As a researcher focused on electrochemical energy storage systems, I am constantly exploring methods to enhance the safety and longevity of lithium-ion batteries, especially within the context of large-scale energy storage battery applications. The push for rapid charging, often driven by the need for fast energy absorption and dispatch in grid applications, introduces significant safety risks, primarily linked to lithium plating on the graphite anode. This side reaction not only accelerates capacity fade but can also lead to internal short circuits and thermal runaway. Therefore, developing online control strategies to enable safe, fast charging for energy storage batteries is of paramount importance. This article details an approach based on a reduced-order electrochemical model to prevent lithium plating during the charging of energy storage batteries.
The core challenge lies in the fact that lithium plating is directly governed by the anode’s electrochemical potential versus a Li/Li+ reference. When the local anode potential falls below 0 V, the thermodynamic driving force favors the reduction of lithium ions to metallic lithium, leading to plating. In a commercial energy storage battery, directly measuring this potential is impractical. Therefore, an accurate, computationally efficient model capable of estimating the internal anode potential in real-time is essential for any online control system aimed at safe operation. This is where physics-based electrochemical models become invaluable, though their complexity must be addressed for practical battery management system (BMS) implementation.
From the Comprehensive P2D Model to a Simplified Framework
The pseudo-two-dimensional (P2D) model, pioneered by Newman and collaborators, provides a rigorous microscopic description of the processes within an energy storage battery. It couples five governing equations describing lithium-ion concentration in the solid and liquid phases, potentials in both phases, and the kinetics of the charge-transfer reaction at the particle surfaces. While highly accurate, solving these coupled partial differential equations (PDEs) is computationally intensive, making direct onboard use in an energy storage battery BMS infeasible. My work focuses on a systematic dimensional reduction and simplification of this model to create a Simplified Pseudo Two-Dimensional (SP2D) model suitable for online control.
The first major simplification addresses the solid-phase diffusion dynamics. Instead of solving Fick’s law in spherical coordinates, the relationship between the volume-averaged lithium concentration and the surface concentration is approximated using two first-order inertial elements. This transforms a PDE into ordinary differential equations (ODEs) that are much simpler to compute. The governing equations are:
$$
\bar{c}_s(t_{k+1}) = \bar{c}_s(t_k) – \frac{3 j_f}{a_s F r_s} \Delta t
$$
$$
c_{s,surf}(t_{k+1}) = \bar{c}_s(t_{k+1}) – \lambda_1 w_1(t_{k+1}) – \lambda_2 w_2(t_{k+1})
$$
where \( \bar{c}_s \) is the average solid-phase concentration, \( c_{s,surf} \) is the surface concentration, \( j_f \) is the local pore wall flux, \( a_s \) is the specific surface area, \( r_s \) is the particle radius, \( F \) is Faraday’s constant, and \( \lambda_1, \lambda_2, w_1, w_2 \) are parameters and states of the inertial blocks.
For the electrolyte phase, the concentration distribution is approximated using low-order polynomial functions across the anode, separator, and cathode domains. This avoids solving the liquid-phase mass conservation PDE directly. The concentration \( c_e(x,t) \) is given by:
$$
c_e(x,t) =
\begin{cases}
a_1(t)x^2 + a_2(t) & \text{(Anode)} \\
a_3(t)x^2 + a_4(t)x + a_5(t) & \text{(Separator)} \\
a_6(t)(L-x)^2 + a_7(t) & \text{(Cathode)}
\end{cases}
$$
where \( L \) is the total thickness from the negative to the positive current collector, \( x \) is the spatial coordinate, and \( a_1 \) to \( a_7 \) are time-dependent coefficients determined from boundary and interface conditions.
The potential distributions are also simplified. The solid-phase potential gradient is approximated using the average current, and the liquid-phase potential is simplified by neglecting the nonlinear effects in highly concentrated solutions under certain assumptions, leading to manageable expressions for the overpotential. Ultimately, the local pore wall flux \( j_f \) can be derived and expressed as a function of known or estimated variables, including the open-circuit potential \( U_{ref} \), which itself is approximated by a cubic polynomial in space for computational ease:
$$
U_{ref}(x,t) = b_3 x^3 + b_2 x^2 + b_1 x + b_0
$$
Through these steps, the complex, coupled P2D model is reduced to a set of algebraic equations and ODEs—the SP2D model—that retains the critical physics linking external current to internal anode potential while being orders of magnitude faster to solve. This makes it a suitable candidate for embedding in the management system of an energy storage battery.
Parameterization of the Model for an Energy Storage Battery
A model is only as good as its parameters. Accurately determining the parameters for the SP2D model requires a multi-faceted approach tailored for a specific energy storage battery cell. The parameters can be categorized based on their source:
- Geometric & Measured: Parameters like electrode dimensions (length, width, thickness), electrode coating area, and separator thickness are obtained through direct physical measurement of a disassembled cell.
- Manufacturer-Provided: Key material properties such as the solid-phase diffusion coefficients (\(D_s\)), initial electrolyte concentration (\(c_{e,0}\)), and particle radii (\(r_s\)) are often supplied by the cell manufacturer based on their material specifications.
- Electrochemical Testing: The equilibrium potential curves for the positive and negative electrodes are fundamental. These are obtained by constructing half-cells (coin cells) with working electrodes harvested from fresh energy storage battery electrodes versus lithium metal counter electrodes. The cells are then cycled at very low currents (C/20 or slower) to approximate equilibrium conditions.
The resulting open-circuit potential (OCP) curves versus lithium stoichiometry are crucial inputs. For a full-cell model, the initial stoichiometries and electrode capacities must be calibrated. This is often done using a “dual-tank” or lumped model fitting against low-rate full-cell data. The model equations are:
$$
y(t) = y_0 – \frac{I_{in} t}{C_p}
$$
$$
x(t) = x_0 + \frac{I_{in} t}{C_n}
$$
$$
V_{out}(t) = U_p(y(t)) – U_n(x(t)) + I_{in} R_{in}
$$
where \(x\) and \(y\) are the anode and cathode stoichiometries, \(C_n\) and \(C_p\) are the electrode capacities, \(U_n\) and \(U_p\) are the OCP functions, and \(R_{in}\) is an internal resistance. By using a genetic algorithm or other optimization techniques to minimize the error between \(V_{out}(t)\) and experimental low-rate voltage data, the parameters \(x_0, y_0, C_n, C_p, R_{in}\) are identified. From \(C_n\) and \(C_p\), the maximum lithium concentrations in the electrodes (\(c_{s,max}\)) can be calculated, providing vital scaling for the SP2D model.
Other parameters, like the electrochemical reaction rate constants (\(k\)), are tuned within physically reasonable bounds to achieve the best fit between the SP2D model’s voltage/overpotential predictions and experimental data from cells instrumented with reference electrodes. A representative set of parameters for an NCM-based energy storage battery is summarized below.
| Parameter | Description | Value | Source |
|---|---|---|---|
| \(A_{cell}\) | Electrode area | 5.05 × 10-2 m² | Measurement |
| \(L_n, L_s, L_p\) | Anode, separator, cathode thickness | 86, 20, 68 μm | Measurement |
| \(r_{s,n}, r_{s,p}\) | Anode & cathode particle radius | 1.5 × 10-5 m | Vendor |
| \(\epsilon_{s,n}, \epsilon_{s,p}\) | Solid phase volume fraction | 0.4 | Vendor |
| \(c_{s,max,n}\) | Max Li conc. in anode (graphite) | ~36,400 mol/m³ | Calibration |
| \(c_{s,max,p}\) | Max Li conc. in cathode (NCM) | ~98,300 mol/m³ | Calibration |
| \(D_{s,n}, D_{s,p}\) | Solid-phase diffusion coefficient | ~1.2×10-13, ~4.5×10-13 m²/s | Vendor/Literature |
| \(k_n, k_p\) | Reaction rate constant | ~1.8×10-10, ~2.3×10-11 m/s | Electrochemical Fitting |
Model Validation: A Critical Step for Energy Storage Battery Management
Before the SP2D model can be trusted for online control in an energy storage battery system, it must be rigorously validated against experimental data. The most critical validations involve comparing model predictions of terminal voltage and, more importantly, anode potential against measured values. Anode potential is measured using a stable reference electrode (e.g., Li metal or Li alloy) embedded in the cell. Experiments involve charging the cell at various constant current (CC) rates (e.g., 0.33C, 0.5C, 1C, 1.5C) from a low state-of-charge (SOC) to the upper voltage cutoff.
The Root Mean Square Error (RMSE) between the model output and experimental data serves as the key metric. Successful validation for an energy storage battery SP2D model typically shows RMSE values for terminal voltage below 15 mV and for anode potential below 10 mV across a range of currents. This level of accuracy confirms that the simplified model reliably captures the internal electrochemical state, particularly the anode potential that governs lithium plating. The table below exemplifies typical validation results, demonstrating the model’s fitness for its purpose in managing the charging of an energy storage battery.
| Charging Rate (C-rate) | Terminal Voltage RMSE (mV) | Anode Potential RMSE (mV) |
|---|---|---|
| 0.33C | 8.4 | 6.5 |
| 0.5C | 8.1 | 5.6 |
| 1.0C | 9.5 | 7.0 |
| 1.5C | 11.6 | 4.7 |
Online Control Strategy for Safe Charging of Energy Storage Batteries
With a validated SP2D model capable of real-time anode potential estimation, an online feedback control loop can be constructed. The control objective is to adjust the charging current \(I_{charge}(t)\) such that the estimated anode potential \( \eta_{anode}(t) \) is maintained at or above a safe threshold \( \eta_{lim} \), typically chosen as 0 V vs. Li/Li+ plus a small safety margin (e.g., 10-30 mV).
A simple yet effective Proportional (P) controller is often sufficient for this task. The control law is:
$$
I_{k} = I_{k-1} + K_p \cdot (\eta_{lim} – \eta_{anode,k})
$$
where \(I_k\) is the current command for the next control interval, \(K_p\) is the proportional gain, and \(\eta_{anode,k}\) is the anode potential estimated by the SP2D model for the current interval. A saturation block is added to limit the maximum current to the capabilities of the charger and the energy storage battery itself. The block diagram for this controller is shown below, integrated with the SP2D model running in the BMS.
[Control Block Diagram Logic: External Current Command -> Saturation Block -> Energy Storage Battery & SP2D Model -> Estimated Anode Potential -> Comparison with Safety Threshold -> Error -> P-Controller -> Feedback to adjust Current Command.]
In a simulation, this controller starts with an initial current. As charging proceeds and the anode potential begins to dip toward the threshold, the controller reduces the current to keep the potential safe. Conversely, if the potential is safely high, the controller can increase the current. This results in a charging profile that is not constant current, but dynamically optimized to charge as fast as possible without crossing the lithium plating boundary. A typical simulation outcome shows an initial high-current phase, followed by a gradually decreasing current as the cell approaches full charge, all while the anode potential hovers just above the safety limit.
Experimental Verification of the Safe Charging Method
Simulation is promising, but experimental verification on actual energy storage battery cells is essential. The optimal current profile derived from simulation and control logic must be translated into a sequence that a battery test station can execute, often as a multi-step constant current profile. This profile is then applied to fresh energy storage battery cells for cycle life testing.
The primary verification metrics are:
- Capacity Retention: Cells charged using the model-based safe profile should exhibit superior capacity retention over many cycles compared to cells charged with a standard, non-optimized fast-charge protocol (e.g., a simple 1C CC-CV). This indicates reduced degradation, consistent with the suppression of lithium plating.
- Post-Mortem Analysis: After several hundred cycles, cells charged with the safe protocol can be disassembled. The surface of the graphite anode should appear uniform and dark, without the characteristic grayish or silvery deposits of plated lithium metal that are often visible after abusive fast charging.
- Non-Destructive Plating Detection: Advanced diagnostic techniques, such as analyzing the relaxation voltage curve after charging, provide further evidence. The derivative of the relaxation voltage (\(dV/dt\)) for a cell that has undergone lithium plating shows a distinct minimum corresponding to the re-intercalation of reversible “dead” lithium. The absence of this feature in cells charged with the safe protocol supports the claim of plating mitigation.
Results from such tests consistently show that the model-based charging strategy enables significantly faster charging than a conservative low-rate method, while achieving cycle life comparable to or better than standard fast-charge methods. This directly validates the SP2D model’s utility in extending the safe operational envelope of an energy storage battery.
Conclusion and Future Perspectives
The integration of reduced-order electrochemical models like the SP2D model into the battery management system represents a significant leap forward for the safe and efficient operation of energy storage batteries. By enabling real-time estimation of the critical internal state—the anode potential—these models facilitate closed-loop control strategies that actively prevent lithium plating during fast charging. This approach moves beyond traditional, static charging protocols to an adaptive, physics-informed method that maximizes charge rate while respecting the fundamental safety limits of the chemistry.
The success of this methodology hinges on accurate model parameterization and validation. As the technology matures, future work will focus on further reducing model complexity, developing robust online parameter identification algorithms to account for aging in the energy storage battery, and integrating thermal effects for a complete electro-thermal-safe charging strategy. Ultimately, the widespread adoption of such intelligent charging controls will enhance the safety, longevity, and economic viability of large-scale lithium-ion energy storage battery systems, playing a crucial role in the global transition to sustainable energy grids.