coulomb Counting Limitations & Drift Compensation

Coulomb counting is a method to estimate how much charge has gone in or out of a battery, but it drifts over time because small measurement errors add up — like walking with slightly uneven steps and gradually veering off course.

🎯 Learning Objectives

  • Calculate cumulative Coulomb counting error over 72 hours given specified current sensor offset and sampling rate
  • Design a drift-compensation schedule using voltage-triggered SoC recalibration thresholds per IEEE 1625 guidelines
  • Analyze SoC estimation divergence between Coulomb counting and Kalman-filtered methods under partial-state cycling
  • Explain the physical origins of current-sense amplifier drift and their impact on energy accounting in mining UPS systems

📖 Why This Matters

In remote mining operations, battery energy storage systems (BESS) power critical ventilation, communications, and emergency lighting — often for days without grid support. An undetected 5% SoC drift can cause unexpected blackouts during shift changes or blasting sequences. Unlike consumer electronics, mining BESS cannot rely on periodic full charging for recalibration; duty cycles are irregular and deep discharges are avoided for longevity. Understanding and mitigating Coulomb counting drift isn’t theoretical — it’s a safety-critical design requirement.

📘 Core Principles

Coulomb counting rests on charge conservation: ΔQ = ∫ I(t) dt. In practice, measured current Î(t) = I(t) + ε_offset + ε_gain·I(t) + ε_noise, where ε terms represent sensor imperfections. Offset error (e.g., ±2 mA) integrates linearly with time — causing SoC drift proportional to duration, not load. Gain error scales with current magnitude but introduces nonlinearity. Temperature shifts further modulate both offset and gain (e.g., ±0.5 µV/°C input offset drift in shunt amplifiers). Drift compensation strategies fall into two categories: (1) periodic absolute reference (e.g., voltage-based SoC anchors at known OCV points), and (2) model-based correction (e.g., augmenting with thermal-aware impedance models). Mining BESS uniquely require hybrid approaches due to wide ambient temperature swings (-30°C to +50°C) and infrequent full-charge opportunities.

📐 Cumulative Drift Estimation

The dominant source of long-term drift is current-sense offset error. This formula quantifies worst-case SoC error accumulation over time, assuming constant offset and no correction.

Coulomb Counting Drift Error

ΔSoC = (ε_offset × t) / C_nom × 100

Estimates percentage state-of-charge error due to constant current-sense offset over time

Variables:
SymbolNameUnitDescription
ΔSoC SoC drift error % Absolute percentage deviation from true state of charge
ε_offset Current sensor offset error A DC bias error in measured current (typ. ±1–5 mA for industrial shunt amplifiers)
t Integration time h Duration over which offset accumulates
C_nom Nominal battery capacity Ah Rated capacity at specified discharge rate and temperature
Typical Ranges:
Industrial shunt-based current sensing (±3σ): ±1.0 – ±5.0 mA
Mining BESS duty cycle (uninterrupted operation): 24 – 168 h

💡 Worked Example

Problem: A mining site UPS uses a 100 Ah LiFePO₄ battery with a current sensor having a guaranteed offset of ±3 mA. The system logs current every 10 seconds and operates continuously for 48 hours with average load current of 2 A. Estimate maximum SoC drift.
1. Step 1: Convert offset to ampere-hours: |ε_offset| = 0.003 A × 48 h = 0.144 Ah
2. Step 2: Normalize to battery capacity: Drift (%) = (0.144 Ah / 100 Ah) × 100 = 0.144%
3. Step 3: Account for bidirectional current: In real operation, charge/discharge asymmetry may double effective drift if offset polarity aligns with net current direction — worst case ≈ 0.29% SoC error after 48 h
Answer: The maximum SoC drift is 0.29%, which exceeds the 0.2% accuracy target required by IEC 62933-2-2 for critical backup power systems.

🏗️ Real-World Application

At Newmont’s Boddington Gold Mine (Western Australia), a 2.4 MWh Li-ion BESS powers underground ventilation during grid outages. Field data revealed 1.8% SoC drift over 120 hours of partial cycling — traced to uncorrected shunt amplifier offset drift at 42°C ambient. Engineers implemented voltage-triggered recalibration only at SoC >90% and <10%, combined with temperature-compensated offset nulling every 8 hours. This reduced median SoC error from ±1.6% to ±0.13% over six months — enabling reliable 72-hour autonomy claims verified by third-party audit (DNV GL Report No. AU-MIN-2023-088).

📚 References