Network-formation type wind storage collaborative control method based on energy coordination state and shared virtual rotor

Abstract

The application discloses a kind of based on energy coordination state and sharing virtual rotor network configuration type wind storage collaborative control method, energy SOC, wind turbine rotor kinetic energy and load shedding reserve are unified as energy coordination state by centralization logarithmic potential function;On this basis, strict convex power distribution model is established and constraint analytical solution is obtained.Sharing virtual rotor is responsible for station frequency formation, differential angle state is responsible for internal power redistribution, and relative swing of two independent virtual rotors is avoided from structure.Derivation of zero net power recovery after disturbance can restore SOC and speed under the condition of approximately not changing the total power of station.Theoretical analysis shows that the proposed distribution solution is unique, the energy boundary has positive invariance, the sharing virtual rotor is stable within the feasible power angle range, and the total energy potential is monotonically non-increasing during the recovery phase.Follow-up work should be combined with detailed electromagnetic transient model, hardware-in-the-loop and wind turbine load model to verify parameter robustness, fault current limiting and multi-machine expansion.

Description

Technical Field

[0001] This invention relates to the field of grid-connected control technology for new energy power generation, and in particular to a grid-connected wind and energy storage collaborative control method based on energy co-state and shared virtual rotor. Background Technology

[0002] As the proportion of new energy sources in the power system continues to increase, the voltage and frequency support of the power system is gradually shifting from traditional synchronous generators to power electronic converters. Grid-forming control, by forming a controlled voltage source at the AC end, can provide effective synchronization and voltage support in scenarios such as weak grids, islanding, and black start. For full-power converter wind turbines, grid-forming control also needs to handle multiple time-scale tasks simultaneously, including maximum power point tracking, DC voltage stabilization, rotor kinetic energy release, pitch control, and fault current limiting.

[0003] After configuring grid-based energy storage in wind farms, the energy storage has rapid bidirectional power regulation capabilities, while the wind turbines have rotor kinetic energy and aerodynamic backup, complementing each other in physical properties. However, existing coordination methods typically employ fixed droop coefficients, virtual inertia parameters, or power decomposition strategies based on high and low frequencies. These methods treat the energy storage state of charge (SOC), turbine speed, and reserve power as independent variables, requiring additional threshold judgments and mode switching; when the operating point, wind speed, or energy storage state changes, the preset allocation relationship is prone to mismatch. For example, some existing technologies use multi-segment SOC zoning control strategies, requiring the setting of multiple switching thresholds. When the energy storage state of charge switches between different zones, the control parameters need to change accordingly, easily causing power fluctuations and control discontinuities. Another example is the parameter threshold coordination strategy based on the amplitude of frequency deviation, which determines the magnitude of the disturbance based on whether the frequency deviation exceeds the preset upper and lower limits of the frequency threshold, and then decides whether the energy storage module should operate; this type of method relies on fixed threshold judgments, and the threshold parameters are difficult to adjust adaptively when the system operating point changes, and hard threshold switching itself is prone to causing sudden power changes.

[0004] If wind turbines and energy storage are configured with independent virtual synchronous generators (VSGs), while each retains its grid-connecting capability, it introduces additional internal relative oscillation modes. Active power circulation and control competition may arise between the two virtual rotors, requiring additional virtual impedance and damping mechanisms for suppression, increasing the complexity of controller structure and parameter tuning. Matching control, virtual oscillator control, and two-port grid-connecting control extend grid-connecting control theory from the perspectives of energy coupling, nonlinear oscillators, and AC / DC multi-port coordination, respectively. Meanwhile, maintaining the angle formation capability of the grid-connecting converter during current saturation has become an important research direction. These studies provide a theoretical foundation for wind-storage grid-connected synergy, but further research is needed on "how heterogeneous energy participates in power distribution with the same index," "how multiple grid-connecting units share frequency formation while retaining independent power regulation," and "how support and recovery are uniformly described by the same control law."

[0005] Furthermore, after the disturbance ends, energy storage needs to restore its State of Charge (SOC), and wind turbines need to restore their speed and aerodynamic standby. In traditional recovery strategies, SOC recovery and speed recovery are usually performed independently. When they absorb or release energy from the grid, they are prone to power fluctuations at the common grid connection point, causing secondary frequency disturbances. At the same time, existing solutions involve multiple controllers operating in a spliced ​​manner, such as transient support control, SOC management loop, and speed recovery loop. This requires tuning multiple sets of control parameters separately, which not only increases the difficulty of engineering commissioning but may also lead to mode competition and control conflicts between the control loops.

[0006] Therefore, there is an urgent need for a wind-storage collaborative control method that can uniformly describe the heterogeneous energy state, eliminate the relative oscillation of multiple virtual rotors, and connect the entire process of support and recovery. Summary of the Invention

[0007] The purpose of this invention is to provide a grid-type wind-storage collaborative control method based on energy co-state and shared virtual rotor.

[0008] To achieve the above objectives, the present invention is implemented according to the following technical solution:

[0009] This invention includes the following steps:

[0010] S1: Obtain the operating status parameters of the wind turbine and energy storage system, including the energy storage state of charge, wind turbine rotor mechanical angular velocity, wind turbine output power, energy storage system output power, and common grid connection point frequency;

[0011] S2: Construct the directional effective energy of energy storage and the directional effective energy of wind turbine according to the operating state parameters, and normalize the directional effective energy to obtain normalized effective energy;

[0012] The energy storage directional effective energy includes upward supporting directional effective energy. and downward adjustment of directional effective energy The directional effective energy of the wind turbine includes upward support directional effective energy. and downward adjustment of directional effective energy ,in For the rated energy of energy storage, In the state of energy storage charge, , For the lower and upper bounds of the state of charge, , For energy storage discharge efficiency and charging efficiency. The equivalent rotational inertia of the wind turbine. The mechanical angular velocity of the fan rotor. , For the lower and upper limits of rotor angular velocity safety, For energy assessment time window, , To adjust the pneumatic standby power up and down.

[0013] The normalized effective energy is:

[0014]

[0015] in As directional effective energy, For the safety energy that must be reserved, The desired operating energy;

[0016] The centered logarithmic energy potential function is:

[0017] ,

[0018] The energy costate is:

[0019]

[0020] in This is the efficiency weighting coefficient.

[0021] S3: Construct a centered logarithmic energy potential function based on the normalized effective energy, and define an energy costate with the negative gradient of the centered logarithmic energy potential function with respect to the directional effective energy;

[0022] S4: Generate the total active power support requirements of the power station based on the frequency deviation of the common grid connection point;

[0023] S5: Using the energy costate as the marginal energy pressure index, construct a strict convex power allocation optimization model with the wind turbine support power amplitude and the energy storage system support power amplitude as decision variables, and solve the optimization model to obtain the support power command of the wind turbine and the energy storage system.

[0024] The strictly convex power allocation optimization model is as follows:

[0025]

[0026] The constraints are , , ,in , These are the supporting power amplitudes for energy storage systems and wind turbines, respectively. , The quadratic cost coefficient is positive. , The energy costates of the energy storage system and the wind turbine are respectively. This represents the total active power support amplitude of the station. , This represents the upper limit of the maximum supporting power.

[0027] Solving the optimization model involves: ignoring power upper and lower bound constraints, and obtaining an unconstrained analytical solution based on the KKT conditions.

[0028]

[0029] Then, based on the power upper and lower limit constraints, analytical projection is performed to obtain the constrained optimal solution.

[0030]

[0031] ,

[0032] The upper limit of the maximum support power It is determined by at least one of the following: energy constraint boundary, current constraint boundary, power change rate constraint boundary, and mechanical constraint boundary: The energy constraint boundary is , To predict the protection time.

[0033] S6: Establish the power-frequency dynamic equation of the shared virtual rotor, and control the angular frequency and phase angle of the shared virtual rotor according to the total active power support requirements of the wind storage station and the actual total power of the wind storage station;

[0034] The power-frequency dynamic equation of the shared virtual rotor is:

[0035]

[0036] in To share virtual inertia, To share the virtual rotor angular frequency, This serves as a valuable reference for all stations. This represents the actual total power of the wind storage station. The damping coefficient is... For frequency deviation, The rated angular frequency, This is a variable inertia compensation term.

[0037] The shared virtual inertia Based on comprehensive energy availability Continuous adjustment: ,in , These are the minimum and maximum values ​​of the virtual inertia. The overall energy availability , , These are the normalized effective energy of the energy storage system and the wind turbine, respectively.

[0038] S7: Generate voltage reference signals for each grid-type converter based on the supporting power command and the phase angle of the shared virtual rotor, so as to control the wind turbine and energy storage system to output power according to the supporting power command.

[0039] The invention further includes: superimposing an internal differential angle on the phase angle of the shared virtual rotor. The phase angles of the energy storage converters are generated respectively. Phase angle of wind turbine converter ,in To share the virtual rotor phase angle, The allocation coefficient is used; the differential angle reference value is calculated based on the support power command. ,in , The synchronization power factor at the current operating point. The internal differential angle is controlled to track the differential angle reference value to achieve internal power redistribution.

[0040] The invention also includes a zero net power recovery step: after the disturbance ends, an internal recovery optimization model is constructed based on the energy costate difference between the energy storage system and the wind turbine.

[0041]

[0042] The solution yields the restored switching power. The recovery power of the energy storage system is controlled to be The recovery power of the wind turbine is This causes the recovery power to cancel each other out at the common grid connection point.

[0043] The beneficial effects of this invention are:

[0044] This invention constructs directional effective energy, mapping energy storage SOC, wind turbine rotor kinetic energy, and load shedding reserve into a unified normalized effective energy, and further defines energy costate as a unified marginal energy pressure index. Compared with existing technologies that set thresholds for each state variable for zoned control, this invention eliminates the need to set switching thresholds and fixed filter boundaries for energy storage SOC and wind turbine speed, fundamentally avoiding allocation mismatch problems caused by multivariate threshold combination explosion and operating point drift. Energy costate characterizes the "urgency" of each device in a continuous dimensional form, enabling power allocation to change smoothly and continuously with operating conditions, avoiding power surges caused by hard threshold switching.

[0045] This invention constructs a strictly convex power allocation model with energy co-linear terms and quadratic cost terms, and provides a constrained analytical projection solution. Compared with existing schemes that rely on numerical optimization or heuristic rules, this invention can obtain a unique optimal allocation result in closed-form solution within each control cycle, without iterative calculation, with low computational cost, meeting real-time control requirements; at the same time, strict convexity guarantees the uniqueness of the solution, avoiding power jumps caused by ambiguity of multiple solutions.

[0046] This invention uses a single shared virtual rotor instead of separate virtual synchronous machines for wind and energy storage, fundamentally eliminating the relative oscillation mode between the two virtual rotors and avoiding active power circulation and control competition problems. Simultaneously, a compensation term is introduced into the variable inertia regulation. This ensures that the virtual inertia does not inject additional energy or generate additional damping disturbances into the system when it changes online, thus guaranteeing power-frequency stability during the dynamic adjustment of inertia. The mechanism of continuously adjusting the inertia magnitude based on overall energy availability enables the system to provide strong inertia support when energy is sufficient, and automatically reduce the inertia commitment at energy boundaries, avoiding exceeding actual energy capabilities.

[0047] This invention superimposes an internal differential angle on a shared phase angle and determines the allocation coefficient according to the synchronization power coefficient. This ensures that the shared virtual rotor phase angle only determines the power-frequency dynamics of the power station, while the differential angle only changes the internal power distribution within the wind and energy storage system. The two are structurally decoupled in a first-order sense. This structure retains the stability advantage of a single frequency mode while providing independent internal power adjustment freedom. Compared to existing independent virtual synchronizing machine schemes, it achieves stable power distribution adjustment without the need for additional virtual impedance or damping components.

[0048] This invention derives a zero net power recovery law from the same energy costate. During the recovery process, the wind and energy storage recovery power cancels each other out at the common grid connection point, and the total power of the power station remains approximately constant. Compared with the existing technology where the SOC recovery loop and speed recovery loop operate independently, this invention avoids control conflicts caused by multi-loop splicing and eliminates secondary frequency disturbances caused by the injection of recovery power into the grid. The theoretical guarantee that the total energy potential function monotonically decreases before reaching the limit ensures that the recovery process always converges towards the costate equipotential state.

[0049] This invention uses a single centralized logarithmic energy potential function to connect the three stages of power support, equipment boundary constraints, and post-disturbance recovery. The energy costate of the transient support stage is continued in the recovery stage, eliminating the need for control law switching or parameter resetting. Compared with existing multi-controller splicing schemes that require separate tuning of droop control parameters, SOC recovery loop parameters, and speed recovery loop parameters, this invention significantly reduces the number of parameters to be tuned, thus lowering the difficulty of engineering debugging.

[0050] This invention provides complete theoretical support from four aspects: strict convex optimization uniqueness, positive invariance of energy boundary (the forward invariance of the safe set is guaranteed by the energy constraint boundary), Lyapunov stability of shared virtual rotor, and monotonic convergence of energy potential in the recovery phase, ensuring the stability and safety of the control method across the entire operating range. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the topology and control structure of the grid-connected wind storage system of the present invention;

[0052] Figure 2 This is a schematic diagram of the real-time process of the shared virtual rotor-energy co-state control of the present invention;

[0053] Figure 3 The comparison results show the PCC frequency response during the disturbance period;

[0054] Figure 4 This represents the distribution of active power support from wind turbines and energy storage during the disturbance period;

[0055] Figure 5 The results show the comparison of the net recovery power of PCC after the disturbance ends. Detailed Implementation

[0056] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0057] The core idea of ​​this invention is to construct only one energy potential function and form an energy costate with its gradient; wind-storage transient power allocation, equipment boundary avoidance, shared virtual inertia adjustment, and energy recovery after disturbance are all derived from this costate. The main work includes:

[0058] By constructing directional effective energy and a centralized logarithmic potential function, the energy storage SOC, wind turbine rotor kinetic energy, and load reduction reserve are unified into marginal energy pressure.

[0059] A strictly convex wind-storage support power allocation model is established, and the analytical optimal solution with equipment boundaries is derived to avoid fixed frequency division and hard threshold switching.

[0060] The design shares virtual rotor and differential angle states, decoupling the total power-frequency dynamics of the power station from the internal power redistribution dynamics of the wind and storage system in the first-order sense.

[0061] The zero net power recovery law is derived from the same energy costate, and it is proven that the total energy potential does not increase monotonically during the recovery phase.

[0062] The research object consists of a full-power permanent magnet synchronous generator wind turbine, a battery energy storage system, two grid-connected converters, and a common grid connection point, such as Figure 1 As shown. Both the wind turbine grid-side converter and the energy storage converter operate as controlled voltage sources, and the station controller outputs share the virtual rotor angle, internal power redistribution angle, and voltage amplitude reference. It is specified that the wind turbine and energy storage output active power to the PCC respectively have the following parameters: and The total output of the station is .

[0063] Directional effective energy storage:

[0064] Assume the rated energy of the energy storage is If the SOC is z, then the internal energy of the battery is:

[0065] (1)

[0066] Using the positive direction definition of AC side power and considering charging and discharging efficiency, the dynamic energy storage is as follows:

[0067] (2)

[0068] To simultaneously describe both power deficit and power surplus operating conditions of the system, directional effective energy is defined for upward support and downward regulation, respectively:

[0069] (3)

[0070] In the formula, and These represent the SOC security level and the upper limit, respectively. and These represent discharge and charging efficiencies, respectively. This indicates the energy that the stored energy can continue to release to the AC side. This indicates the AC energy that the energy storage can continue to absorb.

[0071] Directional effective energy of the wind turbine:

[0072] The short-term power support of the wind turbine mainly comes from the rotor kinetic energy, while continuous support relies on the aerodynamic reserve formed by load reduction. The rotor mechanical dynamics are as follows:

[0073] (4)

[0074] In the formula, For equivalent rotational inertia, The rotor's mechanical angular velocity, , and These are aerodynamic power, electromagnetic power, and mechanical losses, respectively. Let... For energy assessment time window, To upgrade the pneumatic backup, To determine the minimum base power generation, the directional effective energy of a wind turbine is defined as follows:

[0075] (5)

[0076] (6)

[0077] Equation (5) adds the releasable rotor kinetic energy to the continuous reserve within the predicted time window, avoiding estimation of the wind turbine support capacity solely based on speed deviation. Equation (6) corresponds to the down-regulation capability when the system has excess power, where the first term is the space of kinetic energy that the rotor can absorb. In actual operation, the sign of the total power increment is selected. or From now on, it will be uniformly recorded as .

[0078] The main symbols and their physical meanings are shown in Table 1:

[0079] Table 1. Main Symbols and Physical Meanings

[0080]

[0081] Shared Virtual Rotor—Energy Co-state Equipotential Control:

[0082] Centralized energy potential and energy costate

[0083] To eliminate the dimensional differences between SOC, speed, and reserve power, the directional effective energy is normalized as follows:

[0084] (7)

[0085] in, For the safety energy that must be reserved, The desired operating energy is determined by selecting a centralized logarithmic energy potential:

[0086] (8)

[0087] in, For cell index, ; Indicates energy storage system, Indicates wind turbine unit; For the first The positive weighting coefficients of the unit-centered logarithmic energy potential function, and ; For the first The efficiency weighting coefficient of the unit.

[0088] The function satisfies , ( ), and when Approaching 0 It tends towards positive infinity, thus possessing both the attraction property of the reference state and the safety boundary barrier property. Its partial derivative with respect to the effective energy is:

[0089] (9)

[0090] Define an energy costate as an efficiency-weighted form of the energy potential with respect to the negative gradient of the effective energy:

[0091] (10)

[0092] when hour, When the effective energy of the equipment is lower than the reference value, The closer the equipment is to the safety boundary, The larger. Therefore, This can be interpreted as the marginal energy pressure caused by the unit continuing to bear the unit supporting power. If the effective energy of the unit is dynamically written as:

[0093] (11)

[0094] in This indicates that wind power supplementation, charging, or other external energy input is used during the rapid support phase. When it is relatively small, there is .thus, The rate of increase in energy potential after the unit assumes the supporting power is directly characterized.

[0095] Power requirements for power station support

[0096] The frequency deviation and integral state of the shared virtual rotor are defined as follows:

[0097] (12)

[0098] Generate the station's total active power reference based on the frequency deviation:

[0099] (13)

[0100] In the formula, To disrupt the planned power of the front station, and Let be the primary frequency modulation coefficient and the steady-state recovery coefficient, respectively. Let . Then the support direction, total support amplitude, and power balance constraints are:

[0101] (14)

[0102] when When the wind storage increases output; when At this time, the wind storage reduces output or absorbs excess power. Power amplitude is used. This allows both directions to share the same optimization model.

[0103] Energy Costate Driven Analytical Power Allocation

[0104] Under the condition of meeting the total support requirements of the station, construct an instantaneous strictly convex optimization problem:

[0105] (15)

[0106] in, and Positive terms are used to characterize secondary costs such as energy storage power throughput, battery aging, changes in wind turbine electromagnetic torque, and mechanical fatigue; linear terms... Let be the rate of increase of the energy potential. Ignoring the upper and lower limits, the Lagrange function is:

[0107] (16)

[0108] right and Taking the partial derivative and setting it to zero, we obtain the KKT stationary point condition:

[0109] (17)

[0110] By applying the power balance constraints simultaneously, we can obtain the unconstrained analytical solution:

[0111] (18)

[0112] Equation (18) shows that power allocation is determined by the base cost. Costate difference Jointly decided. When energy storage approaches the lower limit of SOC. As the wind speed increases, the energy storage support automatically decreases and continuously shifts to the wind turbine; the reverse occurs when the wind turbine speed approaches its lower limit. This process does not require setting a SOC threshold, speed threshold, or fixed filter cutoff frequency.

[0113] Considering the maximum support capacity of the unit, the feasible range for energy storage is [max(0, - ), min( , Since the problem contains only two variables, its optimal solution can be obtained by analytical projection:

[0114] (19)

[0115] (20)

[0116] Analytical projection guarantee Therefore, the overall power balance of the station will not be disrupted by unit limiting; when the total available power is insufficient, the upper-level controller should take appropriate measures. Perform a feasibility reduction and trigger an insufficient capacity flag.

[0117] Equipment constraint boundaries:

[0118] Energy, current, rate of change of power, and mechanical constraints are all unified and written as the upper bound of supporting power:

[0119] (twenty one)

[0120] Energy constraints are based on predicted protection time. The construction is as follows:

[0121] (twenty two)

[0122] This boundary satisfies the worst dynamical requirement for effective energy. This ensures that the energy lower bound is not exceeded when the initial state is safe. Considering the apparent power capacity occupied by reactive power commands, the upper bound of active power support corresponding to the current constraint is:

[0123] (twenty three)

[0124] The upper bound of the power change rate can be calculated from the command of the previous control cycle and the allowable ramp rate; the mechanical boundary of the wind turbine should also take into account the maximum electromagnetic torque, minimum speed, and available aerodynamic reserve. Thus, equipment constraints are part of the optimized feasible region, rather than being simply truncated at the controller end.

[0125] Shared virtual rotor and variable inertia compensation

[0126] The wind turbine and energy storage no longer have separate virtual rotors, but instead share a common angle and frequency state:

[0127] (twenty four)

[0128] The power-frequency dynamics of the shared virtual rotor are:

[0129] (25)

[0130] Additional terms in the formula Used to compensate for energy injection or extraction caused by variable inertia. The overall energy availability is defined as:

[0131] (26)

[0132] Shared virtual inertia is continuously adjusted by overall energy availability:

[0133] (27)

[0134] When the effective energy of the wind storage is sufficient, a larger virtual inertia can be configured to suppress the rate of frequency change; when the overall energy approaches the boundary, the virtual inertia automatically decreases to avoid the controller promising an inertial response beyond the actual energy capacity. In engineering implementation, this should be addressed... Amplitude limiting is used to prevent rapid parameter changes.

[0135] Internal differential angle and power decoupling

[0136] Sharing the same phase angle makes it impossible to independently adjust the active power of the wind turbine and energy storage, while setting up two independent virtual rotors would introduce relative oscillation. Therefore, in the shared angle... An internal differential angle is superimposed on top. :

[0137] (28)

[0138] Under inductive network and small phase angle conditions, the active power increment of wind storage is linearized as follows:

[0139] (29)

[0140] in, To share the power angle between the virtual rotor and the PCC equivalent grid, This is the increment of the work angle relative to the current working point. and Let be the operating point synchronization power coefficient.

[0141] (30)

[0142] Then the total power increment of the station Strict offsetting terms:

[0143] (31)

[0144] The incremental energy storage power can be written as:

[0145] (32)

[0146] therefore, Only the aggregated power-frequency dynamics of the site are determined. Only the internal power distribution within the wind storage is changed. Based on the optimal distribution result of equation (20), the differential angle reference is:

[0147] (33)

[0148] To improve robustness to changes in line parameters and power measurement errors, first-order tracking and power error correction are employed:

[0149] (34)

[0150] in This is a normalized correction value constructed from the energy storage power distribution error. The structure retains a single shared frequency mode while providing a stable internal power regulation degree of freedom.

[0151] Reactive power and voltage control:

[0152] The total reactive power reference of the power station is formed by the PCC voltage deviation:

[0153] (35)

[0154] Reactive power is allocated according to the remaining capacity of each converter after deducting active power:

[0155] (36)

[0156] in, ; For unit The remaining apparent power capacity index after executing active power instructions For unit The reactive power distribution factor, , These are the remaining capacity indicators for energy storage systems and wind turbines, respectively. A small regularization constant is set to prevent the denominator from being zero; For unit Maximum apparent power capacity, For unit Active power command, For unit reactive power command, This serves as a reference for the total reactive power of the station.

[0157] The network voltage vector reference is:

[0158] (37)

[0159] In the formula, Generated by reactive power-voltage regulation loop, For virtual impedance matrix, For the output current vector. If reactive current priority is adopted during the fault, equation (23) will simultaneously reduce the active power support boundary, and the analyzer will then transfer the active power task to the unit with larger remaining capacity.

[0160] Zero net power energy recovery:

[0161] After the disturbance ends, energy storage needs to restore its State of Charge (SOC), and wind turbines need to restore their speed and aerodynamic standby. If both recover energy from the grid separately, it can easily cause secondary frequency fluctuations. This paper requires that the recovered power cancel each other out at the PCC (Power Constraints).

[0162] (38)

[0163] Using the energy costate from the transient support phase, we construct an internal recovery optimization problem:

[0164] (39)

[0165] Its band-limited analytical solution is:

[0166] (40)

[0167] like This indicates that the marginal energy pressure of energy storage is relatively large, as given by equation (40). That is, the wind turbine increases its output and replenishes the energy storage; if The reduced electromagnetic power of the wind turbine is temporarily compensated by the energy storage, allowing the rotor to recover. When the limit is not reached, the derivative of the total energy potential is:

[0168] (41)

[0169] Therefore, the recovery process automatically reduces the costate difference and tends towards... Transient support and post-disturbance recovery are derived from the same potential function, avoiding control conflicts caused by setting up separate SOC recovery loops and speed recovery loops.

[0170] Stability and constraint property analysis:

[0171] Existence and uniqueness of the solution:

[0172] The Hessian matrix of equation (15) is .when , At this time, the objective function is strictly convex; the power balance constraint is an affine constraint, and the upper and lower power limits form a closed convex set. Therefore, as long as If the total available support capacity of wind storage does not exceed the total available support capacity, the optimization problem has a unique global optimal solution, and equations (19) and (20) are the analytical expressions of this solution.

[0173] Energy boundary positive invariance:

[0174] From equation (22), we can obtain the worst-case scenario. Applying the principle of comparison, we have:

[0175] (46)

[0176] If the initial state satisfies Then at any finite time, we have Therefore, the constructed supporting power boundary ensures the positive invariance of the energy safety set. This property is based on the condition that energy estimation and actuator tracking errors are bounded, and should be applied in engineering applications. Reserve a margin for model error.

[0177] Shared virtual rotor stability

[0178] Define the kinetic energy function of the shared virtual rotor:

[0179] (42)

[0180] Differentiating it and substituting it into equation (25), we get:

[0181] (43)

[0182] It can be seen that the variable inertia compensation term cancels out the effect. ,make Online changes do not generate additional negative damping. For the case where the power station is connected to an equivalent inductive power grid, let... The work point is Constructing incremental potential energy:

[0183] (44)

[0184] Take the Lyapunov function and find its derivative:

[0185] (45)

[0186] when and hour, It is positive definite in the neighborhood of the working point. Semi-negative constant. Combining the principle of invariant sets, we can see that the shared virtual rotor angular frequency error converges to zero, and the power angle converges to a feasible equilibrium point.

[0187] Internal angular modes and recovery convergence:

[0188] From equation (30), it can be seen that the differential angle The first-order model does not include the total power of the power station. When the inner loop bandwidth of the power is higher than... During loop operation, the internal error is dynamically approximated as follows: Its characteristic roots are located in the left half-plane. On the other hand, Equation (41) shows that the total energy potential does not increase monotonically during the zero net power recovery phase, and tends to the co-equipotential state when the limit is not reached and the system parameters are stable.

[0189] Real-time control implementation:

[0190] The controller employs a multi-timescale implementation. The inner current loop and voltage loop operate within the converter control cycle; shared virtual rotor, differential angle, and power distribution operate at the millisecond level; energy assessment and co-state updates can be synchronized with active power control; pitch control and long-term standby recovery operate at the second level. The main steps of each control cycle are as follows:

[0191] 1) Collection , , , ,z, , , and Isostate variables;

[0192] 2) Calculate the directional effective energy of wind storage based on the support direction. and and update , and ;

[0193] 3) Generated by frequency deviation Calculate s and ;

[0194] 4) Calculate energy, current, rate of change of power, and mechanical constraint boundaries. , ;

[0195] 5) Analytical allocation using equations (18) to (20) and ;

[0196] 6) Update , and And generated by equation (33) , and ;

[0197] 7) Generating active, reactive, and voltage vector references is performed by two sets of grid-connected converters;

[0198] 8) After the disturbance ends, switch to zero net power recovery of equation (40) until SOC, speed and reserve capacity return to the allowable range.

[0199] Simulation verification and result analysis:

[0200] To verify the control effectiveness of the grid-based wind-storage coordinated control method based on energy co-state and shared virtual rotor proposed in this invention in frequency support, wind-storage power coordination and allocation, and energy recovery after disturbance, a joint simulation system of wind and storage was established, consisting of a full-power permanent magnet synchronous wind turbine, a battery energy storage system, two sets of grid-based converters, a common grid connection point (PCC), and an equivalent AC grid. Both the wind turbine grid-side converter and the energy storage converter operate in a grid-based controlled voltage source mode, sharing the angular frequency and phase angle of a common virtual rotor, and achieving active power redistribution between the wind and storage units through internal differential angles. According to the control method of this invention, the releasable / absorbable energy corresponding to the energy storage SOC, along with the wind turbine rotor kinetic energy and aerodynamic reserve, are uniformly constructed as directional effective energy and further normalized to form a unified energy state. The energy co-state of energy storage and the wind turbine is calculated using a centralized logarithmic energy potential function, serving as a unified index characterizing the marginal energy pressure generated by the equipment continuing to bear unit support power. This energy co-state increases as the equipment approaches the energy safety boundary, thereby enabling continuous changes in the wind-storage power sharing ratio and avoiding the hard switching phenomenon caused by traditional SOC zoning or speed threshold control.

[0201] The simulation uses per-unit values ​​for calculation, with a 50 Hz system as the illustrative rated frequency. The system is set to a stable operating state before the disturbance, and an active power deficit-type frequency disturbance is applied at approximately 2 seconds to examine the dynamic frequency support performance under different control strategies. According to the method of this invention, the total active power support demand of the power station is generated based on the PCC frequency deviation. Under the condition that the sum of the energy storage power and the wind turbine power equals the total support power of the power station, a rigorous convex optimization model containing a quadratic power cost and an energy co-linear term is constructed. An unconstrained analytical solution is obtained through KKT conditions. Then, the analytical solution is projected based on the energy, current, power change rate, and mechanical constraints of the energy storage and wind turbines to obtain the optimal support power command for each grid unit. This method has a unique globally optimal solution when the optimization problem is feasible and does not require online iterative solving, making it suitable for millisecond-level real-time control. Meanwhile, the shared virtual inertia is continuously adjusted according to the overall energy availability of wind and storage, increasing the inertia support capacity when energy is sufficient and actively reducing the inertia commitment when the system approaches the energy boundary; the variable inertia compensation term in the shared virtual rotor is used to avoid additional energy injection or additional damping disturbance caused by online changes in inertia.

[0202] Figure 3 The comparison results of PCC frequency response during disturbances are presented. Different line types, rather than colors, are used in the figure to distinguish different control strategies. The solid black line represents the "shared virtual rotor + energy co-mode" control of this invention, the dashed line represents the traditional dual VSG cooperative control, and the dotted-dashed line represents fixed droop distribution control. As can be seen from the figure, after a frequency disturbance, the decrease in PCC frequency deviation under the control of this invention is significantly smaller, with the lowest frequency point being approximately -0.18 Hz, and it can smoothly recover to near the rated frequency in a short time, with virtually no obvious secondary oscillations during the recovery process. In contrast, the traditional dual VSG scheme further reduces the lowest frequency point and exhibits more significant overshoot and damped oscillations during the recovery process. This is because setting independent virtual rotors for the wind turbine and energy storage may create internal relative oscillation modes, resulting in active power circulation and control competition. This invention, however, retains only one shared virtual rotor frequency state, structurally eliminating the relative oscillation between the two independent virtual rotors. This result is consistent with the design goal of this invention, which aims to reduce internal control competition and improve the power-frequency dynamics of the power plant using a shared virtual rotor. The fixed droop distribution method has relatively poor frequency minimum point and recovery time, indicating that the fixed parameters cannot adaptively adjust the support capacity according to the real-time available energy status of wind storage.

[0203] Figure 4The distribution results of active power support for wind turbines and energy storage during disturbances are further presented. To enhance the legibility under black and white printing conditions, the wind turbine power, energy storage power, conventional wind turbine power, and conventional energy storage power are represented by solid lines, dashed lines, dotted lines, and dotted lines, respectively. The results show that under the control of this invention, the support power of both wind turbines and energy storage changes continuously and smoothly, and their sum always meets the total power support requirements of the site. When the effective energy of energy storage is relatively low and the corresponding energy costate is large, the optimizer automatically reduces the transient support ratio undertaken by energy storage and continuously transfers more power demand to wind turbines; as the energy state of wind and energy storage changes, their power support ratio is adjusted synchronously and continuously. This phenomenon is consistent with the analytical allocation law of this invention, that is, the power allocation depends not only on the secondary cost of the foundation, but also directly on the energy costate difference between energy storage and wind turbines; when a unit approaches its energy safety boundary, its support task will automatically transfer to another unit without setting a separate SOC threshold, speed threshold, or fixed filter boundary frequency. In contrast, traditional fixed threshold or fixed allocation methods are prone to power spikes, plateauing, and mismatches between allocation ratios and actual available energy. This invention, through rigorous convex analytical allocation and embedding device constraints into the feasible region, enables continuous changes in power commands while maintaining power balance at the field station level when cells reach their capacity limits.

[0204] Figure 5 The comparison results of the net recovery power of PCC after the disturbance ends are given. After the disturbance disappears, both the wind turbine rotor speed and the energy storage SOC need to be gradually restored. Traditional independent recovery methods usually set up SOC recovery loops and wind turbine speed recovery loops separately. The two units may absorb or release power from the grid at the same time, thus forming a significant net recovery power at PCC, causing secondary frequency disturbances. To this end, this invention continues to use the energy costate of the transient support stage, calculates the internal recovery exchange power based on the costate difference between energy storage and wind turbine, and makes the energy storage recovery power equal in magnitude and opposite in direction to the wind turbine recovery power, so that the two cancel each other out at PCC. In equation (38), this invention stipulates that the recovery stage satisfies , During the recovery phase, the energy storage recovery power is equal in magnitude and opposite in direction to the wind turbine recovery power, i.e., the energy storage recovery power is... The fan recovery power is And by driving the recovery of exchange power through co-state difference, the total energy potential remains monotonically constant until the limit is reached. Therefore, Figure 5In this invention, the corresponding solid line remains essentially near zero power, while traditional independent recovery and segmented recovery strategies exhibit significant positive and negative power fluctuations in the initial recovery phase and require a considerable amount of time to return to zero. This comparison illustrates that this invention can transform energy storage SOC recovery and wind turbine speed / reserve recovery into an energy exchange process within the wind-storage station, achieving energy state recovery without significantly altering the total active power output (PCC), thereby effectively reducing the secondary frequency disturbances that may result from traditional independent recovery processes.

[0205] The combined simulation results from the three sets demonstrate that this invention establishes a complete closed-loop control chain: "frequency disturbance identification—unified energy state representation—strict convex analytical power allocation—shared virtual rotor frequency formation—internal power redistribution within the differential angle—zero net power energy recovery." During the rapid support phase, the energy co-state can adjust the wind-storage load ratio in real time based on the energy storage SOC, wind turbine rotor kinetic energy, and reserve capacity. The shared virtual rotor avoids relative oscillations between multiple independent virtual rotors and achieves continuous adaptive adjustment of inertia through comprehensive energy availability. In the recovery phase after the disturbance, the same energy co-state continues to participate in internal recovery control without needing to switch control indices again. The real-time implementation scheme provided in the patent also shows that the shared virtual rotor, differential angle, and power allocation can operate within millisecond-level control cycles, energy assessment and co-state updates can be synchronized with active power control, while pitch control and long-term reserve recovery operate on a slower time scale, thus meeting the engineering implementation requirements for multi-time-scale coordinated control of grid-type wind-storage power plants.

[0206] The technical solutions of the present invention are not limited to the specific embodiments described above. Any technical modifications made in accordance with the technical solutions of the present invention fall within the protection scope of the present invention.

Claims

1. A grid-type wind-storage collaborative control method based on energy co-state and shared virtual rotor, characterized in that, Includes the following steps: S1: Obtain the operating status parameters of the wind turbine and energy storage system, including the energy storage state of charge, wind turbine rotor mechanical angular velocity, wind turbine output power, energy storage system output power, and common grid connection point frequency; S2: Construct the directional effective energy of energy storage and the directional effective energy of wind turbine according to the operating state parameters, and normalize the directional effective energy to obtain normalized effective energy; S3: Construct a centered logarithmic energy potential function based on the normalized effective energy, and define an energy costate with the negative gradient of the centered logarithmic energy potential function with respect to the directional effective energy; S4: Generate the total active power support requirements of the power station based on the frequency deviation of the common grid connection point; S5: Using the energy costate as the marginal energy pressure index, construct a strict convex power allocation optimization model with the wind turbine support power amplitude and the energy storage system support power amplitude as decision variables, and solve the optimization model to obtain the support power command of the wind turbine and the energy storage system. S6: Establish the power-frequency dynamic equation of the shared virtual rotor, and control the angular frequency and phase angle of the shared virtual rotor according to the total active power support requirements of the wind storage station and the actual total power of the wind storage station; S7: Generate voltage reference signals for each grid-type converter based on the supporting power command and the phase angle of the shared virtual rotor, so as to control the wind turbine and energy storage system to output power according to the supporting power command.

2. The grid-type wind-storage coordinated control method based on energy co-state and shared virtual rotor as described in claim 1, characterized in that, The energy storage directional effective energy includes upward supporting directional effective energy. and downward adjustment of directional effective energy The directional effective energy of the wind turbine includes upward support directional effective energy. and downward adjustment of directional effective energy ,in For the rated energy of energy storage, In the state of energy storage charge, , For the lower and upper bounds of the state of charge, , For energy storage discharge efficiency and charging efficiency. The equivalent rotational inertia of the fan. The mechanical angular velocity of the fan rotor. , For the lower and upper limits of rotor angular velocity safety, For energy assessment time window, , To adjust the pneumatic standby power up and down.

3. The grid-type wind-storage coordinated control method based on energy co-state and shared virtual rotor as described in claim 1, characterized in that, The normalized effective energy is: in As directional effective energy, For the safety energy that must be reserved, The desired operating energy; The centralized logarithmic energy potential function is: ， The energy costate is: in, For cell index, ; Indicates energy storage system, Indicates wind turbine unit; For the first The positive weighting coefficients of the unit-centered logarithmic energy potential function, and ; For the first The efficiency weighting coefficient of the unit.

4. The grid-type wind-storage coordinated control method based on energy co-state and shared virtual rotor as described in claim 1, characterized in that, The strictly convex power allocation optimization model is as follows: The constraints are , , ,in , These are the supporting power amplitudes for energy storage systems and wind turbines, respectively. , The quadratic cost coefficient is positive. , These are the energy costates of the energy storage system and the wind turbine, respectively. This represents the total active power support amplitude of the station. , This represents the upper limit of the maximum supporting power.

5. The grid-type wind-storage coordinated control method based on energy co-state and shared virtual rotor according to claim 4, characterized in that, Solving the optimization model involves: ignoring power upper and lower bound constraints, and obtaining an unconstrained analytical solution based on the KKT conditions. Then, based on the power upper and lower limit constraints, analytical projection is performed to obtain the constrained optimal solution. ， in, For unit The optimal support power amplitude, Units with direction symbols Optimal active power support increment, where s is the sign of the power station support direction. , These represent the unconstrained optimal support power amplitudes for energy storage systems and wind turbines, respectively. , These are the constrained optimal support power amplitudes.

6. The grid-type wind-storage coordinated control method based on energy co-state and shared virtual rotor according to claim 1, characterized in that, The power-frequency dynamic equation of the shared virtual rotor is: in To share virtual inertia, To share the virtual rotor angular frequency, This serves as a valuable reference for all stations. This represents the actual total power of the wind storage station. The damping coefficient is... For frequency deviation, The rated angular frequency, This is a variable inertia compensation term.

7. The grid-type wind-storage coordinated control method based on energy co-state and shared virtual rotor as described in claim 6, characterized in that, The shared virtual inertia Based on comprehensive energy availability Continuous adjustment: ,in , These are the minimum and maximum values ​​of the virtual inertia. The overall energy availability , , These are the normalized effective energy values ​​of the energy storage system and the wind turbine, respectively. , These are the non-negative weighting coefficients for energy storage systems and wind turbines in the overall energy availability, respectively.

8. The grid-type wind-storage coordinated control method based on energy co-state and shared virtual rotor according to claim 1, characterized in that, Also includes: An internal differential angle is superimposed on the phase angle of the shared virtual rotor. The phase angles of the energy storage converters are generated respectively. Phase angle of wind turbine converter ,in To share the virtual rotor phase angle, The allocation coefficient is used; the differential angle reference value is calculated based on the support power command. ,in , The synchronization power factor at the current operating point. The internal differential angle is controlled to track the differential angle reference value to achieve internal power redistribution. To provide the optimal active power support increment for the energy storage system, To increase the total active power support of the station, The equivalent synchronous power coefficient for power redistribution within the wind storage system.

9. The grid-type wind-storage coordinated control method based on energy co-state and shared virtual rotor according to claim 1, characterized in that, It also includes a zero net power recovery step: after the disturbance ends, an internal recovery optimization model is constructed based on the energy costate difference between the energy storage system and the wind turbine. The solution yields the restored switching power. The recovery power of the energy storage system is controlled to be The recovery power of the wind turbine is This causes the recovery power to cancel each other out at the common grid connection point, where, Optimize the objective function for internal recovery; To restore the decision variables for switching power; >0 represents the secondary cost coefficient for restoring switching power; , These are the energy costates of the energy storage system and the wind turbine, respectively. To achieve optimal recovery of switching power; >0 represents the upper limit of the recovery switching power amplitude; sat is the saturation function; , The recovery power of the energy storage system and the wind turbine are respectively obtained; Later orders Perform recovery control.

10. The grid-type wind-storage coordinated control method based on energy co-state and shared virtual rotor according to claim 4, characterized in that, The upper limit of the maximum support power It is determined by at least one of the following: energy constraint boundary, current constraint boundary, power change rate constraint boundary, and mechanical constraint boundary: The energy constraint boundary is , To predict the protection time, among which, , Indicates energy storage system, Indicates wind turbine unit; For unit Maximum supporting power upper limit; , , , These are the upper limits of the supporting power determined by energy constraints, current constraints, power change rate constraints, and mechanical constraints, respectively. For unit Directional effective energy, For the safety energy that must be reserved, For unit The efficiency weighting coefficient, >0 represents the predicted protection time.

Images