The Four Root Causes of Capacity Imbalance in Large-Format Battery Modules
Capacity imbalance in large-format battery modules reduces usable pack capacity to that of the weakest cell or bank in the series chain — a consequence of four compounding mechanisms: manufacturing variation, differential self-discharge, contact resistance asymmetry, and thermal gradients. Understanding each mechanism is a prerequisite for designing BMS algorithms that can distinguish genuine cell degradation from measurement artefacts.
Manufacturing Variation and Cell-Level Parameter Spread
Even within tightly controlled production batches, individual cells exhibit differences in internal resistance, electrode coating uniformity, and electrolyte distribution. As demonstrated by SK On’s 2021 BMS patent, SOC divergence within a pack emerges progressively as cells with varying state-of-health (SOH) respond differently to identical charge/discharge currents. The system selects “high-risk group battery cells” by applying SOH-dependent corrections to each cell’s SOC, explicitly acknowledging that a single representative SOH value inadequately captures inter-cell heterogeneity. Robert Bosch GmbH formalises this by computing a quality factor for each cell that combines a function of state of charge with a function of the product of current and internal resistance — explicitly coupling electrical state with electrochemical aging.
Manufacturing variation in internal resistance, electrode coating uniformity, and electrolyte distribution causes progressive SOC divergence across cells in large-format battery modules, even when all cells are subjected to identical charge and discharge currents.
Differential Self-Discharge
Cells with manufacturing defects, micro-short circuits, or contamination discharge faster during rest periods than neighbouring cells, producing SOC divergence without any applied load. BMW’s 2021 patent on abnormal self-discharge detection addresses this directly by monitoring balancing charge per cell over time, enabling prediction of safety-critical internal short circuit conditions well before thermal or voltage anomalies become detectable. LG Energy Solution’s 2025 detection of lithium precipitation calculates cumulative deviations between measured terminal voltage and open-circuit voltage (OCV) to flag cells undergoing abnormal electrochemical processes such as lithium plating — a condition that accelerates both self-discharge and capacity fade.
Contact Resistance Asymmetry and Interconnect Degradation
In large-format packs consisting of parallel battery banks, contact resistances at bus bars and connectors vary across banks, causing unequal current distribution. LG Energy Solution’s 2023 battery management patent specifically decomposes measured bank voltages into internal resistance components and external contact resistance components, then calculates compensated bank voltages by removing the voltage drop due to each contact resistance. This compensation is necessary because raw voltage measurements taken at module terminals misrepresent actual cell SOC when resistance is non-uniform. The same methodology appears across multiple LG Energy Solution patent families through 2026, confirming this as a persistent, systematically addressed engineering challenge.
Raw voltage measurements at module terminals include ohmic drops from bus bar and connector resistances that vary across banks. BMS algorithms must decompose these measurements into internal and external resistance components before any balancing decision is executed — otherwise the system may attempt to balance cells that are actually at equal SOC.
Thermal Gradients Within the Module
In large-format modules, cells positioned near the centre experience higher operating temperatures than edge cells, accelerating electrochemical aging and increasing local self-discharge. The hierarchical management patent from inventor Kim Chang-in (2021) deploys temperature sensors at designated locations throughout each module and aggregates temperature values across all sub-management modules to the main management module, managing both voltage balance states and temperature balance states simultaneously. SoluM’s 2024 battery pack device extends this architecture to commercial pack designs, reflecting the industry understanding that temperature imbalance and capacity imbalance are causally linked.
BMS Algorithms for Imbalance Detection and Correction: From SOC Estimation to Statistical Anomaly Detection
BMS algorithms addressing capacity imbalance operate across three functional tiers: measurement and state estimation, fault detection and diagnosis, and active or passive equalization control. Without accurate per-cell SOC, balancing decisions are uninformed and may worsen imbalance by over-discharging weak cells — making the estimation layer the non-negotiable foundation of every other corrective action.
SOC and SOH Estimation as the Foundation of Imbalance Detection
LG Energy Solution’s 2016 BMS patent (originally filed under LG Chem) implements a SOC/SOH estimation loop that triggers a balancing signal when maximum SOC deviation across modules reaches a first limit deviation, and escalates to a diagnostic signal at a second, higher limit deviation. Critically, when modules are replaced with new units having different SOH than the original pack, the algorithm applies a corrected third limit deviation proportional to the SOH deviation between old and new modules — preventing false alarms in heterogeneous-age packs. Samsung SDI’s 2024 approach uses an adaptive filter to update G and H parameters in real time from voltage and current measurements, where G represents voltage sensitivity to current change and H captures the effective potential from local equilibrium potential distribution and resistance distribution, enabling model-based SOH estimation without full-cycle testing.
“When modules are replaced with new units having different SOH than the original pack, the algorithm applies a corrected third limit deviation proportional to the SOH deviation — preventing false alarms in heterogeneous-age packs.”
Voltage History Analysis and Statistical Anomaly Detection
LG Energy Solution’s 2024 BMS patent constructs an observation matrix of voltage vectors measured in a sliding time window for each cell, then generates a restoration matrix of predicted voltage vectors. Differences between observed and restored vectors detect abnormality — cells with internal short circuits or abnormal self-discharge produce reconstruction errors that differ statistically from the healthy population. This moving-window approach, endorsed by standards bodies such as IEEE for fault-tolerant system design, represents a significant advance over fixed-threshold triggers. SK On’s 2021 BMS patent uses maximum voltage change rate across cells to select high-risk candidates and calculate a maximum current limit value that keeps all cells within their operating voltage window — a safety constraint derived directly from inter-cell imbalance severity.
LG Energy Solution’s statistical delta-SOC method for internal short circuit detection computes the SOC change for each cell during a charging sequence, derives an ensemble reference factor from all cells in the pack, and flags individual cells whose delta-SOC deviates from this reference — making detection relative rather than absolute and robust against pack-wide aging.
Internal Short Circuit Detection via SOC Change Tracking
LG Energy Solution’s 2022 BMS patent applies a SOC estimation algorithm during charging, computes the first SOC change (delta-SOC between two charge time points) for each cell, then applies a statistical algorithm to the ensemble to derive a reference factor. Cells whose delta-SOC deviates from the reference factor are flagged for internal short circuit. This statistical reference approach is robust against pack-wide aging because it normalises each cell’s delta-SOC against its peers, making detection relative rather than absolute. The same methodology is patented across LG Energy Solution’s Japanese family (2023), confirming international deployment intent across EV and ESS markets.
Explore the full patent landscape for BMS algorithms and cell balancing circuits in PatSnap Eureka.
Explore Full Patent Data in PatSnap Eureka →Resistance Degradation Diagnosis and Capacity Correction
LG Energy Solution’s 2023 resistance degradation patent implements a pulse discharge protocol on target battery banks, recording bank voltage before and after discharge, deriving the current voltage change amount, and comparing it against a stored reference voltage change amount to estimate resistance degradation degree. This approach enables non-intrusive, in-service resistance mapping without requiring electrochemical impedance spectroscopy (EIS) hardware — a significant operational advantage for deployed ESS systems. As noted in technical guidance from IEC, avoiding invasive diagnostic procedures during normal operation is a key design criterion for grid-scale storage systems. LG Energy Solution’s 2025 apparatus extends this by estimating internal resistance of parallel-connected modules from routine voltage and current measurements, setting a reference resistance from the ensemble, and flagging modules deviating beyond a defined threshold.
Sungrow Power Supply’s 2025 capacity correction patent applies dynamic voltage range gating to establish valid start and end voltage indicators for full discharge and charge sequences, obtaining initial and final charging capacities from these indicators and computing total corrected capacity. This ensures that partial cycles — common in field operation — do not produce cumulative SOC drift errors that masquerade as capacity imbalance.
Balancing Circuit Topologies: From Resistive Dissipation to Hierarchical Active Equalization
BMS algorithms require corresponding hardware to execute calculated balancing commands. The patent dataset reveals a spectrum of passive, active, and hybrid circuit approaches, each with distinct trade-offs in energy efficiency, cost, and scalability — a taxonomy that aligns with classifications maintained by bodies such as IEC and documented in grid storage standards from WIPO-registered international filings.
Passive Balancing via Resistive Dissipation
Passive balancing remains the baseline approach for cost-sensitive applications. The Korea Electrotechnology Research Institute’s 2014 patent establishes the canonical passive algorithm: measure OCV for each cell, designate a balancing voltage from the measured set, and discharge cells above this voltage through switch-controlled resistors until all OCVs converge. While thermally inefficient, this approach is reliable and simple. LG Chem’s 2019 patent adds thermal management integration, routing current from the balancing resistor through a fan cooling circuit and using second switching units to manage the thermal load generated during passive balancing — addressing a practical limitation of resistive dissipation in high-energy modules.
Active Balancing with Equalization Current Injection
Active balancing transfers charge rather than dissipating it, achieving higher energy efficiency. LG Chem’s 2020 apparatus implements a Cell Module Controller (CMC) for each battery module and a central Battery Module Controller (BMC) that calculates SOC for each module, diagnoses abnormal modules, and forms a closed equalization current supply circuit connecting the BMC to the target module via switching units. The position of the abnormal module within the string determines which switching units are activated to route equalization current optimally. LG Energy Solution’s 2024 apparatus introduces a dedicated balancing line, electrically independent from the main power line, that connects positive and negative terminals of multiple cell assemblies in parallel — enabling continuous balancing during charge and discharge operations without interrupting the main current path.
LG Energy Solution’s 2024 active balancing architecture introduces a dedicated balancing line electrically independent from the main power line, connecting positive and negative terminals of multiple cell assemblies in parallel to enable continuous SOC equalization during both charge and discharge operations without interrupting the main current path.
Constant Current Balancing with DC/DC Conversion
Crest Co., Ltd.’s 2022 patent uses a single charging control unit and a single discharging control unit sharing DC/DC converters for each direction, with a switching unit reconfiguring the current path to connect the appropriate converter to the target cell. This topology reduces converter count relative to per-cell active balancing architectures, lowering cost while preserving bidirectional energy transfer capability. Linear Technology LLC’s 2016 voltage-compensated active cell balancing patent addresses a fundamental measurement accuracy problem: cell voltage measured during a balancing current flow is distorted by the ohmic drop across series resistance. The apparatus measures both cell voltage and the voltage drop due to balancing current, then calculates a corrected open-cell voltage by subtracting the ohmic contribution — significantly improving balancing accuracy at high currents.
Linear Technology LLC’s voltage-compensated active cell balancing apparatus measures both cell voltage and the ohmic voltage drop due to balancing current, then bases all balancing decisions on a corrected open-cell voltage rather than the raw measured value. Samsung SDI’s 2026 BMS applies the same connection-resistance correction principle to improve both measurement precision and balancing efficiency.
Hierarchical BMS Architecture for Large ESS Packs
LG Energy Solution’s 2024 hierarchical BMS patent employs a Rack BMS (RBMS) as the first-level management device that determines whether balancing execution conditions are met based on idle state and charging status, sets balancing targets and timing, and commands Pack BMS (PBMS) units residing within each Battery Module Assembly (BMA). The PBMS executes cell-level balancing according to RBMS instructions. This hierarchical control cleanly separates system-level strategy from module-level execution and scales to large ESS rack configurations. The same hierarchical design principle is embodied in Kim Chang-in’s 2021 patents, which implement distributed sub-management modules reporting to a central main management module.
Bypass and Fault Isolation
Thunder Power New Energy’s 2018 patent implements a connecting circuit that normally places each module in series but, upon detection of abnormal operation, disconnects the defective module and directly bridges the preceding and succeeding modules, maintaining series circuit continuity. This hardware bypass function complements BMS diagnostic algorithms by providing a graceful degradation path when balancing alone cannot restore pack operability — a design principle increasingly relevant as pack sizes grow for grid-scale ESS applications monitored by organisations such as IEA.
Search and analyse BMS balancing circuit patents across all major assignees using PatSnap Eureka’s AI-powered patent intelligence platform.
Analyse Patents with PatSnap Eureka →Key Players and Innovation Trends Across 60+ Patent Filings
LG Energy Solution and LG Chem dominate the dataset with more than 25 patent filings spanning SOC/SOH estimation, active balancing circuits, internal resistance diagnosis, short circuit detection, contact resistance compensation, and hierarchical BMS architectures — a breadth that reflects their global deployment in both EV and ESS markets across Korea, Japan, and international jurisdictions.
Assignee Profiles and Technical Specialisations
Samsung SDI contributes to both SOH estimation methodology via adaptive G/H parameter filters and cell balancing hardware, with recent 2026 filings addressing connection-resistance correction and RC-circuit fault compensation in BMS balancing switches. SK On focuses specifically on high-risk cell identification and current limit management, producing BMS algorithms that select cells most likely to exit operating voltage bounds and using these as the basis for conservative current limiting rather than aggressive balancing. Robert Bosch GmbH approaches the problem from a probabilistic switching standpoint, computing quality factors that incorporate both SOC and the product of current and internal resistance to determine connection probabilities for individual cells, enabling dynamic reconfiguration of the series string to compensate for inter-cell variation.
Korea Electrotechnology Research Institute and Linear Technology LLC (now part of Analog Devices) have contributed foundational balancing methodologies — OCV-based passive equalization and voltage-drop-compensated active balancing, respectively — that underpin subsequent industry approaches. Sungrow Power Supply addresses the ESS-specific challenge of capacity correction across multiple modules operating under partial-cycle conditions. BMW introduces a consumer electronics and automotive safety perspective through abnormal self-discharge monitoring, linking per-cell balancing charge monitoring to early prediction of internal short circuit risk.
The Shift from Threshold-Based to Statistical and Model-Based Detection
An important trend visible across filings from 2022 onward is the shift from threshold-based balancing triggers to statistical and model-based anomaly detection. Moving windows, deviation rankings, cumulative deviation tracking, and machine learning-assisted state prediction all appear as next-generation diagnostic tools. LG Energy Solution’s 2025 patent applies machine learning to battery status data to generate predictive state assessments — a development consistent with broader trends in AI-driven predictive maintenance documented by organisations such as IEEE and WIPO. This trajectory suggests that future BMS architectures will increasingly rely on data-driven models trained on fleet-scale telemetry rather than fixed electrochemical thresholds derived from laboratory characterisation.
From 2022 onward, battery management system patents increasingly shift from fixed-threshold balancing triggers to statistical and model-based anomaly detection methods, including moving-window voltage history analysis, cumulative deviation tracking, and machine learning-assisted state prediction, as evidenced by LG Energy Solution’s 2024 and 2025 patent filings.
“The shift from threshold-based balancing triggers to statistical and model-based anomaly detection — moving windows, deviation rankings, cumulative deviation tracking — is the defining innovation trend in BMS architecture from 2022 onward.”
For battery engineers and BMS architects, the practical implication is that competitive differentiation in large-format pack design is migrating from hardware topology — which is increasingly commoditised — to the quality of state estimation and anomaly detection algorithms. Tracking this patent landscape in real time is essential for identifying both freedom-to-operate risks and licensing opportunities in a domain where LG Energy Solution alone holds more than 25 active filings.
PatSnap’s innovation intelligence platform indexes over 2 billion data points across 120+ countries, enabling R&D teams to monitor this evolving BMS patent landscape continuously. The PatSnap IP management suite and the PatSnap R&D intelligence tools provide structured access to the full filing histories, citation networks, and legal status data underlying the analysis in this article.