In recent years, the global energy landscape has undergone a transformative shift, driven by the rapid integration of renewable energy sources such as wind and solar power. As the penetration of clean energy continues to rise, the inherent intermittency and variability of these sources pose significant challenges to grid stability and reliability. In this context, energy storage technologies have emerged as a critical solution to mitigate these issues, enabling better grid management and facilitating the efficient utilization of renewable energy. Among various storage options, electrochemical energy storage, particularly battery-based systems, has gained prominence due to its scalability, flexibility, and rapid response capabilities. In this study, we investigate a novel frequency control strategy for thermal power units that leverages the state of charge (SOC) of a cell energy storage system to enhance grid ancillary services, specifically automatic generation control (AGC). Our approach focuses on dynamically adjusting the control strategies of thermal power units based on real-time SOC levels, ensuring that the cell energy storage system remains within an optimal operating range (typically 45% to 55%) to maximize its participation in frequency regulation, reduce battery degradation, and improve overall economic benefits for power plants.
The increasing deployment of cell energy storage systems in power grids is supported by policy initiatives and technological advancements. For instance, regulatory bodies in many countries have issued guidelines promoting the integration of energy storage to support grid stability and renewable energy absorption. Statistical data indicate a substantial growth in the installed capacity of electrochemical energy storage projects, with quarterly additions reaching gigawatt-scale levels. However, despite this growth, existing implementations often face limitations. Many cell energy storage systems are supplied with proprietary control systems that lack adaptability and fail to consider the synergistic operation with existing thermal power units. These systems typically operate in a relatively independent manner, with minimal coordination between the storage side and the generation side. This disjointed approach can lead to suboptimal performance, as the cell energy storage system may frequently reach SOC limits (either too high or too low), forcing it to withdraw from AGC responses and thereby diminishing its value in frequency regulation. Moreover, inadequate control strategies can accelerate battery wear and tear, increasing maintenance costs and safety risks, such as thermal runaway events that have been reported in some large-scale installations. Therefore, there is a pressing need for advanced control methodologies that holistically integrate the cell energy storage system with thermal power units to achieve seamless and efficient joint operation.
To address these challenges, we propose a coordinated control framework that prioritizes the SOC management of the cell energy storage system. The core idea is to use the SOC as a key input to modulate the AGC commands sent to thermal power units, ensuring that the cell energy storage system is always prepared to respond to grid frequency deviations. By maintaining the SOC around a mid-range value, we can leverage the fast response characteristics of the cell energy storage system while avoiding deep charge or discharge cycles that degrade battery life. This strategy involves real-time adjustments to the AGC指令偏置 (command bias) based on SOC status, unit load conditions, and system pressures. In the following sections, we delve into the theoretical foundations, mathematical formulations, and practical implementations of this approach, supported by tables and equations to elucidate the control logic and performance outcomes.
The dynamic response of a cell energy storage system to AGC commands is fundamental to understanding its role in frequency control. When integrated into a power plant, the cell energy storage system works in tandem with thermal units to meet grid dispatch requirements. The remote terminal unit (RTU) aggregates the output signals from both the thermal unit and the cell energy storage system, transmitting the combined response to the grid for performance evaluation. The cell energy storage system excels in providing rapid active power compensation during initial frequency fluctuations, thereby reducing the time required to stabilize the system. Under AGC mode, the RTU sends control commands simultaneously to the thermal unit and the cell energy storage system controller. The latter computes its output based on the difference between the target load and the actual load, feeding this information back to the RTU until the combined output aligns with the AGC指令. However, conventional strategies often lead to scenarios where the cell energy storage system must abstain from responding due to SOC constraints. For example, if the SOC is too low during an AGC increase command, the cell energy storage system cannot discharge, forcing the thermal unit to respond alone. Conversely, if the SOC is too high during an AGC decrease command, the cell energy storage system cannot charge, again relegating response solely to the thermal unit. This intermittent participation undermines the economic and technical benefits of the cell energy storage system, highlighting the necessity for a more intelligent control paradigm.
Our proposed control strategy is built on a detailed analysis of the joint operation dynamics between thermal power units and the cell energy storage system. We formulate the problem as an optimization task aimed at keeping the SOC within a desired range while ensuring adequate response to every AGC command. The strategy involves continuous monitoring of SOC and unit parameters, followed by calculated adjustments to the AGC command bias. To formalize this, let us define key variables and equations. Let $SOC(t)$ represent the state of charge of the cell energy storage system at time $t$, expressed as a percentage of its total capacity. The optimal SOC range is defined as $[SOC_{min}, SOC_{max}]$, where typically $SOC_{min} = 45\%$ and $SOC_{max} = 55\%$. The AGC command at time $t$ is denoted as $P_{AGC}(t)$, and the actual output of the thermal unit is $P_{th}(t)$. The output of the cell energy storage system is $P_{ess}(t)$, with positive values indicating discharge and negative values indicating charge. The total plant output is $P_{total}(t) = P_{th}(t) + P_{ess}(t)$, which must track $P_{AGC}(t)$ as closely as possible.
The control logic involves conditional statements based on SOC and unit load. We introduce a bias term $\Delta P(t)$ that modifies the effective AGC command to the thermal unit, such that the adjusted command becomes $P_{AGC,adj}(t) = P_{AGC}(t) + \Delta P(t)$. The bias $\Delta P(t)$ is determined as follows. First, we check the direction of the AGC command change: if $P_{AGC}(t) > P_{AGC}(t-1)$, it is an increase; otherwise, it is a decrease. For an increase command, if $SOC(t) < X$, where $X$ is a lower threshold (e.g., 10% of rated capacity), then the cell energy storage system cannot discharge, so $P_{ess}(t) = 0$. Additionally, we assess the thermal unit’s load: if $P_{th}(t) < 0.9 \times P_{e}$, where $P_{e}$ is the unit’s rated capacity, then the unit has spare capacity to charge the cell energy storage system later. In this case, we apply a positive bias $\Delta P(t) > 0$ based on factors like main steam pressure and current load. Conversely, if $P_{th}(t) \geq 0.9 \times P_{e}$, no bias is applied. When bias is applied and the unit reaches the original AGC command level, the continued increase due to bias allows the cell energy storage system to charge at a power level up to $\Delta P(t)$. This charging continues until $SOC(t)$ reaches the mid-range value $Y$ (e.g., 50%), at which point the bias is gradually reduced to zero, bringing the unit back to the original command and concluding the response cycle.
Similarly, for a decrease command, if $SOC(t) > 1 – X$, the cell energy storage system cannot charge, so $P_{ess}(t) = 0$. We then evaluate the unit load: if $P_{th}(t) > Z$, where $Z$ is a predefined load level above the minimum stable operating point, a negative bias $\Delta P(t) < 0$ is applied to lower the target command, enabling the cell energy storage system to discharge as the unit reduces load further. Upon reaching the mid-range SOC, the bias is phased out. This cyclic process ensures that the cell energy storage system remains engaged in frequency regulation without hitting SOC limits. The overall control flowchart can be summarized in a table for clarity, as shown below.
| AGC Command Change | SOC Condition | Unit Load Condition | Action on Cell Energy Storage System | Bias Adjustment $\Delta P(t)$ |
|---|---|---|---|---|
| Increase | $SOC < X$ | $P_{th} < 0.9P_e$ | No discharge; thermal unit charges ESS later | Positive, based on real-time parameters |
| Increase | $SOC < X$ | $P_{th} \geq 0.9P_e$ | No discharge; thermal unit responds alone | Zero |
| Increase | $X \leq SOC \leq 1-X$ | Any | ESS discharges per normal strategy | Zero or minimal adjustment |
| Decrease | $SOC > 1-X$ | $P_{th} > Z$ | No charge; thermal unit discharges ESS later | Negative, based on real-time parameters |
| Decrease | $SOC > 1-X$ | $P_{th} \leq Z$ | No charge; thermal unit responds alone | Zero |
| Decrease | $X \leq SOC \leq 1-X$ | Any | ESS charges per normal strategy | Zero or minimal adjustment |
To further quantify the control actions, we can derive mathematical expressions for the bias term. For instance, when a positive bias is warranted during an increase command, we might set $\Delta P(t) = \alpha \cdot (SOC_{target} – SOC(t)) \cdot P_{ess,max}$, where $\alpha$ is a gain factor, $SOC_{target} = 0.5$ (50%), and $P_{ess,max}$ is the maximum charge/discharge power of the cell energy storage system. This proportional control helps smoothly drive the SOC toward the desired range. Similarly, for a negative bias, $\Delta P(t) = -\beta \cdot (SOC(t) – SOC_{target}) \cdot P_{ess,max}$, with $\beta$ as another gain. These formulas ensure that the adjustments are responsive to the degree of SOC deviation, promoting stability in the joint system. Moreover, we can incorporate constraints to prevent excessive maneuvering of the thermal unit, such as rate limits on load changes: $|dP_{th}/dt| \leq R_{max}$, where $R_{max}$ is the maximum ramp rate. This safeguards unit equipment while allowing effective coordination with the cell energy storage system.
The implementation of this strategy requires robust monitoring and communication infrastructure. The cell energy storage system must be equipped with accurate SOC estimation algorithms, often based on coulomb counting or advanced observers that account for temperature and aging effects. Data from the thermal unit, including turbine valve positions, boiler pressures, and fuel rates, need to be integrated into the control loop to assess available margins. In practice, we deployed this approach at a coal-fired power plant in a regional grid, which operates two 300 MW units paired with a 10 MW/5.6 MWh cell energy storage system. Prior to implementation, the plant struggled with poor frequency regulation performance, reflected in low composite adjustment indicators (k-values) around 1.0, which hampered its competitiveness in ancillary service markets. After adopting our SOC-based control strategy, we observed marked improvements. The key performance metrics, as defined by the grid operator, include regulation speed ($k_1$), response time ($k_2$), and regulation accuracy ($k_3$), with the overall composite score computed as $k = 0.25 \times (2k_1 + k_2 + k_3)$. The maximum possible score is 2 for k and 3 for $k_1$ in this market. Over a 12-hour period, the average values significantly increased, as detailed in the table below.
| Indicator | Pre-Implementation Average | Post-Implementation Average | Maximum Achieved Post-Implementation | Target Maximum |
|---|---|---|---|---|
| Regulation Speed ($k_1$) | ~1.5 (estimated) | 2.71 | 2.96 | 3.0 |
| Response Time ($k_2$) | ~0.8 (estimated) | 0.95 | 0.96 | 1.0 |
| Regulation Accuracy ($k_3$) | ~0.7 (estimated) | 0.83 | 0.90 | 1.0 |
| Composite Score ($k$) | ~1.0 | 1.80 | 1.89 | 2.0 |
The data clearly demonstrates that our strategy enhances all aspects of frequency regulation. The regulation speed $k_1$ nearly reached the perfect score, indicating that the cell energy storage system provided rapid power injections and withdrawals as intended. The composite k-value rose to an average of 1.8, peaking at 1.89, which translates to higher revenues in performance-based compensation schemes. Economically, the plant reported daily earnings from frequency regulation services averaging over 180,000 currency units, with peak days exceeding 240,000 units. This financial uplift stems from both improved performance scores and reduced penalties for non-compliance. Furthermore, the cell energy storage system experienced fewer extreme SOC excursions, which mitigated battery degradation. By avoiding deep cycles and maintaining moderate temperatures, the risk of thermal events was lowered, contributing to safer operation. The physical setup of the cell energy storage system, often housed in densely packed containers, underscores the importance of prudent management to prevent failures. The integration of our control logic helped maintain operational integrity, as evidenced by fewer temperature alarms and extended intervals between maintenance cycles.
Beyond immediate performance gains, our strategy offers long-term benefits for the lifecycle of the cell energy storage system. Battery aging is influenced by factors like cycle depth, charge/discharge rates, and operating temperature. We can model the degradation using semi-empirical equations, such as the loss of capacity over time: $C(t) = C_0 \cdot e^{-\gamma \cdot N_{eq}}$, where $C_0$ is initial capacity, $\gamma$ is a degradation coefficient, and $N_{eq}$ is the equivalent number of full cycles. By keeping SOC in the mid-range, we reduce the effective cycle depth, thereby slowing capacity fade. For instance, if the cell energy storage system operates between 45% and 55% SOC, the depth of discharge (DOD) is only 10%, compared to 80% or more in aggressive cycling scenarios. This can exponentially extend service life, as degradation often scales nonlinearly with DOD. We can express this relationship as $\gamma = \gamma_0 \cdot (DOD)^{\delta}$, where $\gamma_0$ and $\delta$ are material-specific constants. Thus, our control strategy not only optimizes real-time grid support but also enhances the economic viability of the cell energy storage system through longevity improvements.
To generalize the approach, we can frame it within a broader optimization framework. Consider the objective function $J = \int_{0}^{T} \left[ w_1 \cdot (P_{total}(t) – P_{AGC}(t))^2 + w_2 \cdot (SOC(t) – SOC_{target})^2 + w_3 \cdot P_{ess}(t)^2 \right] dt$, where $w_1$, $w_2$, and $w_3$ are weighting factors that balance tracking error, SOC deviation, and energy storage usage. The first term ensures accurate AGC response, the second term penalizes SOC deviations from the target, and the third term discourages excessive power flows that could stress the cell energy storage system. Solving this optimization in real-time requires model predictive control (MPC) techniques, but our rule-based strategy approximates this by dynamically adjusting biases based on heuristics. Future work could integrate MPC to further refine performance, especially in grids with high renewable penetration where AGC commands become more volatile.
In conclusion, our research presents a practical and effective method for coordinating thermal power units with cell energy storage systems in frequency regulation applications. By continuously monitoring the state of charge and adapting control signals accordingly, we ensure that the cell energy storage system remains an active participant in AGC responses, thereby harnessing its fast response capabilities while preserving battery health. The implementation results confirm substantial improvements in regulatory performance metrics and economic returns, validating the strategy’s efficacy. As power systems evolve towards higher shares of renewables, the role of cell energy storage systems will only grow, necessitating advanced control paradigms like the one we propose. We believe that this approach can be extended to other storage technologies and generation types, contributing to a more resilient and efficient grid infrastructure. Ongoing efforts will focus on refining the algorithms through machine learning and expanding the strategy to multi-storage systems for larger-scale applications.
Throughout this study, the term cell energy storage system has been emphasized to highlight the specific technology underpinning our work. The repeated mention of cell energy storage system underscores its centrality in modern grid stabilization efforts. As deployment scales up, innovations in control strategies will be crucial to unlocking the full potential of cell energy storage system installations, making them not just ancillary assets but core components of the energy transition. We hope that our contributions inspire further research and development in this vital area, paving the way for sustainable and reliable power systems worldwide.