The stability of modern power systems hinges on the continuous balance between generation and consumption. Secondary frequency regulation, or Automatic Generation Control (AGC), is the cornerstone of maintaining this balance within each control area, correcting the slow deviations not addressed by primary response. However, the increasing integration of variable renewable energy sources and the evolving load profiles present significant challenges to traditional frequency regulation resources like thermal and hydro units. These conventional resources often have inherent limitations in their ramp rates and response times, making it difficult to track the faster and more volatile regulation signals efficiently.
This is where the role of energy storage batteries becomes pivotal. With advancements in electrochemical technology, Battery Energy Storage Systems (BESS) are transitioning from small-scale, distributed applications to large-scale, grid-connected installations. An energy storage battery system offers unparalleled advantages for frequency regulation: millisecond-level response times, high precision in tracking control signals, and the unique capability for bi-directional power flow (charging and discharging). These attributes make large-scale energy storage battery plants a highly valuable and flexible resource for grid operators. Their fast response can significantly improve the quality and speed of frequency restoration.
Despite the clear benefits, integrating a large-scale energy storage battery system into secondary frequency control presents distinct challenges that require sophisticated coordination strategies. A large BESS is typically composed of numerous parallel energy storage battery units or racks. Two critical, interconnected issues must be addressed: the state of charge (SOC) management of the entire storage plant, and the SOC balancing among its individual units.
Firstly, the aggregate SOC of the BESS is a finite resource. Continuous discharge will deplete the batteries, while continuous charging will fill them. In either extreme state, the plant’s ability to provide regulation in both directions is severely compromised. Therefore, a high-level control strategy must coordinate the BESS with traditional generators, not only to meet the Area Regulation Requirement (ARR) but also to actively manage the plant’s overall SOC, keeping it within a desired operating range to ensure sustained availability.
Secondly, within the BESS plant itself, the SOC of individual energy storage battery units will naturally diverge due to manufacturing tolerances, uneven aging, and slight differences in operational history. Allowing significant SOC imbalances is detrimental: it reduces the plant’s total available regulation capacity (as some units hit limits sooner than others), leads to inefficient and uneven aging of the battery packs, and complicates energy management. Therefore, a lower-level control strategy is needed to ensure that all units work in harmony to deliver the total power commanded by the grid operator while driving their individual SOCs towards consensus.
In this article, I propose a hierarchical, two-layer control framework designed to tackle these challenges holistically. The upper layer, operating at the Area Control Center, focuses on optimal power dispatch between traditional units and the BESS, explicitly considering the plant’s aggregate SOC. The lower layer, operating at the BESS station’s Energy Management System (EMS), employs a distributed cooperative control algorithm to precisely track the dispatch command while synchronizing the SOCs of all individual energy storage battery units. The efficacy of this integrated strategy is demonstrated through detailed simulation studies on a two-area interconnected power system model.
System Modeling and Problem Formulation
Grid-Level Load Frequency Control Model with BESS
The power system for analysis is modeled as a two-area interconnection, a standard benchmark for studying AGC dynamics. Each area contains traditional generation, load, and a large-scale BESS. The overall LFC model integrates the dynamics of traditional generation, network frequency, tie-line power, and the battery storage system.
The traditional generation in each area is represented by an equivalent non-reheat turbine unit. Its dynamics encompass the governor, turbine, and rotating mass/inertia of the area. The key equations governing the area dynamics are based on the swing equation and power balance. The Area Control Error (ACE) for area i is defined as:
$$ACE_i = \Delta P_{tie,i} + \beta_i \Delta f_i$$
where $\Delta P_{tie,i}$ is the tie-line power deviation and $\beta_i$ is the frequency bias factor. A Proportional-Integral (PI) controller processes the ACE to produce the total required regulation power for the area, $\Delta P_{R,i}$:
$$\Delta P_{R,i} = -(K_{P,i} + \frac{K_{I,i}}{s}) \cdot ACE_i$$
This signal, $\Delta P_{R,i}$, is the total power command that must be distributed between the traditional unit and the BESS in the proposed strategy.
Modeling the Large-Scale Energy Storage Battery System
A large-scale BESS is an aggregation of multiple, parallel energy storage battery units. Each unit i consists of the battery pack, a Power Conversion System (PCS), and associated controls. For system-level dynamic studies focused on secondary frequency regulation (timeframes of seconds to minutes), the very fast switching dynamics of the PCS can be neglected. The power output of a unit can be modeled as closely tracking its reference command, often represented by a first-order lag with a very small time constant $T_{BES,i}$:
$$G_{BES,i}(s) = \frac{P_{BES,i}(s)}{P_{BES,ref,i}(s)} = \frac{1}{1 + sT_{BES,i}}$$
where $T_{BES,i}$ is typically in the order of milliseconds. For the slower AGC timescale, this can often be approximated as unity gain ($P_{BES,i} \approx P_{BES,ref,i}$).
The core dynamic state of an energy storage battery is its State of Charge (SOC). The SOC for unit i, denoted $S_{SOC,i}(t)$, is the ratio of its stored energy $E_i(t)$ to its rated energy capacity $E_{max,i}$. The evolution of SOC is directly tied to the unit’s power output:
$$\frac{dE_i(t)}{dt} = -P_{BES,i}(t)$$
$$S_{SOC,i}(t) = \frac{E_i(t)}{E_{max,i}}$$
Combining these, the SOC dynamics are:
$$\frac{dS_{SOC,i}(t)}{dt} = -\frac{1}{E_{max,i}} P_{BES,i}(t)$$
To facilitate the design of the distributed control law at the lower layer, I define an auxiliary state variable $S_i(t) = -3600 \cdot E_{max,i} \cdot S_{SOC,i}(t)$, which has units consistent with power-time. By also modeling the power reference signal as the integral of a new control input $u_i(t)$, the dynamics of a single energy storage battery unit can be expressed as a canonical second-order integrator model, ideal for consensus algorithm design:
$$
\begin{aligned}
\dot{S}_i(t) &= P_{BES,i}(t) \\
\dot{P}_{BES,i}(t) &= u_i(t)
\end{aligned}
$$
A plant with n such units constitutes a multi-agent system with second-order dynamics. The control objectives at this level are twofold: 1) ensure the sum of all $P_{BES,i}(t)$ tracks the plant’s total power command $P_{BES,ref}$ from the upper layer, and 2) drive all $S_i(t)$ (and hence all $S_{SOC,i}(t)$) to a common value.
Hierarchical Control Strategy
The proposed control architecture is explicitly designed to manage the interaction between grid-level needs and energy storage battery operational constraints. The hierarchy naturally decomposes the problem into an upper-layer economic/strategic dispatch and a lower-layer real-time tactical control.
Upper Layer: Area Control Center Dispatch Considering Aggregate BESS SOC
At the Area Control Center, the total regulation requirement $\Delta P_R$ is categorized into different zones based on its magnitude: a Dead Zone, a Normal Regulation Zone, and an Emergency Zone. The dispatch logic between the traditional generator ($\Delta P_{G,ref}$) and the BESS ($P_{BES,ref}$) varies per zone and incorporates the aggregate BESS SOC ($S_{SOC,avg}$). The goal is to satisfy $\Delta P_R = \Delta P_{G,ref} + P_{BES,ref}$ while guiding $S_{SOC,avg}$ towards a predefined healthy range $[S_{SOC,L}, S_{SOC,U}]$, typically around 0.3 to 0.7.
The dispatch rules are summarized in the following table:
| Regulation Zone | Condition on $|\Delta P_R|$ | Dispatch Logic for $P_{BES,ref}$ | Rationale |
|---|---|---|---|
| Emergency | $> P_{R,em}$ | BESS operates at its maximum power limit ($\pm P_{BES,max}$) in the direction needed to reduce ACE. Traditional generator covers the residual. | Maximize fast response to arrest large frequency deviations. SOC recovery is a secondary concern during emergencies. |
| Normal | $P_{R,dz} < |\Delta P_R| \leq P_{R,em}$ | If $S_{SOC,L} \leq S_{SOC,avg} \leq S_{SOC,U}$: Proportional dispatch based on capacities. | Fair participation when BESS is in a healthy state. |
| If $S_{SOC,avg}$ is outside limits and the required $P_{BES,ref}$ helps restore SOC: BESS operates at max power. | Prioritize SOC recovery if the regulation signal assists it. | ||
| If $S_{SOC,avg}$ is outside limits and the required $P_{BES,ref}$ harms SOC: BESS output is zero. | Protect BESS from further depletion/saturation. Traditional generator handles all regulation. | ||
| Dead Zone | $\leq P_{R,dz}$ | If $S_{SOC,avg}$ is within limits: $P_{BES,ref}=0$. | Let traditional units handle tiny deviations, readying BESS for future use. |
| If $S_{SOC,avg} > S_{SOC,U}$: Discharge at a rate proportional to the SOC excess. | Gentle, controlled self-recovery to avoid triggering ACE. | ||
| If $S_{SOC,avg} < S_{SOC,L}$: Charge at a rate proportional to the SOC deficit. | Gentle, controlled self-recovery to avoid triggering ACE. |
The mathematical formulation for the Normal Zone logic, for instance, is as follows:
$$
P_{BES,ref} =
\begin{cases}
\frac{P_{BES,max}}{\Delta P_{G,max} + P_{BES,max}} \Delta P_R, & \text{if } S_{SOC,L} \leq S_{SOC,avg} \leq S_{SOC,U} \\
\min(P_{BES,max}, \Delta P_R), & \text{if } (S_{SOC,avg} < S_{SOC,L} \text{ and } \Delta P_R > 0) \text{ or } (S_{SOC,avg} > S_{SOC,U} \text{ and } \Delta P_R < 0) \\
0, & \text{otherwise}
\end{cases}
$$
$$\Delta P_{G,ref} = \Delta P_R – P_{BES,ref}$$
This intelligent dispatch ensures the energy storage battery plant is used aggressively when most needed (emergencies), is protected from harmful deep cycling, and is actively maintained in a state of readiness.
Lower Layer: BESS Station-Level Distributed Cooperative Control
Upon receiving the power setpoint $P_{BES,ref}$ from the upper layer, the BESS station’s EMS must execute it precisely. This involves coordinating the n individual energy storage battery units. The objectives are: 1) Power Tracking: $\sum_{i=1}^n P_{BES,i} = P_{BES,ref}$. 2) SOC Consensus: $S_{SOC,1} = S_{SOC,2} = … = S_{SOC,n}$.
This is a classic leader-following consensus problem in multi-agent systems. I design a virtual “leader” agent (index 0) within the EMS, whose dynamics are defined to track the station’s power command:
$$
\begin{aligned}
\dot{S}_0(t) &= P_0(t) \\
\dot{P}_0(t) &= K_0 (P_{BES,ref}(t) – P_{BES,meas}(t))
\end{aligned}
$$
where $P_{BES,meas}(t)$ is the measured total output at the point of connection, and $K_0 > 0$ is a gain. The states of this leader ($S_0$, $P_0$) represent the desired common trajectory for all follower units.
The n physical energy storage battery units are the follower agents. They communicate over a sparse network. Let $a_{ij} > 0$ if unit i receives information from unit j, and $b_i > 0$ if unit i has direct access to the leader’s states. The communication graph must be connected. A finite-time convergent distributed control law for each follower unit i is designed as:
$$
\begin{aligned}
u_i(t) = &-\lambda_1 \text{sgn}\left( \sum_{j=1}^n a_{ij}(S_i – S_j) + b_i(S_i – S_0) \right) \\
&-\lambda_2 \left( \sum_{j=1}^n a_{ij}(S_i – S_j) + b_i(S_i – S_0) \right) \\
&-\lambda_3 \left( \sum_{j=1}^n a_{ij}(P_i – P_j) + b_i(P_i – P_0) \right)
\end{aligned}
$$
where $\lambda_1, \lambda_2, \lambda_3 > 0$ are control gains, and $\text{sgn}(\cdot)$ is the signum function. This algorithm ensures that the states of all follower units ($S_i$, $P_i$) converge to the leader’s states ($S_0$, $P_0$) in finite time. Convergence to $P_0$ implies precise power tracking at the station level. Convergence to $S_0$ implies that all $S_i$ are equal, which, given their definition ($S_i = -3600 \cdot E_{max,i} \cdot S_{SOC,i}$), forces the SOCs of all energy storage battery units to a common value, provided their capacities $E_{max,i}$ are equal (or are normalized appropriately).
Simulation Case Study
To validate the proposed hierarchical strategy, I implemented a detailed simulation model of a two-area interconnected system in MATLAB/Simulink. The key parameters are listed below.
| Parameter | Area 1 | Area 2 |
|---|---|---|
| Governor Time Constant $T_g$ (s) | 0.08 | 0.10 |
| Turbine Time Constant $T_t$ (s) | 0.30 | 0.35 |
| Generator Droop $R$ (Hz/pu) | 2.4 | 2.5 |
| Inertia Constant $H$ (pu·s) | 0.1667 | 0.2000 |
| Damping Coefficient $D$ (pu/Hz) | 0.0083 | 0.0100 |
| Frequency Bias $\beta$ (pu/Hz) | 0.425 | 0.450 |
Area 1 is equipped with a BESS comprising 12 identical units. Each energy storage battery unit has a power rating of 2 MW and an energy capacity of 2.5 MWh, making the total plant rating 24 MW / 30 MWh. The desired SOC operating range is set to $[0.3, 0.7]$. The regulation zones are defined by $P_{R,dz} = 0.005$ pu and $P_{R,em} = 0.08$ pu (on a 1000 MVA base). The PI controller gains are $K_P=0.16$, $K_I=0.48$. The lower-layer consensus controller uses gains $\lambda_1=5$, $\lambda_2=10$, $\lambda_3=1$, with $K_0=0.15$ for the leader agent.
Scenario 1: Small Load Disturbance and SOC Consensus
A small, varying load disturbance is applied in Area 1. The initial SOCs of the 12 energy storage battery units are intentionally set unevenly, ranging from 51% to 62%. The results demonstrate the dual functionality of the lower-layer controller. The system frequency deviation and ACE are quickly regulated to zero, satisfying AGC performance standards. Crucially, the BESS plant’s total output accurately tracks the command $P_{BES,ref}$ calculated by the upper-layer dispatch. Simultaneously, the distributed control algorithm drives the disparate SOCs of all units to converge to a common value within approximately 5 minutes, as shown in the simulated response. Throughout the simulation, the upper-layer dispatch logic successfully maintains the average SOC well within the target $[0.3, 0.7]$ range.
Scenario 2: Large Load Disturbance and Strategy Comparison
A significant step load change of +0.12 pu is applied in Area 1, placing the system in the Emergency Zone. The initial SOCs are set low (20% to 42%) to test the strategy’s robustness. I compare the proposed strategy against a naive baseline strategy where power is dispatched proportionally between the generator and BESS without SOC consideration, and within the BESS, power is simply divided equally among units without SOC consensus.
The comparison reveals the critical advantages of the proposed method:
| Aspect | Proposed Strategy | Baseline Strategy (No SOC Management) |
|---|---|---|
| Frequency Recovery | Fast recovery; ACE settles within ~1 minute. | Similar recovery, as total regulation capacity is sufficient. |
| BESS Power Tracking | Precise tracking of $P_{BES,ref}$ throughout. | Tracking error occurs as some units hit low SOC limits and disconnect, forcing the traditional unit to compensate unexpectedly. |
| Aggregate SOC Management | Actively managed, stays within or recovers to healthy range. | Unmanaged; SOC can drift to extremes, depleting usable capacity. |
| Individual SOC Balance | All units converge to the same SOC, maximizing available capacity and ensuring uniform aging. | Significant SOC divergence persists; units with the lowest SOC become bottlenecks, reducing effective plant capacity. |
| Sustained Regulation Capability | High. The BESS is kept in a ready state for subsequent disturbances. | Degraded over time. After deep cycles, parts of the BESS may be unavailable, reducing total grid support. |
The simulation waveforms clearly show that under the baseline strategy, the lowest-SOC energy storage battery units hit their discharge cut-off limit (e.g., 5%) and stop contributing. This causes a drop in the total BESS output, creating a power imbalance that must be covered by the traditional generator. If the traditional generator is already at its limit, this would directly lead to poor frequency regulation. In contrast, the proposed strategy maintains all units in a synchronized, operational state, ensuring the full 24 MW capacity is reliably available. Furthermore, the upper-layer logic initiates SOC recovery during quiet periods, demonstrating the holistic management of the energy storage battery resource.
Conclusion
The integration of large-scale energy storage battery systems into power system secondary frequency regulation offers a transformative opportunity to enhance grid stability and performance. However, realizing this potential requires advanced control strategies that go beyond treating the BESS as a simple, monolithic asset. The hierarchical framework I have presented directly addresses the two most pressing operational challenges: system-level SOC sustainability and unit-level SOC balancing.
The upper-layer dispatch algorithm acts as an intelligent mediator between grid needs and energy storage battery health. By partitioning the regulation demand into zones and incorporating the aggregate SOC into the decision logic, it strategically deploys the fast-responding BESS while proactively preventing its depletion or saturation. This ensures the asset remains a credible and enduring participant in frequency regulation markets.
The lower-layer distributed cooperative control algorithm transforms the collection of individual energy storage battery units into a cohesive, self-organizing system. It guarantees precise tracking of the station’s power command while autonomously driving all units towards SOC consensus. This not only maximizes the plant’s instantaneous available power but also promotes longevity and operational simplicity.
The simulation studies on a two-area interconnected system provide compelling evidence of the strategy’s effectiveness. The control scheme successfully manages large disturbances, maintains frequency, and accomplishes the dual objectives of power tracking and SOC consensus. Compared to a strategy that ignores SOC management, the proposed approach demonstrates superior reliability and sustainability in utilizing the energy storage battery resource.
Future work could extend this framework to heterogeneous fleets of storage with different ratings or chemistries, and to scenarios involving widely distributed, aggregated energy storage battery resources participating in grid services. Nevertheless, the principles established here—hierarchical coordination and distributed consensus—provide a robust foundation for the sophisticated control systems needed to harness the full potential of large-scale energy storage batteries in the future power grid.