In modern power systems, the energy storage battery plays a critical role in frequency regulation due to its fast response and bidirectional power capability. However, the operational life of the energy storage battery remains a primary concern, as frequent deep charge/discharge cycles accelerate degradation. This paper presents an adaptive comprehensive frequency regulation strategy based on state-of-charge (SOC) feedback, which optimally allocates power output between virtual inertia control and virtual droop control. The adaptive factor is determined by the input coefficient from a fuzzy logic controller and the feedback coefficient from the real-time SOC of the energy storage battery. The input coefficient is governed by the frequency regulation output proportion coefficient \(K\), which balances the two control modes. A simulation model is built on the Matlab/Simulink platform using actual operational data from a power plant frequency regulation energy storage battery. The rain-flow counting method is employed to predict the service life of the energy storage battery under different control strategies. Results demonstrate that the proposed adaptive strategy extends the energy storage battery life by 25.53% compared with the current plant strategy, and by 38.19% and 22.42% over the fixed \(K\) method and variable \(K\) method, respectively. The findings confirm the effectiveness of the adaptive strategy in improving both frequency regulation performance and energy storage battery longevity.
1. Introduction
The energy storage battery, as an advanced energy storage and release device, offers advantages such as rapid response, accurate tracking, and bidirectional regulation, making it an ideal choice for grid peak shaving and frequency modulation. In practice, virtual droop control and virtual inertia control are the two main approaches for energy storage battery participation in frequency regulation. Reasonable allocation of power output between these two modes can smooth the energy storage battery’s power output, maintain frequency stability, and prevent overcharge/overdischarge, thereby prolonging the service life of the energy storage battery. To achieve this, an adaptive factor is introduced to dynamically adjust the output proportion of the two control modes based on frequency parameters and the SOC of the energy storage battery. Building a model that accurately reflects the real-world operation of the energy storage battery in a power plant is essential for improving its comprehensive regulation performance. For life prediction of the energy storage battery, both performance-based and experience-based methods exist. Among experience-based approaches, the cycle counting method, particularly the rain-flow counting method, is widely adopted in the current electricity market environment because it can accurately represent the complex charge/discharge processes of the energy storage battery and improve the precision of life estimation. Deep and frequent charge/discharge cycles severely degrade the energy storage battery, and the rain-flow counting method, by extracting peak and valley values of cycles and selecting extreme points, enables a comprehensive evaluation of the energy storage battery’s life. This study focuses on a lithium iron phosphate energy storage battery and adopts an adaptive comprehensive frequency regulation strategy to smoothly allocate the output proportion \(K\) between virtual inertia control and virtual droop control. The rain-flow counting method is used to obtain the equivalent cycle number of the energy storage battery, and the service life is predicted using actual data from a power plant. The proposed strategy is compared with fixed \(K\) and variable \(K\) methods to validate its effectiveness.
2. Frequency Regulation Control Strategy
Virtual inertia control can effectively suppress further deterioration of the frequency deviation rate, while virtual droop control can reduce the steady-state frequency deviation. Combining the advantages of both and rationally distributing their proportions is key to achieving optimal frequency regulation.
2.1 Control Model of Energy Storage Battery Participating in Frequency Regulation
The frequency deviation \(\Delta f\) is calculated as:
\[
\Delta f = \frac{\Delta P_b + \Delta P_g – \Delta P_{agc}}{H_s + D}
\]
where \(\Delta P_b\) is the active power output change of the energy storage battery system, \(\Delta P_g\) is the active power output change of the thermal power unit (considered known in this study), \(\Delta P_{agc}\) is the change in the grid AGC dispatch command, and \(H_s\) and \(D\) are the generator inertia constant and damping coefficient, respectively.
The actual output of the energy storage battery system consists of two parts: virtual inertia control output and virtual droop control output. An adaptive factor \(\mu\) is introduced to allocate the output proportion between them:
\[
\begin{aligned}
\Delta P_b &= \Delta P_{b,I} + \Delta P_{b,D} \\
\Delta P_{b,I} &= K’ \frac{d \Delta f}{dt} G_{bess}(s) = \mu_1 \cdot M_b \cdot \frac{d \Delta f}{dt} G_{bess}(s) \\
\Delta P_{b,D} &= K \cdot \Delta f \cdot G_{bess}(s) = \mu_2 \cdot K_b \cdot \Delta f \cdot G_{bess}(s)
\end{aligned}
\]
Here, \(\Delta P_{b,I}\) and \(\Delta P_{b,D}\) are the active power output changes of virtual inertia control and virtual droop control after adaptive adjustment. \(K’\) and \(K\) are the output proportion coefficients for virtual inertia and virtual droop control, respectively. \(M_b\) and \(K_b\) are the power regulation coefficients. \(G_{bess}(s)\) is the transfer function of the energy storage battery system:
\[
G_{bess}(s) = \frac{1}{T_b s + 1}
\]
where \(T_b\) is the time constant of the energy storage battery system. The adaptive factor \(\mu\) is composed of the input coefficient \(\alpha\) and the feedback coefficient \(\beta\):
\[
\mu_1 = \alpha_1 \cdot \beta_1, \quad \mu_2 = \alpha_2 \cdot \beta_2
\]
with \(\alpha_1 + \alpha_2 = 1\).
2.2 Input Coefficient via Fuzzy Logic Controller
The critical threshold for switching between control modes is difficult to quantify precisely, which may cause frequency instability and accelerate the degradation of the energy storage battery. To avoid this, a fuzzy logic controller is used to adjust parameters. The input variables are frequency deviation \(\Delta f\) and frequency deviation change rate \(d\Delta f/dt\). The output is the input coefficient \(\alpha_1\) (the proportion of virtual inertia control). Based on the grid standard and actual data, the ranges are set: \(\Delta f \in [-0.5, 0.5]\), \(d\Delta f/dt \in [-1, 1]\), and \(\alpha_1 \in [0, 1]\). Triangular membership functions are used. The fuzzy sets for inputs are {NB, NM, NS, Z, PS, PM, PB} (negative big to positive big), and for the output are {Z, S, M, B, VB} (zero, small, medium, big, very big). The fuzzy control rules are shown in Table 1.
| \(d\Delta f/dt\) | \(\Delta f\) | ||||||
|---|---|---|---|---|---|---|---|
| NB | NM | NS | Z | PS | PM | PB | |
| NB | VB | B | B | VB | Z | Z | Z |
| NM | B | B | M | B | Z | Z | S |
| NS | M | M | M | M | Z | S | S |
| Z | Z | S | Z | Z | Z | S | Z |
| PS | S | S | Z | M | M | M | M |
| PM | S | Z | Z | B | M | B | B |
| PB | Z | Z | Z | VB | B | B | VB |
Defuzzification uses the centroid method. The design principle: when \(\Delta f\) is small but \(d\Delta f/dt\) is large, larger \(\alpha_1\) (more virtual inertia) is used to prevent frequency rate deviation from worsening. When \(\Delta f\) is large but \(d\Delta f/dt\) is small, smaller \(\alpha_1\) (more virtual droop) is used to restore frequency quickly. When both are small, the energy storage battery output is reduced to preserve its life.
2.3 Feedback Coefficient Based on SOC of Energy Storage Battery
To prevent overcharge/overdischarge, a feedback coefficient based on the SOC of the energy storage battery is introduced. The Logistic function is used to define charging coefficient \(K_c\) and discharging coefficient \(K_d\):
\[
K_c =
\begin{cases}
\frac{P_0 K_{\text{max}} \exp\left(v_0 \frac{0.8 – S_{SOC}}{(0.8-0.23)/2}\right)}{K_{\text{max}} + P_0 \left[\exp\left(v_0 \frac{0.8 – S_{SOC}}{(0.8-0.23)/2}\right) – 1\right]}, & S_{SOC} \in [0.23, 0.8] \\
K_{\text{max}}, & S_{SOC} \in [0, 0.23) \\
0, & S_{SOC} \in (0.8, 1]
\end{cases}
\]
\[
K_d =
\begin{cases}
\frac{P_0 K_{\text{max}} \exp\left(v_0 \frac{S_{SOC} – 0.23}{(0.8-0.23)/2}\right)}{K_{\text{max}} + P_0 \left[\exp\left(v_0 \frac{S_{SOC} – 0.23}{(0.8-0.23)/2}\right) – 1\right]}, & S_{SOC} \in [0.23, 0.8] \\
0, & S_{SOC} \in [0, 0.23) \\
K_{\text{max}}, & S_{SOC} \in (0.8, 1]
\end{cases}
\]
where \(K_{\text{max}} = 1\), \(P_0 = 0.01\), \(v_0 = 15\). The feedback coefficients \(\beta_1\) and \(\beta_2\) are then determined linearly:
\[
\beta_1 =
\begin{cases}
\lambda K_c, & \frac{d\Delta f}{dt} \geq 0 \\
\lambda K_d, & \frac{d\Delta f}{dt} < 0
\end{cases}
\quad
\beta_2 =
\begin{cases}
K_c, & \Delta f \geq 0 \\
K_d, & \Delta f < 0
\end{cases}
\]
where \(\lambda\) is the ratio of virtual inertia coefficient to virtual droop coefficient, set to 1/3 based on the actual grid disturbance characteristics.
2.4 Adaptive Comprehensive Frequency Regulation Control
The final virtual droop control coefficient \(K\) and virtual inertia coefficient \(K’\) with the adaptive factor are:
\[
K’ = \mu_1 \cdot M_b = \alpha_1 \cdot \beta_1 \cdot M_b
\]
\[
K = \mu_2 \cdot K_b = \alpha_2 \cdot \beta_2 \cdot K_b = (1 – \alpha_1) \cdot \beta_2 \cdot K_b
\]
By adjusting \(K\) dynamically, the energy storage battery output is allocated between the two control modes. Three methods are compared: fixed \(K\) (constant \(K = K_b\)), variable \(K\) (only SOC feedback, \(K = \beta_2 \cdot K_b\)), and the proposed adaptive method (\(K = \alpha_2 \cdot \beta_2 \cdot K_b\)). The average service life of the energy storage battery over the operation period is used as the evaluation metric, with the cycle number \(R_{cyc}\) representing the effective cycles at different depths of discharge \(D_{DOD}\).
3. Simulation Results and Analysis
A model is built on Matlab/Simulink using actual data from a power plant on August 20, 2023. The thermal unit rated capacity is 330 MW, and the energy storage battery system capacity is 10 MW / 5 MWh. The grid nominal frequency is 50 Hz.
Figure above illustrates the active power output of the energy storage battery under three strategies. The green line is the energy storage battery output, the red line is the planned output (AGC command minus unit output), and the blue line is the actual unit output. The fixed \(K\) method closely tracks the planned output but may cause overcharge/overdischarge. The variable \(K\) method considers SOC and keeps the energy storage battery within safe limits. The adaptive method further smooths the output and avoids operating at rated power, thus extending the energy storage battery life. Long-term simulation over 55,000 s shows that the adaptive strategy maintains good tracking performance and stability.
4. Life Calculation of Energy Storage Battery Using Rain-Flow Counting Method
The rain-flow counting method is applied to the SOC curve obtained from simulation to extract charge/discharge cycles. The SOC maintenance index is defined as:
\[
Q_{SOC} = \frac{1}{n} \sum_{i=1}^{n} (Q_i – Q_{ref})^2
\]
where \(Q_{ref} = 0.55\) (midpoint of allowed range [0.23, 0.8]). The adaptive strategy yields a 5.91%–7.78% improvement in SOC maintenance over the variable \(K\) method, indicating longer operation in the optimal SOC region.
The rain-flow counting method generates the cycle number histogram for each strategy. A summary of the maximum cycle numbers and average service life is given in Table 2.
| Strategy | Maximum Cycle Number | Average Service Life (years) | Life Improvement vs Current Plant Strategy (%) |
|---|---|---|---|
| Fixed \(K\) | 6,422 (but SOC below 0.23) | 8.93 | -9.15 |
| Variable \(K\) | 5,556 | 10.08 | +2.54 |
| Adaptive (proposed) | 8,783 | 12.34 | +25.53 |
The fixed \(K\) method causes the energy storage battery to frequently reach the lower SOC limit (0.22), which is not allowed in practice, thus it is considered invalid. The adaptive strategy achieves the highest maximum cycle number and the longest average life (12.34 years), which is 25.53% higher than the current plant operation strategy (9.83 years), 38.19% higher than the fixed \(K\) method (8.93 years), and 22.42% higher than the variable \(K\) method (10.08 years). These results confirm the superiority of the adaptive comprehensive frequency regulation strategy in preserving the energy storage battery life.
5. Conclusion
An adaptive comprehensive frequency regulation strategy based on SOC feedback is proposed for the energy storage battery in power plants. By dynamically allocating the output between virtual inertia and virtual droop control via an adaptive factor combining fuzzy logic input and SOC-based feedback, the strategy effectively balances frequency regulation performance and energy storage battery health. Simulations using actual plant data demonstrate that the adaptive strategy improves SOC maintenance by 5.91%–7.78% over the variable \(K\) method and extends the energy storage battery life by 25.53% compared with the current plant strategy. The rain-flow counting method provides an accurate life prediction tool. The proposed approach is promising for enhancing both grid stability and economic benefits of the energy storage battery system.