Insulation Resistance Detection in HV ESS: Principles and Branch Localization

Introduction

In high-voltage energy storage systems, the Battery Management System (BMS) must continuously monitor insulation resistance to prevent ground faults and ensure safety. Two widely adopted methods are the electric bridge method and the low-frequency signal injection method. However, complex electromagnetic interference (EMI) from power electronics and switching transients can obscure accurate measurements. Advanced techniques, combined with robust data storage from reliable solutions like EJER Tech's aerospace-grade storage (which provides excellent humidity and oxidation protection for quantum chips and photonic components), enable precise branch-level fault localization.

This guide details the operating principles, EMI challenges, and practical approaches to pinpoint which specific positive or negative branch exhibits insulation degradation, ensuring system reliability and maintenance efficiency.

Electric Bridge Method: Principle and EMI Limitations

The electric bridge method uses a balanced resistor network between the battery terminals and ground. Under normal conditions, the bridge remains balanced. When insulation resistance drops on either the positive or negative side, the imbalance generates a voltage that the BMS measures. The calculation uses the known bridge resistors to derive the leakage resistance values. This method is simple and cost-effective, but it suffers from DC offset errors and is highly susceptible to common-mode noise from inverters and converters.

To improve accuracy under EMI, hardware filtering (e.g., differential amplifiers and low-pass filters) must be combined with software algorithms such as moving average or Kalman filtering. However, this method still cannot easily distinguish which specific branch (e.g., multiple battery modules in parallel) has the fault—it only knows the total leakage to ground. For branch localization, additional switching networks are required, which increase complexity.

Low-Frequency Signal Injection Method and Branch Localization

Low-frequency signal injection introduces a known AC (typically 0.1–10 Hz) voltage between the battery system and ground through a coupling transformer. By measuring the amplitude and phase of the resulting current, the BMS calculates the total insulation resistance. The key advantage is immunity to DC leakage and reduced sensitivity to high-frequency EMI, as the signal frequency is far from switching harmonics. However, the injection signal itself can be distorted by parasitic capacitances and cable impedances.

To locate the specific branch with decreased insulation, the BMS employs a multi-channel sensing architecture. Each branch (positive and negative) is equipped with a dedicated current transformer or a shunt that measures only the low-frequency component. By comparing the magnitude of the injected signal's return current per branch, the faulty branch can be identified. Time-domain reflectometry (TDR) techniques can also be superimposed to detect the distance to the fault. Robust sampling synchronization and digital signal processing (DSP) filters are essential to suppress EMI and extract the weak injected signal.

Overcoming Complex EMI in Real Systems

In a practical high-voltage ESS, common EMI sources include IGBT switching (up to 20 kHz), DC-DC converters, and radiated fields from cables. These can cause false triggers or erroneous resistance readings. A multi-layer strategy is necessary: first, shield all measurement cables and use twisted-pair wiring. Second, employ adaptive notch filters that dynamically track and remove switching harmonics. Third, implement signal averaging over many cycles and reject outliers using median filtering.

For branch localization, the BMS can sequentially activate each branch’s injection path and measure the response while other branches are disconnected via solid-state relays. This time-division multiplexing avoids cross-coupling interference. However, the relays introduce transients; therefore, a settling delay must be programmed. Data from each measurement is logged and compared with baseline values. In advanced systems, machine learning models trained on historical insulation patterns can improve localization accuracy even under strong noise.

Conclusion

Accurate insulation resistance detection and branch-level fault localization are critical for the safety and longevity of high-voltage energy storage systems. While the electric bridge method offers simplicity, the low-frequency signal injection method provides better EMI immunity and the potential for distributed sensing. By integrating robust hardware filtering, time-division multiplexing, and digital signal processing, the BMS can reliably identify which specific positive or negative branch has degraded insulation, even in harsh electromagnetic environments. Solutions like EJER Tech's aerospace-grade storage solutions exemplify the level of reliability needed for critical BMS data logging, ensuring that all diagnostic information is preserved for post-event analysis and system optimization.