We present an integrated solution combining optimized energy storage battery design and coordinated system voltage control to address the issues of energy waste and grid voltage fluctuations caused by the intermittency and uncertainty of photovoltaic power generation. By establishing a hybrid AC/DC energy storage architecture, we clarify the functional division where the DC-side energy storage battery smooths power fluctuations while the AC-side energy storage battery undertakes grid frequency regulation. A quantitative model is developed to determine the power and capacity configuration of the energy storage battery systems. At the control level, a global voltage control model based on photovoltaic-energy storage battery collaboration is constructed: an improved extremum seeking method is adopted on the photovoltaic side to achieve rapid and accurate maximum power point tracking under dynamic conditions, while a nonlinear voltage-current dual-loop controller based on differential flatness theory is designed on the energy storage battery side to achieve precise regulation of DC bus energy. Experimental results demonstrate that our proposed scheme not only reduces the system frequency regulation actions to nearly zero, decreases the voltage fluctuation range to within ±2%, and shortens the recovery time to 0.5 s, but also improves the overall system energy efficiency to 93.8%, effectively balancing operational stability, energy utilization efficiency, and grid ancillary service capability.
Natural conditions significantly impact photovoltaic power generation, primarily through intermittency and uncertainty, which lead to energy waste and grid instability. As power systems impose increasingly stringent requirements on photovoltaic power stations for ancillary services, these stations must possess adaptive frequency regulation capabilities. The deployment of energy storage battery systems offers a solution, but it requires rational design of the configuration of the energy storage battery and effective control of system voltage to ensure stable station operation. Existing research on energy storage battery and voltage control for photovoltaic stations mainly focuses on two dimensions: optimal configuration of energy storage battery capacity and coordinated control strategies. However, there remain gaps in addressing the coordinated energy management between AC and DC sides and global voltage stability. In terms of energy storage battery configuration, most studies emphasize single-objective optimization. For example, some researchers constructed a photovoltaic-hydrogen energy storage battery two-layer optimization model to address peak shaving and fluctuation smoothing, but their focus was on economic configuration without exploring active voltage control strategies. Others proposed a method based on operating mode identification for energy storage battery operation and capacity configuration in high-capacity-ratio photovoltaic stations, optimizing the whole lifecycle net present value. Yet, they did not fully consider the dynamic impact of energy storage battery actions on the grid connection point voltage, lacking targeted suppression of voltage fluctuations. In control strategies, existing methods attempt to use the energy storage battery or photovoltaic inverter for voltage regulation, but coordination mechanisms need improvement. Some researchers designed a voltage regulation strategy based on cluster partitioning of distributed energy storage battery systems, effectively solving voltage over-limit problems caused by high-penetration photovoltaic access. However, this method mainly relies on local regulation of distributed energy storage battery systems, failing to fully exploit the synergistic potential of centralized energy storage battery systems, photovoltaic inverters, and other control resources within the station. Other researchers further used consensus algorithms to coordinate photovoltaic inverter reactive power and energy storage battery active power, achieving grouped coordinated voltage regulation. Nevertheless, this strategy heavily depends on real-time communication system reliability and does not analyze robustness against communication delays or data packet losses, which are common engineering faults. Moreover, its control structure is still confined to the AC side. In summary, existing studies share the following common limitations: most separate the configuration of the energy storage battery from voltage control, failing to integrate the design of the energy storage battery with dynamic voltage stability performance; configuration of the energy storage battery is mostly concentrated on the AC side or DC side, lacking systematic design of hybrid AC/DC energy storage battery structures in terms of functional complementarity, lifespan extension, and economic optimization; at the control level, there is a lack of a global voltage control strategy that achieves fast, accurate, and robust coordination among the photovoltaic side, the energy storage battery side, the DC bus, and the AC grid.
To overcome these problems, we propose an integrated method for photovoltaic station energy storage battery design and system voltage control. The core innovations are twofold. First, we propose a collaborative design architecture for hybrid AC/DC energy storage battery systems. This architecture clarifies the functional division and cooperation mechanism between the DC-side energy storage battery and the AC-side energy storage battery: the DC-side energy storage battery primarily focuses on smoothing photovoltaic output fluctuations and absorbing surplus energy, reducing power disturbances at the source. The AC-side energy storage battery undertakes system frequency regulation tasks, meeting relevant policy requirements, and by prioritizing response to grid commands, reduces the number of operations of the DC-side energy storage battery. Through this division of labor, the overall system operational flexibility and ancillary service capability are enhanced while considering the economic and reliability aspects of the whole lifecycle. Second, we construct a global voltage control model based on photovoltaic-energy storage battery collaboration. On the photovoltaic side, an improved extremum seeking method is used for maximum power point tracking. By introducing an adaptive delay and error feedback mechanism, tracking accuracy under sudden irradiance changes and current balance capability of the converter are improved, ensuring stable input from the photovoltaic side. On the energy storage battery side, a nonlinear control method based on differential flatness theory is introduced to design a voltage-current dual-loop control strategy, achieving rapid and precise regulation of DC bus energy, thereby effectively suppressing system voltage fluctuations caused by photovoltaic fluctuations and load changes.
Photovoltaic Energy Storage Station Topology
The topology of the photovoltaic energy storage station is illustrated conceptually. Photovoltaic arrays generate electricity, which is collected by combiner boxes, processed by DC lightning protection distribution units, and then fed into inverters to convert DC to AC. The output from inverters, after AC lightning protection and metering, is partially sent to the public grid via a 0.4 kV/20 kV step-down transformer and partially supplied to local loads. Simultaneously, the energy storage battery system is connected to the AC side through inverters, forming a hybrid AC/DC structure. This topology supports bidirectional energy flow, providing the foundation for energy storage battery design and voltage control.
Energy Storage Battery Design for Photovoltaic Stations
Based on the energy distribution characteristics of the above topology, to meet the requirements of energy absorption (reducing surplus energy loss) and system frequency regulation, and to comply with the minimum local requirements for energy storage battery configuration at photovoltaic stations, we design the configuration of the hybrid AC/DC energy storage battery. The following sections clarify the power and capacity configuration logic for the DC-side and AC-side energy storage battery systems.
In the hybrid AC/DC energy storage battery system design, the selection of the energy storage battery type directly affects system performance and economy. The DC side is mainly responsible for smoothing photovoltaic fluctuations and energy buffering, and we select lithium-ion batteries, whose high energy density and high efficiency (>95%) facilitate optimal energy management, though attention must be paid to cycle life and temperature sensitivity. The AC side mainly handles power-type tasks such as grid frequency regulation, where we select lead-carbon batteries, which are low-cost, have long cycle life, and good power characteristics, suitable for frequent charging and discharging.
DC-Side Energy Storage Battery Design
The power design of the DC-side energy storage battery must satisfy the difference between the maximum power of the photovoltaic array and the maximum allowed power on the AC side. For the j-th photovoltaic generation unit, let the AC-side power be \(P_{aj}\) and the DC-side generation power be \(P_{sj}\). Let the capacity ratio and power loss be \(s_j\) and \(\beta_j\), respectively. To prevent loss from DC-side over-configuration, the DC-side energy storage battery power \(P_{Zj}\) must satisfy:
$$P_{Zj} \ge P_{sj} – P_{aj}, \quad P_{sj} = s_j P_{aj} – s_j P_{aj} \beta_j$$
Based on the energy conservation principle, this ensures that the DC-side energy storage battery can absorb the energy that cannot be fully consumed by the AC side. Considering that the inverter and box-type transformer have 1.1 times overload capacity, the power of the energy storage battery controller should be no less than the difference between the maximum DC-side power and the maximum allowed AC-side power. Selecting the day with optimal generation efficiency, we calculate the surplus energy during periods when \(P_{sj} > P_{aj}\). The capacity \(F_{Zj}\) of the DC-side energy storage battery must satisfy:
$$F_{Zj} \ge \sum_{t=1}^{N} (P_{sj} – 1.1 P_{aj})_t$$
where \(t\) is time (h) and \(N\) is the total number of steps. The total DC-side energy storage battery power \(P_Z\) (kW) and total capacity \(F_Z\) (kWh) are the accumulations over m units:
$$P_Z = \sum_{j=1}^{m} P_{Zj}, \quad F_Z = \sum_{j=1}^{m} F_{Zj}$$
AC-Side Energy Storage Battery Design
The configuration of the AC-side energy storage battery must first meet the minimum local requirements for energy storage configuration at the photovoltaic station. Based on the total installed capacity of the station, the minimum required energy storage battery power \(P_B\) and capacity \(F_B\) are determined. The AC-side energy storage battery power \(P_{AB}\) and capacity \(F_{AB}\) must satisfy:
$$P_{AB} \ge P_B – P_Z, \quad F_{AB} \ge F_B – F_Z$$
If the system is required to participate in primary frequency regulation, let the maximum frequency regulation power required by the power system be \(P_g\) and capacity be \(F_g\). Then:
$$P_{AB} \ge \max\{P_g – P_Z,\ 0.5P_g,\ P_B – P_Z\}, \quad F_{AB} \ge \max\{F_g – F_Z,\ 0.5F_g,\ F_B – F_Z\}$$
Prioritizing the AC-side for primary frequency regulation reduces the number of charging/discharging cycles on the DC side, which helps extend its cycle life, in line with whole-lifecycle economic considerations.
System Voltage Control Based on Photovoltaic-Energy Storage Battery Collaboration
Reasonable design of the energy storage battery provides hardware foundation for stable operation, while precise control of system voltage is key to ensuring power quality and preventing faults caused by photovoltaic output fluctuations or load changes. We construct a system voltage control model based on photovoltaic-energy storage battery collaboration, designing voltage control strategies for both the photovoltaic side and the energy storage battery side that are adapted to the characteristics of the photovoltaic-energy storage battery system.
Photovoltaic Side Converter Control: Improved Extremum Seeking Method
In the photovoltaic energy storage station, the control of the photovoltaic three-phase interleaved boost converter directly affects photovoltaic energy collection and system voltage foundation. To maximize solar energy utilization and improve photoelectric conversion efficiency, we adopt an improved extremum seeking method for maximum power point tracking (MPPT). Traditional extremum seeking methods determine the voltage adjustment direction by calculating the sign of the derivative of photovoltaic power, combined with the previous voltage change \(dv_q/dt\), where \(v_q\) is the photovoltaic cell output voltage. However, under drastic irradiance changes, the power-voltage curve shifts rapidly, leading to misjudgment of the power extremum point, causing oscillation of the tracking point and resulting in photovoltaic side power fluctuations. Additionally, in a three-phase interleaved parallel converter structure, the traditional method lacks active control of current balance among phases, which may cause uneven current distribution and induce local overcurrent, threatening stable operation. To solve these problems, we make two key improvements to the traditional extremum seeking method:
1) Introduction of an adaptive delay: When a drastic irradiance change is detected (i.e., the power change rate exceeds a set threshold of 30 kW/s), the algorithm automatically inserts an adaptive waiting period, temporarily suspending the voltage adjustment command. This improvement allows the system to wait for the power-voltage curve to stabilize under new irradiance conditions, avoiding erroneous searching during dynamic curve shifts, significantly improving the tracking accuracy and anti-interference capability of MPPT in dynamic environments, and fundamentally reducing power oscillations caused by misjudgment, enhancing the stability of the photovoltaic side input. The adaptive delay time \(T_w\) is inversely proportional to the disturbance intensity:
$$T_w = K / |dP/dt|$$
where \(K\) is the gain coefficient in the range of 0.1 to 10. This design ensures that the stronger the disturbance, the shorter the waiting time, achieving a balance between fast response under dynamic conditions and accurate tracking in steady state.
2) Introduction of error signals: Applied to the three-phase interleaved boost converter, the voltage signal \(v_{q\max}\) output by the MPPT control is superimposed with error signals. Define the phase current error \(\gamma_n\) as the difference between the phase current reference \(i_{qs}\) and the actual value \(i_{qn}\) (\(n=1,2,3\)):
$$\gamma_1 = i_{qs} – i_{q1},\quad \gamma_2 = i_{qs} – i_{q2},\quad \gamma_3 = i_{qs} – i_{q3}$$
Then, based on the error signals, the voltage signal for each phase control is adjusted to:
$$v_{q1} = v_{q\max} + \gamma_1,\quad v_{q2} = v_{q\max} + \gamma_2,\quad v_{q3} = v_{q\max} + \gamma_3$$
This design automatically reduces the voltage command for the phase with larger current, thus decreasing its output current and achieving automatic current balancing among the three phases. This optimizes losses in magnetic cores and switching devices and effectively prevents protection actions or device damage caused by single-phase overcurrent, greatly improving the operational reliability of the converter and overall system stability. The improved extremum seeking method, through adaptive delay and error feedback, balances dynamic tracking accuracy and static current balance, providing more stable and high-quality photovoltaic side power input for subsequent voltage control.
Energy Storage Battery Side Converter Control: Differential Flatness Control Method
Uncertainty in photovoltaic output can easily cause voltage fluctuations. The energy storage battery three-phase interleaved bidirectional converter, with its fast charging/discharging characteristics, adopts a voltage-current dual-loop control strategy to ensure system voltage stability, where the outer loop is a differential flatness control method. Traditional voltage control methods, such as PI control based on small-signal linearization, have inherent limitations when dealing with complex conditions like photovoltaic stations that exhibit significant nonlinearity and large disturbances. On one hand, the controller parameters of these methods are typically designed around a specific operating point. When the system operating point deviates significantly due to power transients, control performance degrades markedly, leading to slower response or larger overshoot. On the other hand, these methods have limited ability to handle system nonlinearities, making it difficult to achieve consistent high dynamic performance over the entire operating range, especially in photovoltaic stations that must cope with rapidly changing irradiance and load demands. In contrast, the differential flatness control method we adopt is a nonlinear exact feedback linearization method based on differential geometry. Its core advantage is that by carefully selecting an appropriate flat output, the original nonlinear system can be completely equivalently transformed into a linear controllable system. This transformation is not only accurate in theory but also significant in practice. Based on this transformed linear system, we can directly design a control law with global stability and desired dynamic response (e.g., exponential convergence). This fundamentally overcomes the shortcomings of traditional linear control methods that rely on local operating points and lack robustness, making it particularly suitable for applications like photovoltaic-energy storage battery systems that need to maintain stable operation over a wide range of power fluctuations.
System Characteristic Analysis
Differential flatness is a control theory framework applicable to a class of nonlinear systems. The core idea is: for a nonlinear system, if a specific output (called the “flat output”) can be found such that all system state variables and control inputs can be expressed as algebraic functions of this output and its finite-order derivatives, then the system is said to be differentially flat. This property allows us to map the complex nonlinear control problem into a completely equivalent linear controllable system space through this output, enabling the application of all mature linear system theories to design control laws with global stability and desired dynamic performance. For the system \(\dot{y} = k(y, l)\), where \(k\) is a nonlinear smooth vector field (representing the influence of the energy storage battery charging/discharging characteristics on the system state), if there exists an output \(x = \epsilon(y, l, \dot{l}, \ldots, l^{(n)})\) such that the system state \(y\) and input \(l\) (the power command of the energy storage battery converter) can be expressed by \(x\) and its finite-order derivatives as \(y = \mu(x, \dot{x}, \ldots, x^{(n)})\), then the system is a differential flat system and can be equivalently transformed into a linear controllable system. Here, \(\eta\) denotes the derivative order.
Differential Flatness Modeling
For the energy storage battery converter, we choose the DC bus capacitor energy \(x = F_s\) as the flat output, where \(F_s\) is the stored energy of the DC bus capacitor \(E_B\). Define the energy storage battery side converter control input \(l = p_E\) (power interaction between the energy storage battery converter and the DC bus, negative for charging, positive for discharging) and the state variable \(y = l_E\) (a power command related to the energy storage battery energy and power interaction). According to the law of power conservation, the rate of change of DC bus capacitor energy equals the sum of all power flowing into that node, i.e., photovoltaic output power minus load power plus the charging/discharging power of the energy storage battery system. By selecting the capacitor energy as the flat output, the state variables can be directly expressed through the physical definition of capacitor energy. The control input, i.e., the power command of the energy storage battery converter, can be expressed as a function of the first derivative of the flat output and known external disturbances through the above power balance relation. This key step proves that all key variables of the system can be algebraically expressed by the selected flat output and its derivatives, thus satisfying the complete definition of differential flatness. Combining the above derivations, we obtain the reversible dynamic equations for the energy storage battery side converter control (negative for charging, positive for discharging):
$$y = \frac{2x}{\sqrt{E_B}}, \quad l = \pm 2 p_{E\max} \left(1 – \sqrt{1 – \frac{(\dot{x} + p_d – p_0)}{p_{E\max}}}\right)$$
where \(p_{E\max}\) is the maximum interactive power of the energy storage battery converter, \(p_d\) and \(p_0\) correspond to load power and photovoltaic output power, respectively, which are external disturbances affecting DC bus energy changes, and \(\dot{x}\) is the first derivative of \(x\). Since state and control variables can be expressed by \(x\) and its derivatives, the energy storage battery converter system is a differential flat system.
Voltage Control Law Design
When photovoltaic output or load changes abruptly, the DC bus energy of the power system changes, causing voltage fluctuations. To make the bus energy track a reference trajectory, we design the control law based on the above differential flatness model. The core advantage of differential flatness theory is that once the system is proven to be flat, a controller with desired dynamics can be designed for the flat output. We design a reference trajectory \(x_s\) for the flat output and directly construct its derivative relationships, thereby achieving exact linearization control of the nonlinear system. The DC bus capacitor energy is \(E_c\), which is the steady-state target of the control system, directly corresponding to the desired rated DC bus voltage \(U_d\). It is calculated by the capacitor’s energy storage characteristic: \(E_c = \frac{1}{2} C U_d^2\), where \(C\) is the total DC bus capacitance. This establishes the energy reference for voltage control. Based on this, we design the voltage feedback control law:
$$\ddot{x} – \ddot{x}_s + \kappa_1(\dot{x} – \dot{x}_s) + \kappa_2(x – x_s) = 0$$
where \(x_s\) is the DC bus capacitor energy reference value, and \(\kappa_1, \kappa_2\) are feedback control law parameters. \(\kappa_1\) dominates the system dynamic response speed and damping characteristics, while \(\kappa_2\) is used to compensate steady-state errors. The integral term compensates measurement/modeling deviations, ensuring zero steady-state error. The tracking error \(e = x – x_s\) satisfies \(\ddot{e} + \kappa_1 \dot{e} + \kappa_2 e = 0\). Combined with the desired characteristic polynomial, we obtain the optimal voltage control parameters: \(\kappa_1 = 2\zeta \omega_m\), \(\kappa_2 = \omega_m^2\), where \(\zeta\) is the damping coefficient (typically between 0.7 and 1.0, with \(\zeta=0.7\) giving moderate overshoot and shorter settling time, and \(\zeta=1.0\) giving critical damping with no overshoot), and \(\omega_m\) is the natural frequency (rad/s).
$$P(r) = r^2 + 2\zeta \omega_m r + \omega_m^2$$
The outer loop voltage control output serves as the reference for the inner current loop, which uses a linear PI controller to make the energy storage battery current \(i_E = p_E / v_E\) (where \(v_E\) is the energy storage battery voltage), achieving precise voltage regulation.
Stability Verification
We use Lyapunov’s direct method to rigorously prove global asymptotic stability of the closed-loop system. Consider the Lyapunov function candidate:
$$V(e, \dot{e}) = \frac{1}{2} \dot{e}^2 + \frac{1}{2} \kappa_2 e^2$$
This function \(V\) is positive definite. Taking its time derivative and substituting the closed-loop error dynamics \(\ddot{e} = -\kappa_1 \dot{e} – \kappa_2 e\) yields:
$$\dot{V} = \dot{e} \ddot{e} + \kappa_2 e \dot{e} = \dot{e}(-\kappa_1 \dot{e} – \kappa_2 e) + \kappa_2 e \dot{e} = -\kappa_1 \dot{e}^2$$
Since \(\dot{V}\) is negative semidefinite, according to Lyapunov’s stability theorem, the system is stable at the equilibrium \((e, \dot{e}) = (0,0)\). To further prove asymptotic stability, we apply LaSalle’s invariance principle. In the set \(\{(e, \dot{e}) \mid \dot{V}=0\} = \{(e, \dot{e}) \mid \dot{e}=0\}\), substituting into the system dynamics gives \(\ddot{e} = -\kappa_2 e\). For the system trajectory to remain in this set, we require \(\ddot{e} \equiv 0\), which further requires \(e \equiv 0\). Therefore, the largest invariant set is only the origin. By LaSalle’s principle, the closed-loop system is globally asymptotically stable. The stability of the nonlinear system of the energy storage battery converter is determined by controller parameters and control frequency. When \(\zeta > 0\) and \(\omega_m > 0\) (i.e., \(\kappa_1, \kappa_2 > 0\)), the system voltage can achieve stable control.
Coordination Strategy and Protection Mechanism
The coordination strategy of the hybrid AC/DC energy storage battery follows the principle of “functional layering, power mutual assistance.” The system sets the AC-side energy storage battery to respond first to grid frequency regulation commands, while the DC-side energy storage battery is primarily responsible for smoothing photovoltaic fluctuations; when the AC-side power is insufficient, the DC side can provide backup support. The coordination process uses an event-triggered communication mechanism to ensure real-time dynamic response. In extreme fault scenarios, the system automatically isolates the faulty unit and switches to independent control mode to maintain basic operation, then orderly restores the coordinated state after the fault is cleared.
Frequency Regulation Threshold of Hybrid Energy Storage Battery
The frequency regulation process of this hybrid energy storage battery structure has a clear dynamic threshold mechanism, mainly reflected in the power distribution logic and the state of the energy storage battery itself. In power distribution, the AC-side energy storage battery is set as the primary regulation unit; only when its power output reaches a preset upper limit (e.g., 90% of rated power) or its state of charge (SOC) deviates from the safe operating range does the DC-side energy storage battery provide supplementary power support. The SOC, maximum charging/discharging power, and converter capability of each energy storage battery unit constitute the hard thresholds for frequency regulation capability. The control system monitors these states in real time to ensure that frequency regulation commands are always executed within the comprehensive safe operating domain of the hybrid energy storage battery.
Protection Strategy and Recovery Capability under Extreme Conditions
To deal with extreme scenarios such as energy storage battery faults, grid faults, and superposition of multiple disturbances, we design a hierarchical protection and self-recovery strategy. Once a fault in any energy storage battery unit is detected, the system immediately isolates it and activates a power redistribution algorithm, where other energy storage battery units take over its core functions to ensure uninterrupted system operation. In the event of a grid fault, the system switches to island operation mode, using the photovoltaic and hybrid energy storage battery to cooperatively maintain power supply to critical loads. For the superposition of irradiance transients and load shocks, the preceding control law based on differential flatness has good robustness and can respond quickly to maintain voltage stability. After fault clearance, the system follows a preset sequence for orderly recovery: grid-side synchronization and reconnection; the isolated faulty energy storage battery unit, after completing diagnosis and reset, undergoes controller verification before being flexibly reconnected; the system state transitions smoothly to normal operation. The entire recovery process is controlled within seconds to minutes, ensuring system resilience and rapid recovery capability.
Experimental Analysis
We conducted simulation experiments based on the following parameters: photovoltaic arrays using monocrystalline silicon modules, each rated 300 W, open-circuit voltage 40 V, short-circuit current 10 A. The original station included 10 units, each with DC-side generation power of 300 kW and AC-side output of 250 kW, total installed capacity 2.5 MW. For the energy storage battery, the DC side selected lithium-ion batteries, cell voltage 3.2 V, cycle life greater than 2000 cycles; the AC side selected lead-carbon batteries, cell voltage 2 V, cycle life greater than 3000 cycles. In the simulation settings, we used the ode23tb solver with a fixed step size of 0.001 s, simulation time set to 10 s. Using MATLAB/Simulink, we built a high-fidelity digital simulation model, with component parameters and control systems consistent with the theoretical design.
| Design Dimension | Core Objective | Design Result | Remarks |
|---|---|---|---|
| DC side | Prevent DC-side over-configuration loss, absorb surplus energy | Single unit: 25 kW power, 50 kWh capacity; 10 units total: 250 kW, 500 kWh | Adapts to photovoltaic unit generation characteristics, reduces surplus energy loss, provides basic energy buffer for voltage stability |
| AC side | Meet minimum local configuration requirements, support primary frequency regulation | 2.5 MW station: 15% minimum configuration → 375 kW, 750 kWh (2h discharge); if primary frequency regulation required (max 400 kW) → 400 kW, 800 kWh | Prioritizes primary frequency regulation, reduces DC side charge/discharge cycles, extends DC-side energy storage battery life, balances economy and ancillary service capability |
Based on this energy storage battery design, we tested the frequency regulation counts of the photovoltaic energy storage station system under three strategies: single energy storage battery control, single photovoltaic control, and photovoltaic-energy storage battery coordinated control. The results are summarized in the following table.
| Control Strategy | Frequency Regulation Count (times/h) | DC Bus Voltage Fluctuation Amplitude (%) | Voltage Recovery Time (s) | System Steady-State Voltage Deviation (%) |
|---|---|---|---|---|
| Traditional PI control | 15 | ±12 | 2.0 | ±6 |
| Sliding Mode Control (SMC) | 8 | ±8 | 1.2 | ±4 |
| Photovoltaic-energy storage battery coordinated control (our method) | 0.2 | ±2 | 0.5 | ±1 |
The experimental data show that our method significantly outperforms the other two strategies in all metrics. The frequency regulation count is drastically reduced to nearly zero, indicating that the system voltage is effectively stabilized, keeping the frequency very close to the rated value. The voltage fluctuation amplitude is limited to ±2%, much smaller than the ±12% of PI control and ±8% of SMC. The voltage recovery time after a load step change is only 0.5 s, compared to 2.0 s for PI and 1.2 s for SMC. The steady-state voltage deviation is also reduced to ±1%, demonstrating excellent regulation precision. These results strongly validate the superiority of the proposed photovoltaic-energy storage battery coordinated control approach.
To further verify the engineering feasibility, we compared the implementation complexity of different control strategies from three dimensions: state observation complexity, parameter sensitivity and robustness, and computational complexity. The results are presented in the table below.
| Evaluation Dimension | Specific Indicator | Traditional PI Control | OFL Control (with observer) | Our Method (Differential Flatness) |
|---|---|---|---|---|
| State observation complexity | Number of sensors required | 2 (voltage, current) | 2 (voltage, current) + state observer | 2 (voltage, current) |
| State acquisition method | Direct measurement | Depends on complex observation algorithm | Direct calculation | |
| Parameter sensitivity | Overshoot with +20% capacitance error (%) | > 15 | 4.5 | 2.2 |
| Settling time with -20% capacitance error (s) | > 3.0 | about 1.3 | 0.61 | |
| Computational complexity | Single execution time (μs) | 6.2 | 25.8 | 8.5 |
| Code size (kb) | 5.1 | 18.9 | 6.3 |
The results demonstrate that our differential flatness control strategy offers significant advantages in engineering implementation. Regarding state observation, our method requires only conventional sensors, like PI control, without the need for complex observers as in OFL control. Under ±20% parameter mismatch, our method still maintains the best dynamic performance (overshoot < 2.5%, settling time < 0.61 s), significantly surpassing PI control and avoiding the complex tuning of OFL. In terms of computational efficiency, our method has a single execution time of only 8.5 μs and code size of 6.3 kb, far lower than OFL control and comparable to PI control. This fully proves that the differential flatness control strategy possesses both theoretical advancement and engineering convenience, enabling seamless implementation in real systems.
To quantify the improvement in overall energy efficiency brought by the AC/DC coordinated scheme, we defined the system comprehensive efficiency, which covers all energy losses from photovoltaic array output to final delivery to grid or load, including converter efficiency, transformer losses, and energy storage battery charging/discharging cycle efficiency. The experimental results under different operating modes are shown below.
| Operating Mode | Traditional Scheme System Efficiency (%) | Our Coordinated Scheme System Efficiency (%) | Efficiency Improvement (%) |
|---|---|---|---|
| Normal grid-connected generation | 91.5 | 93.8 | 2.3 |
| Participating in primary frequency regulation | 89.2 | 92.5 | 3.3 |
| Absorbing surplus energy | – (energy wasted) | 90.1 (stored and used) | >10 (equivalent) |
The data indicate that our method significantly improves system energy efficiency across all modes due to the synergy between DC-side and AC-side energy storage battery systems. In normal grid-connected mode, the improvement stems from the DC-side energy storage battery smoothing photovoltaic fluctuations locally, reducing extra losses in converters and transformers under non-steady conditions. In frequency regulation mode, the improvement is more evident because the AC-side energy storage battery, with its high power efficiency and long cycle life, efficiently handles frequent charging/discharging actions, avoiding efficiency loss from frequent photovoltaic inverter adjustments. In absorbing surplus energy, the scheme increases the utilization of otherwise wasted energy from nearly zero to 90.1%, which is of great significance for improving the overall economic benefit of the station.
Conclusions
In this paper, we have addressed the issues of energy waste, poor grid stability, and insufficient ancillary service capability caused by the intermittency and uncertainty of photovoltaic power generation. Considering the defects of existing research in energy storage battery configuration and voltage control, we conducted a study on the design of energy storage battery and system voltage control for photovoltaic stations. The main conclusions are as follows:
1) In energy storage battery design: To rationally design the hybrid AC/DC energy storage battery, we used quantitative formulas to determine the power and capacity of the DC-side and AC-side energy storage battery systems. The DC-side energy storage battery is responsible for solving the surplus energy loss caused by capacity ratio differences; the AC-side energy storage battery is responsible for combining policy configuration requirements and participating in primary frequency regulation. Experiments show that our method significantly improves the reliability of the photovoltaic energy storage station system while satisfying energy absorption and frequency regulation needs.
2) In system voltage control: We constructed a system voltage control model based on photovoltaic-energy storage battery collaboration. On the photovoltaic side, an improved extremum seeking method was used for MPPT, balancing tracking accuracy and current distribution under sudden irradiance changes. On the energy storage battery side, a differential flatness control method was introduced to regulate DC bus energy, achieving fast response and stable control of system voltage. Experimental results show that through this collaborative control model, our method effectively suppresses system voltage fluctuations under both load transients and steady-state conditions, with the overall efficiency reaching 93.8% and remarkable improvements in frequency regulation and voltage stability.