With the deepening of the “dual carbon” strategy and the reform of the electric power system, an increasing number of renewable energy sources are being integrated into the power grid. This integration makes the frequency stability of traditional regulating units more vulnerable. There is an urgent need for new frequency regulation control strategies to address this challenge. The battery energy storage system (BESS), capable of rapid charging and discharging and providing fast response to grid regulation, plays a crucial role in improving the frequency regulation performance of regional power grids under these new circumstances. Current research on energy storage participating in power frequency regulation control is progressing vigorously, but further investigation is still required to enhance the response speed of battery energy storage system involvement in grid frequency regulation and its practical application. Based on an analysis of the operational mechanism of energy storage participating in power system frequency regulation, this paper establishes a simulation model for the frequency regulation characteristics of a battery energy storage system in the grid and proposes a control optimization strategy. Simulation experiments demonstrate that the proposed control strategy is practical and efficient, providing a valuable reference for studying the control strategies and practical applications of battery energy storage system participation in grid frequency regulation.
Power System Frequency Control Model
Traditional Frequency Control Process
Grid frequency regulation is divided into primary and secondary frequency regulation. When the balance of active power in the grid is disturbed by load changes, the primary frequency regulation system of the unit governor initiates action. It adjusts the valve opening based on the frequency deviation to change the unit’s active power output. Primary frequency control requires a rapid response to frequency disturbances, typically within seconds. However, due to the characteristics of generators, it is difficult to achieve zero-error regulation. Therefore, secondary frequency regulation is needed to further correct the power signal. The basic principle of grid control is illustrated below:
Equivalent Model Establishment
In the traditional frequency regulation process, power changes of generators are slow. With the involvement of a battery energy storage system, the active power can be rapidly controlled, enabling the unit to achieve fast response regulation. A typical frequency control model consists of several modules: the governor model, the turbine model, and the generator-load model. Their transfer functions and equivalent models are well-defined. Therefore, this paper focuses on the equivalent model of the battery energy storage system. The BESS model is derived from the charging and discharging processes of the battery. Following common practice, a first-order inertial link function is used to represent the BESS:
$$
G_B(s) = \frac{1}{1 + s T_B}
$$
where \(T_B\) is the time response constant of BESS output. The state of charge (SOC) is defined as \(S(t) = E(t)/E_N\), with \(E_N\) being the rated capacity. The initial SOC is \(S_0\). The charging/discharging power is denoted as \(P_B(t)\). The dynamic equation for SOC is:
$$
\frac{dS(t)}{dt} = -\frac{P_B(t)}{E_N}
$$
Based on this equivalent model, the load frequency control model with BESS participation is constructed. The overall transfer function representation includes the governor, turbine, generator-load, and BESS as interconnected blocks. The battery storage system contributes a parallel path for fast power injection.
Control Strategy for BESS Participation in Grid Frequency Regulation
Control Framework
Considering the frequency regulation requirements, the rapid response advantages of BESS, and the coordination characteristics among multiple storage units, a two-layer distributed frequency control strategy is proposed. The control structure consists of an upper layer (regional dispatching layer) and a lower layer (storage system control layer). The upper layer is responsible for coordinating the distribution of frequency regulation signals between traditional units and BESS. The lower layer handles the communication control among multiple distributed BESS units participating in the frequency regulation.
Regional Dispatching Layer Control Strategy
The role of the regional dispatching layer is to divide the regulation interval based on the Area Control Error (ACE) signal and provide optimal reference output power for each regulation zone through a coordinated control strategy. The control logic for different zones is described below.
Dead Zone and Emergency Zone Control
When the ACE signal falls within the dead zone (small disturbance), the system external disturbance is small. In this case, the BESS recovers its SOC to the rated state as soon as possible, without actively participating in frequency regulation. When the ACE signal falls into the emergency zone (large disturbance), the BESS reference output power is determined by:
$$
\Delta P_{B,ref} = \text{sign}(\Delta P_{AGC}) \cdot \min\left( |\Delta P_{AGC}|, P_{B,max}^{discharge} \right) \quad \text{or} \quad \min\left( |\Delta P_{AGC}|, P_{B,max}^{charge} \right)
$$
where \(\Delta P_{AGC}\) is the total grid frequency control demand power, \(P_{B,max}^{discharge}\) and \(P_{B,max}^{charge}\) are the maximum discharging and charging powers of the BESS, respectively. Positive values indicate discharging, and negative values indicate charging.
Normal Regulation Zone Control
When the ACE signal lies within the normal regulation zone, the control logic is as follows. If the BESS SOC is within the ideal range \([SOC_0, SOC_1]\), the BESS responds to the frequency regulation signal with priority. If the SOC is outside the ideal range but the current power demand helps restore the SOC towards the ideal range, the BESS participates with its maximum allowable power. Otherwise, to protect the battery, the BESS does not participate in this regulation cycle. The conditions are expressed as:
$$
\begin{cases}
P_{B,ref} = \Delta P_{AGC}, & \text{if } SOC_0 \leq S \leq SOC_1 \\
P_{B,ref} = \text{sign}(\Delta P_{AGC}) \cdot P_{B,max}, & \text{if } S < SOC_0 \text{ and } \Delta P_{AGC} < 0 \text{ (charging helps)} \\
P_{B,ref} = \text{sign}(\Delta P_{AGC}) \cdot P_{B,max}, & \text{if } S > SOC_1 \text{ and } \Delta P_{AGC} > 0 \text{ (discharging helps)} \\
P_{B,ref} = 0, & \text{otherwise}
\end{cases}
$$
Storage System Layer Control Strategy
After the regional layer determines whether the BESS should participate and the reference power level, the next step is coordination among multiple distributed BESS units. A distributed consensus algorithm based coordination control is proposed to minimize the overall operating cost while fulfilling the frequency regulation requirements. The cost function for each BESS unit \(i\) at time step \(k\) is defined as:
$$
C_{i,k} = a_i \left( \frac{P_{Bi,k}}{E_{BN,i}} \right)^2 + b_i \left( S_{bi,k} – S_{bi}^* \right)^2
$$
where \(a_i\) and \(b_i\) are weighting factors for power output cost and SOC deviation cost, respectively; \(P_{Bi,k}\) is the active power output of unit \(i\) at time \(k\); \(E_{BN,i}\) is the rated energy capacity; \(S_{bi,k}\) is the state of charge; and \(S_{bi}^*\) is the optimal SOC (e.g., 0.6).
To account for battery lifetime degradation, the depth of discharge (DOD) indicator is introduced. A larger DOD leads to faster degradation. The lifetime model is given by:
$$
\delta = \frac{1}{\lambda_1 \cdot DOD + \lambda_2}
$$
where \(\lambda_1\) and \(\lambda_2\) are coefficients. The initial investment cost of unit \(i\) is \(C_{0,i}\). Under constant voltage and constant current operation, the output energy is proportional to DOD. The battery loss cost related to output energy is:
$$
C_{loss,i,k} = \frac{E_{Bi,k}}{E_{BN,i}} \cdot \frac{C_{0,i}}{\lambda_1 \delta + \lambda_2}
$$
where \(E_{Bi,k}\) is the energy output of unit \(i\) at time \(k\). The comprehensive cost for BESS participating in frequency regulation is then:
$$
J = \sum_{i} \left( C_{i,k} + C_{loss,i,k} \right)
$$
To minimize \(J\) subject to the power balance constraint \(\sum_i P_{Bi,k} = P_{total,ref,k}\), a Lagrangian function is formulated:
$$
\mathcal{L} = \sum_i \left( C_{i,k} + C_{loss,i,k} \right) + \lambda \left( \sum_i P_{Bi,k} – P_{total,ref,k} \right)
$$
The optimality condition requires:
$$
\frac{\partial C_{i,k}}{\partial P_{Bi,k}} + \frac{\partial C_{loss,i,k}}{\partial P_{Bi,k}} + \lambda = 0
$$
Let the incremental cost \(\mu_i = \frac{\partial C_{i,k}}{\partial P_{Bi,k}} + \frac{\partial C_{loss,i,k}}{\partial P_{Bi,k}}\). The consensus algorithm updates each unit’s incremental cost through communication with neighbors. The update rule is:
$$
\mu_i^{(t+1)} = \mu_i^{(t)} + \epsilon \sum_{j \in \mathcal{N}_i} w_{ij} \left( \mu_j^{(t)} – \mu_i^{(t)} \right)
$$
where \(\epsilon\) is a step size, \(\mathcal{N}_i\) is the set of neighbors, and \(w_{ij}\) are communication weights. When all \(\mu_i\) converge to a common value \(\mu^*\), the power outputs are determined by solving the local optimality condition. This consensus-based approach ensures that the total power is shared among BESS units in a cost-efficient manner while respecting SOC constraints.
Simulation and Analysis
A Simulink model is built based on the proposed two-layer control structure. Four distributed BESS units are configured with parameters listed in the table below.
| Unit | Power Rating (MW) | Rated Capacity (MWh) | Charge/Discharge Efficiency | Time Constant (s) | Power Cost Coefficient \(a_i\) | SOC Deviation Cost Coefficient \(b_i\) | Initial SOC |
|---|---|---|---|---|---|---|---|
| 1 | ±15 | 3.5 | 0.90/0.90 | 0.01 | 0.90 | 0.65 | 0.6 |
| 2 | ±14 | 3.0 | 0.95/0.90 | 0.01 | 0.90 | 0.73 | 0.6 |
| 3 | ±13 | 2.5 | 0.90/0.85 | 0.01 | 0.90 | 0.86 | 0.6 |
| 4 | ±12 | 2.2 | 0.90/0.90 | 0.01 | 0.85 | 1.35 | 0.6 |
The system is subjected to a step load disturbance of 0.2 pu (per unit) at t = 5 s. The simulation compares the traditional frequency control model (without BESS) with the proposed BESS-integrated control. The results are shown in the following curves. The frequency deviation plot demonstrates that with BESS participation, the maximum frequency deviation is significantly reduced (from about 0.04 Hz to 0.015 Hz) and the settling time is shortened. The power output curve indicates that BESS units respond quickly, providing immediate power support before the traditional generators ramp up. The SOC trajectories of the four BESS units show that each unit remains within safe limits (0.2–0.9) throughout the regulation process, and the consensus algorithm ensures balanced participation without any unit being drained excessively.
Further simulation cases with different load patterns (random fluctuations, ramp disturbances) confirm the robustness of the proposed strategy. The results consistently show that the two-layer distributed control effectively utilizes the fast response capability of battery energy storage system while managing SOC and battery lifetime. Compared to a centralized approach, the distributed method reduces communication overhead and improves scalability.
Conclusion
This paper addresses the challenge of how distributed battery energy storage system units can optimally participate in grid frequency control. A two-layer distributed control strategy is proposed. The upper layer defines zone-based regulation rules to protect the battery and allocate power between conventional units and BESS. The lower layer employs a consensus algorithm to coordinate multiple BESS units, minimizing a combined cost that includes power deviation, SOC regulation, and battery degradation. Simulation results demonstrate that with BESS participation, the frequency response is faster, the maximum deviation is smaller, and the control is more efficient. The consensus-based coordination ensures that each unit shares the regulation burden in a cost-effective manner without overloading any single battery. This work provides a practical framework for integrating battery energy storage system into modern power system frequency regulation. Future work will consider communication delays, hardware-in-the-loop validation, and adaptive tuning of cost coefficients.