The Bucket Effect: Why Cell Inconsistency Strands Capacity

In any series-connected high-voltage battery string, usable pack capacity is bounded by the lowest-SOC or most degraded cell — a constraint known as the “bucket effect” or “weakest-cell bottleneck.” No matter how much charge resides in the other cells, the pack terminates discharge the moment the weakest cell reaches its lower voltage limit, leaving that energy permanently inaccessible during that cycle.

Passive balancing, the legacy solution, addresses this by dissipating excess charge from higher-SOC cells through resistors. According to the active balancing battery management system described by Jiangsu Hanhai Core Cloud Network Technology (2016), passive methods operate at maximum currents of approximately 200 mA, making them both slow and thermally wasteful. The energy that could have been redistributed to the weakest cell is instead converted to heat. Active balancing breaks this constraint by transferring charge bidirectionally — from high-SOC donors to low-SOC recipients — with efficiencies routinely exceeding 80%, as documented in the same patent. The practical result is a higher effective floor of available capacity across the entire pack.

The scale of the problem grows with pack size. High-voltage EV packs may contain hundreds to thousands of cells in series. Even small manufacturing variations in cell capacity, combined with differential ageing rates, accumulate into significant SOC spread over time. According to research tracked by IEEE, cell-to-cell capacity variance of just a few percent can translate into meaningful reductions in accessible pack energy over thousands of charge cycles. Active balancing topologies are therefore not a marginal refinement — they are a structural requirement for maintaining pack-level energy availability across the vehicle’s service life.

Circuit Topologies: From Buck-Boost to Transformer-Based Balancing

The most prevalent hardware approach in the patent dataset is the bidirectional DC/DC converter topology, in which each cell in a series string is assigned a corresponding bidirectional DC/DC module and control switch. Hangzhou Huasu Technology’s 2024 patent describes the operating logic precisely: the controller continuously samples real-time cell voltages, identifies the lowest-voltage cell as the first target, selects the higher-voltage adjacent cell as the energy donor, and switches the DC/DC module between Buck and Boost modes to charge or discharge the appropriate cell. This adjacent-cell diffusion approach propagates energy across the pack iteratively, improving voltage uniformity and ensuring the limiting cell receives charge rather than simply being protected from over-discharge.

The key limitation of adjacent-cell Buck-Boost diffusion is that non-adjacent balancing incurs incremental conversion losses at every intermediate stage. The matrix switching approach, described in Daechang Motors’ 2017 patent, addresses this directly: an open-circuit voltage detection unit for each cell is combined with a matrix switching unit that forms any-cell-to-any-cell charge/discharge paths, paired with a bidirectional DC/DC converter that calculates the precise balancing action required. This eliminates multi-hop energy loss for non-adjacent transfers.

“A full switch-matrix topology requires 12 series MOSFETs for a single charge/discharge balancing cycle, suffers from low conversion efficiency, and cannot balance multiple cell groups simultaneously.”

Transformer-based topologies offer a higher-efficiency alternative for non-adjacent cell balancing at reduced switch count. Hangzhou Xenergy Technology’s 2023 patent proposes cascaded basic balancing units, each connecting four series cells via a first-order transformer and a switch group, enabling energy transfer between any two cells with an energy imbalance through the transformer’s transmission path. The patent explicitly contrasts this with the switch-matrix approach, noting the 12-MOSFET requirement and the inability to balance multiple groups simultaneously as its primary drawbacks.

Multi-winding transformer magnetic energy balancing is developed further by Shenzhen Voltek Technology (2024). This approach converts electrical energy from multiple cell groups into magnetic energy stored in the transformer core simultaneously, then releases it preferentially into the lowest-voltage groups via MOS switch control — achieving group-to-group active equalisation without the large bidirectional switch matrices demanded by cell-level switch-matrix designs. The tradeoff identified in the patent record is that high winding counts cause poor coupling and high leakage inductance, increasing switch voltage stress in multi-winding configurations.

Hierarchical Control Architectures and Measured SOC Improvements

High-voltage EV packs containing hundreds to thousands of cells make flat single-level balancing architectures computationally and electrically impractical. The patent record from 2021 onward demonstrates a strong convergence toward hierarchical balancing systems that address cell-level, module-level, and pack-level imbalances in coordinated layers, with each layer operating in parallel rather than sequentially.

The hierarchical architecture described in Chang-in Kim’s 2021 patent structures the pack as M battery modules, each containing N cells, managed by M sub-management modules that measure individual cell voltages and temperatures. These sub-controllers all feed a main management module that collects (M×N) cell voltages and executes both voltage and temperature balance management across the full pack. A companion patent by the same inventor adds a communication module enabling remote state management and real-time balance status reporting. This decomposition reduces the combinatorial complexity of cell selection and allows sub-controllers to operate in parallel, substantially accelerating convergence.

The AutoGiGi Co. (2024) driver-integrated battery balancing device takes simultaneous operation further by linking a cell balance control part directly to a module charging/discharging control part via a bidirectional driver. This allows both intra-module cell imbalances and inter-module imbalances to be addressed concurrently rather than sequentially, improving overall balancing efficiency and reducing the time during which some cells remain at the extremes of the SOC range.

BMW’s 2024 patent describes a particularly elegant approach to inter-module SOC balancing during driving operation that requires no dedicated converter. When a module’s SOC falls significantly below the others, its current path is switched off via the switching system and remains disconnected until the remaining higher-SOC modules discharge down to near the same level, at which point the previously isolated module is reconnected. The equalization runs independently of charging cycles and is reported to be imperceptible to the driver — the vehicle’s traction load acts as the equalisation medium.

Inter-Pack Balancing in Multi-Pack High-Voltage Systems

Modern high-voltage EV architectures increasingly combine multiple battery packs to achieve greater energy capacity or architectural flexibility — for example, 400V/800V switchable systems. This creates a distinct class of balancing challenge at the pack-to-pack level that cell-level and module-level topologies cannot address alone. Cross-pack equalisation topologies have consequently become an active area of patent filing since 2019.

Aiways Automobile’s 2019 patent implements cross-pack equalisation at the individual cell level across a primary-secondary pack configuration. The system calculates residual capacity differences between corresponding cells in the main and auxiliary packs, opens a cross-pack equalisation channel when the difference exceeds a preset threshold, and transfers energy from the auxiliary pack’s cells to the main pack’s matching cells. The benefit for EV range is explicit in the patent: online charging of the main pack extends its discharge time and total deliverable energy.

Jiangsu XCMG Construction Machinery Research Institute’s 2025 patent introduces a three-level architecture for multi-branch parallel systems: battery packs as the first level, Battery Distribution Units (BDUs) as the second level, and a high-voltage junction box as the third level, with a bidirectional DC/DC module bridging the high-voltage battery system and the vehicle’s low-voltage battery. In discharge equalisation mode, high-SOC branches discharge through the DC/DC module into the 12V battery; in charge equalisation mode, the low-voltage battery supplies energy back through the DC/DC to charge low-SOC branches. This dual-purpose design simultaneously achieves inter-branch SOC equalisation and low-voltage battery supplementation.

CATL addresses the most challenging scenario — connecting an additional battery pack to an already-operating circuit — in a 2025 patent that predicts the reflux current that will occur when a slave BMS connects a new pack to the operating circuit, and controls connection timing to a window where the predicted reflux value is within a safe range. This eliminates the need for a dedicated balancing relay and prevents current surges that would otherwise damage cells or trigger protection shutdowns. According to WIPO‘s global patent trend data, predictive protection strategies in battery management systems have been among the fastest-growing sub-categories in EV power electronics filings since 2022.

Tesla’s 2026 pending patent frames the cross-pack connection problem around open-circuit voltage (OCV) matching: the system determines whether the OCV of a main pack and an auxiliary pack match before permitting parallel connection and initiating parallel charging. This OCV-gated approach prevents the inrush currents that would occur if packs at significantly different SOC levels were connected without prior conditioning — a prerequisite for safe range extender integration.

CATL’s 2025 powertrain patent takes a different approach to inter-pack energy transfer by using the vehicle’s motor and motor controller as the energy transfer medium. The controller determines a first equalisation current and equalisation time for Pack 1, routes Pack 1’s energy through the motor controller into the motor as a temporary energy buffer, and subsequently discharges the motor’s stored energy into Pack 2 at the same controlled current. Intra-pack cell balancing via resistive dissipation runs concurrently and independently, forming a complementary multi-layer system that eliminates the need for a dedicated inter-pack balancing converter.

Algorithmic Path Optimisation: Depth-First Search and Predictive Control

Beyond hardware topology, significant patent activity addresses the algorithmic problem of selecting the most efficient balancing sequence, particularly when energy must traverse multiple intermediate cells to reach a distant target. The choice of balancing path directly determines how much of the transferred energy actually reaches the target cell versus being lost as heat in intermediate conversion stages.

Hefei University of Technology’s 2017 patent formalises the balancing path selection problem as a directed weighted graph. Cell terminal voltages are nodes, and edge weights represent both the energy consumed and the energy transferred in one balancing cycle between any pair of cells, calculated from measured internal resistance and DC impedance of wiring and switching components. A depth-first search with backtracking exhaustively enumerates paths to find the one with highest energy-transfer efficiency. This ensures that, for non-adjacent cells, the system selects multi-hop paths that minimise cumulative losses rather than following the nearest-neighbour default, which would incur unnecessary intermediate conversion steps.

Chongqing University’s 2023 patent addresses the practical reality that equaliser efficiency varies nonlinearly with equalisation current. The patent constructs a state-space model for a bus-type equalisation system that integrates this nonlinearity, then applies Model Predictive Control (MPC) to minimise equalisation losses while also accounting for time efficiency — treating faster convergence as a weighted optimisation objective alongside energy conservation. This dual-objective formulation reflects the operational reality of EV use: prolonged balancing during a charging session that blocks the start of a drive is itself a capacity-availability penalty. Standards bodies including IEC have increasingly emphasised the importance of time-bounded battery management operations in EV safety and performance standards.

Zhengzhou University’s 2025 patent identifies a critical failure mode in single-topology active equalisers: when only over-charged cells or only under-charged cells exist in the pack, Any-Cell-to-Any-Cell (AC2AC) mode equalisers cannot form valid energy transfer paths, causing the balancing function to fail entirely. The proposed multi-mode topology supports Cell-to-String (C2S), String-to-Cell (S2C), and direct Cell-to-Cell modes via a configurable bidirectional switch array, ensuring that complex over-charge/under-charge combinations are handled without routing energy through intermediate strings at reduced efficiency.

Shenzhen Leran Technology’s 2025 patent extends algorithmic optimisation to the cross-pack level with a scheduling framework that first maps cell configurations across all packs, constructs cross-pack equalisation path sets ranked by “path benefit,” selects paths that satisfy all triggering conditions simultaneously, and manages conflicts between intra-pack and cross-pack equalisation operations through a queuing system. The architecture ensures that intra-pack equalisation takes priority, that cross-pack channels targeting cells currently undergoing intra-pack balancing are deferred to a waiting queue, and that once intra-pack work is complete, cross-pack paths are re-evaluated — preventing destructive interactions between the two balancing layers.

“Prolonged balancing during a charging session that blocks the start of a drive is itself a capacity-availability penalty — making faster convergence a weighted optimisation objective alongside energy conservation.”

Patent Landscape: Key Players and Innovation Trends

The patent dataset of approximately 55 sources filed between 1999 and 2026 reveals a clear geographic and corporate concentration in active cell balancing innovation, with a distinct evolution from single-level, single-pack cell balancing (pre-2020) toward hierarchical multi-level systems that simultaneously address cell-level, module-level, and pack-level imbalances (2021–2026).

LG Chem / LG Energy Solution (Korea) is the most prolific assignee in the dataset, with patents covering voltage balancing apparatuses for battery modules (2018, 2020), inter-rack voltage balancing (2017), relay-controlled rack connection with reference voltage simulation (2016, 2017), and master-slave module equalisers (2019). Their approach consistently uses equalisation current supply circuits to correct SOC abnormalities in individual modules within larger packs, targeting pack-level usable capacity recovery.

CATL / Ningde Times (China) appears with multiple active patents covering discharge/charge control systems for parallel battery packs (2025) and motor-mediated inter-pack balancing (2025, 2026), reflecting their scale and systems-integration focus. Jiangsu XCMG Construction Machinery Research Institute (China) is a notable industrial equipment filer, with both a multi-branch parallel active balancing architecture patent (2025) and a multi-pack passive balancing system patent (2025), indicating active exploration of both paradigms.

Hyundai Motor Company (Korea) addresses the pre-charge balancing problem for parallel pack connection (2024, pending) and relay-switched series/parallel reconfiguration for module-level balancing (2025), reflecting OEM-level integration of balancing into powertrain design. BMW differentiates with a driving-operation-integrated balancing approach that avoids dedicated converters entirely (2024), while Tesla focuses on OCV-matched pre-conditioning for safe auxiliary pack parallel connection (2026).

Trend analysis across the dataset indicates that the period 2021–2026 is characterised by three converging developments: hierarchical multi-level control architectures that decompose the balancing problem across parallel sub-controllers; cross-pack equalisation channels that recover capacity stranded in auxiliary or lower-SOC packs; and algorithm-driven path selection using techniques such as depth-first search and Model Predictive Control to minimise conversion losses and convergence time. These developments are consistent with the broader direction identified by OECD in its energy technology innovation tracking, which notes accelerating IP activity in battery management systems as EV adoption scales globally. For engineers and IP professionals tracking this space, PatSnap’s R&D intelligence platform provides real-time access to the full patent record across all relevant jurisdictions.