A virtual synchronous generator cooperative adaptive control method based on reinforcement learning

Abstract

The application discloses a kind of virtual synchronous generator collaborative adaptive control methods based on reinforcement learning.First, the mathematical model of virtual synchronous machine is constructed and the active and reactive regulation scheme is determined;Then, through small signal model analysis, the influence of virtual inertia and damping coefficient on the characteristics of virtual synchronous generator is determined, and the selection method of rated virtual inertia and rated damping coefficient is proposed;Finally, based on the reinforcement learning algorithm, the virtual inertia controller and the damping coefficient controller are designed, and the optimal values of inertia and damping under different states are generated collaboratively and adaptively.The collaborative adaptive control method generates corresponding inertia and damping values according to different operating states of the virtual synchronous generator, realizes optimal control of virtual inertia and damping coefficient, and effectively improves the dynamic response capability and stability of the system.

Description

Technical Field

[0001] This invention belongs to the field of distributed energy grid connection, and in particular relates to a cooperative adaptive control method for virtual synchronous generators based on reinforcement learning. Background Technology

[0002] With increasing concerns about energy conservation and environmental pollution, microgrids have become a field of significant interest. In microgrids, distributed power sources are integrated with the grid through devices such as inverters. However, these inverters have an inherent limitation: a lack of inertia and damping characteristics. Their stability is poor when the system experiences disturbances. To address this, researchers both domestically and internationally have developed virtual synchronous generator control methods to enable grid-connected inverters to simulate the external characteristics of synchronous generators, effectively suppressing frequency fluctuations and improving system stability during disturbances.

[0003] Currently, many achievements have been made in virtual synchronous generator control technology. Existing technologies can fully consider the control flexibility of inverters and adaptively generate target values ​​for virtual inertia and damping coefficients. However, the selection of virtual inertia and damping coefficients is generally controlled independently based on their relationship with the system frequency, and the control threshold and other parameters are generally derived from experience or simulation. This results in the existing control strategies being unable to select the optimal virtual inertia and damping coefficients and having insufficient generalization ability. They also cannot adaptively adjust the dynamic response speed and stability of the system based on different system requirements. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a cooperative adaptive control method for virtual synchronous generators based on reinforcement learning, which realizes the cooperative adaptive generation of inertia and damping targets, and effectively improves the stability and dynamic response capability of the system.

[0005] This invention adopts the following technical solution: a cooperative adaptive control method for virtual synchronous generators based on reinforcement learning, comprising the following steps:

[0006] Step 1: Construct a mathematical model of the virtual synchronous engine;

[0007] Step 2: Through small-signal model analysis, determine the influence of virtual inertia and damping coefficient on the characteristics of virtual synchronous generator, and propose a method for selecting rated virtual inertia and rated damping coefficient;

[0008] Step 3: Based on reinforcement learning algorithms, design a collaborative adaptive target generation algorithm for virtual inertia and damping coefficients to generate optimal values ​​of inertia and damping under different states.

[0009] Furthermore, step 1, which involves constructing the mathematical model of the virtual synchronous generator, includes the following sub-steps:

[0010] Step 1.1: Determine the correspondence between the virtual synchronous generator model and the synchronous generator model, and establish the mathematical model of the virtual synchronous generator;

[0011] Step 1.2, based on the mechanical torque T m The adjustment is used to regulate the active power output of the virtual synchronous generator model, thereby implementing the inverter's active power regulation command.

[0012] Step 1.3: Based on the adjustment of the excitation, the reactive power output of the virtual synchronous generator model is adjusted to obtain the potential voltage of the virtual synchronous generator.

[0013] In step 1.1, determining the correspondence between the virtual synchronous generator model and the synchronous generator model...

[0014] The mechanical equations of the virtual synchronous generator are as follows:

[0015]

[0016] In the formula, J is the moment of inertia, D is the damping coefficient, ω is the angular velocity of the virtual synchronous generator, ω0 is the grid synchronization angular velocity, and T... m T e and T D They are mechanical torque, electromagnetic torque, and damping torque, respectively; δ is the power angle of the virtual synchronous generator; and dω, dt, and dδ are the differentials of the angular velocity, time, and power angle of the virtual synchronous generator, respectively.

[0017] Among them, T e The calculation formula is:

[0018]

[0019] In the formula, P e To output electromagnetic power, e a e b e c These are the three-phase output potentials, i a i b i c These are the three-phase output currents.

[0020] The active power regulation described in step 1.2 is achieved by using mechanical torque T. m The active power output of the virtual synchronous generator is controlled by adjusting the parameters, as shown in the following expression:

[0021] T m =T ref +△T (3)

[0022] In the formula, T refThe rated mechanical torque is T, where ΔT is the mechanical torque deviation. ref The expressions for ΔT are:

[0023]

[0024] In the formula, P ref K is the rated active power. ω This is the adjustment coefficient;

[0025] The reactive power regulation described in step 1.3 is achieved by adjusting the reactive power output through adjusting the potential E, as shown in the following formula:

[0026] E = E0 + K q (Q ref -Q)+K u (U ref -U) (5)

[0027] In the formula, E0 is the no-load potential, U ref U and K are the rated and actual values ​​of the virtual synchronous machine terminal voltage, respectively; q Q is the reactive power regulation coefficient. ref Rated reactive power; Q is instantaneous reactive power, which can be expressed as:

[0028]

[0029] In the formula, u a u b u c These are the three-phase node voltages;

[0030] Therefore, the potential-voltage vector of the virtual synchronous generator can be obtained as follows:

[0031]

[0032] in, This represents the phase of the virtual synchronous generator.

[0033] Furthermore, the small-signal model analysis described in step 2 includes the following sub-steps:

[0034] Step 2.1: Construct a second-order transfer function based on the small-signal model analysis method and the model of the virtual synchronous generator;

[0035] Step 2.2: Calculate the rated virtual inertia based on the transfer function;

[0036] Step 2.3: Calculate the rated damping coefficient based on the transfer function.

[0037] The second-order transfer function mentioned in step 2.1 is as follows:

[0038]

[0039] Where E is the voltage amplitude of the virtual synchronous generator, U is the terminal voltage of the virtual synchronous generator, Z is the impedance of the filter circuit of the virtual synchronous generator, s is the complex variable in the transfer function, G(s) is the second-order transfer function, and P(·) function is the input and output of the second-order transfer function.

[0040] The natural oscillation frequency is determined by referring to the synchronous generator, and the rated virtual inertia J0 mentioned in step 2.2 and the rated damping coefficient D0 mentioned in step 2.3 are obtained by the following formula (9);

[0041]

[0042] In the formula, ω n Let J be the natural oscillation angular frequency, and the J obtained by solving is the rated virtual inertia J0;

[0043] ξ is the damping coefficient. Based on the preset ξ value, D obtained by solving according to the above formula is the rated damping coefficient D0 mentioned in step 2.3.

[0044] Furthermore, step 3, the virtual inertia and damping coefficient collaborative adaptive target generation algorithm, includes the following sub-steps:

[0045] Step 3.1: Set weight coefficients: Based on the intelligent decision-making capability of reinforcement learning algorithm, design the state space of virtual inertia controller and damping coefficient controller, and use weight coefficients to balance the system's emphasis on dynamic response capability and stability.

[0046] Step 3.2, Inertia Target Generation: Design a suitable action space and reward function for the state space of the virtual inertia controller, so as to select appropriate actions in different state spaces, and generate an inertia target by combining the rated virtual inertia mentioned above;

[0047] Step 3.3, Damping Target Generation: Design a suitable action space and reward function for the state space of the damping coefficient controller, so as to select appropriate actions in different state spaces, and generate the damping target by combining the rated virtual inertia mentioned above.

[0048] The state spaces of the virtual inertia controller and damping coefficient controller mentioned in step 3.1 are as follows:

[0049]

[0050] In the formula, S J S is the state space of the virtual inertia controller. DLet be the state space of the damping coefficient controller, and a be the weighting coefficient, a∈[0.1,1]. When a=0.1, the system focuses on stronger response capability, and when a=1, the system focuses on stronger stability.

[0051] The inertia target generation method described in step 3.2 is as follows:

[0052] Set the state space S of the virtual inertia controller. J The discrete set, and the action set A of the controller. J The A J The output of the controller is the real-time set of scheduling instructions ΔJ, defined as A. J The action space;

[0053] The current state space of the virtual inertia controller is determined, and the action that yields the maximum reward value in the current state space is selected using the reward function. The inertia target is as follows:

[0054] J target =J0+ΔJ (11)

[0055] In the formula, J0 is the rated inertia, ΔJ is the output value of the virtual inertia controller, and J target The target value for the virtual inertia of the virtual synchronous generator.

[0056] The damping target generation method described in step 3.3 is as follows:

[0057] Set the state space S of the damping coefficient controller. D The discrete set, and the action set A of the controller. D The A D The output of the controller is the real-time set of scheduling instructions ΔD, defined as A. D The action space;

[0058] Determine the current state space of the damping coefficient controller, and select the action that yields the maximum reward value in the current state space using the reward function. The damping objective is as follows:

[0059] D target =D0+ΔD (12)

[0060] In the formula, ΔD is the damping coefficient controller output value, D0 is the rated damping coefficient, and D target This represents the target value for the damping coefficient of the virtual synchronous generator.

[0061] Compared with the prior art, the present invention, employing the above technical solution, has the following technical effects:

[0062] 1. This invention constructs a weighting coefficient connecting the state spaces of two controllers, realizing the coordinated control of virtual inertia and damping coefficient. By changing the weighting coefficient, it meets the different requirements of different systems for inertia and damping, thereby improving the accuracy of generating system inertia and damping targets.

[0063] 2. The present invention provides a multi-parameter cooperative adaptive control method for virtual synchronous generators based on reinforcement learning, comprising: designing appropriate state space, action space and reward function to achieve cooperative adaptive generation of inertia and damping targets. Through simulation comparison analysis, it is found that this method can change the system's emphasis on dynamic response capability and stability, and effectively improve the system's dynamic response capability and stability. Attached Figure Description

[0064] Figure 1 This is a flowchart of the virtual synchronous generator cooperative adaptive control method of the present invention;

[0065] Figure 2 This is a control block diagram of the virtual synchronous generator in this invention;

[0066] Figure 3 This is a comparison chart of different weight coefficients in a simulation example of the virtual synchronous generator cooperative adaptive control method of the present invention;

[0067] Figure 4 This is a comparison chart of traditional virtual synchronous machine control, virtual synchronous machine virtual inertia and damping coefficient adaptive control, and reinforcement learning-based virtual synchronous machine virtual inertia and damping coefficient cooperative adaptive control in the simulation examples of this invention. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All non-innovative embodiments implemented by other researchers in the art based on these embodiments are within the scope of protection of this invention. Furthermore, the step numbers in the embodiments of this invention are only for ease of explanation and do not limit the order of the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0069] like Figure 1 As shown, the present invention provides a cooperative adaptive control method for virtual synchronous generators based on reinforcement learning, comprising the following steps:

[0070] Step 1: Construct a mathematical model of the virtual synchronous engine;

[0071] Step 2: Through small-signal model analysis, determine the influence of virtual inertia and damping coefficient on the characteristics of virtual synchronous generator, and propose a method for selecting rated virtual inertia and rated damping coefficient;

[0072] Step 3: Based on reinforcement learning algorithms, design a collaborative adaptive target generation algorithm for virtual inertia and damping coefficients to generate optimal values ​​of inertia and damping under different states.

[0073] In one embodiment of the present invention, step 1, constructing a mathematical model of a virtual synchronous generator, includes the following sub-steps:

[0074] Step 1.1: Correspondence between the virtual synchronous generator model and the synchronous generator model

[0075] The mechanical equations of the virtual synchronous generator are as follows:

[0076]

[0077] In the formula, J is the moment of inertia; D is the damping coefficient; ω0 is the angular velocity of the power grid synchronization; T m T e and T D Mechanical torque, electromagnetic torque, and damping torque are respectively represented; δ is the power angle of the synchronous generator. Where T... e The calculation formula is:

[0078]

[0079] Step 1.2: Active Power Regulation

[0080] Through mechanical torque T m The active power output of the virtual synchronous generator is controlled by adjusting the parameters, and the expression is:

[0081] T m =T ref +△T;

[0082] In the formula, T ref ΔT represents the rated mechanical torque, and ΔT represents the mechanical torque deviation. ref The expressions for ΔT are:

[0083]

[0084] In the formula: P ref K is the rated active power. ω This is the adjustment coefficient.

[0085] Step 1.3: Reactive Power Adjustment

[0086] The reactive power output is adjusted by regulating the potential E, and its expression is as follows:

[0087] E = E0 + K q (Q ref -Q)+K u (U ref -U);

[0088] In the formula: E0 is the no-load potential; U ref U and U are the rated and actual values ​​of the virtual synchronous machine terminal voltage, respectively; k q Q is the reactive power regulation coefficient. ref Rated reactive power; Q is instantaneous reactive power, which can be expressed as:

[0089]

[0090] Therefore, the potential-voltage vector can be obtained as follows:

[0091]

[0092] in, This represents the phase of the virtual synchronous generator.

[0093] Furthermore, step 2, the small-signal model analysis, includes the following three sub-steps:

[0094] Step 2.1: Construct the second-order transfer function

[0095] The transfer function for constructing the virtual synchronous generator is shown below:

[0096]

[0097] Step 2.2: Method for Selecting Rated Virtual Inertia

[0098] The natural oscillation frequency is determined with reference to a synchronous generator. In this embodiment, the natural oscillation frequency is 0.628-15.7 rad / s. The rated virtual inertia J0 is obtained using the following formula:

[0099]

[0100] In the formula, ω n ξ is the natural oscillation angular frequency; ξ is the damping coefficient.

[0101] Step 2.3: Method for Selecting Rated Damping Coefficient

[0102] Specifically, in this embodiment, ξ = 0.707 is taken, and the rated damping coefficient D0 is solved using the above formula.

[0103] Furthermore, step 3, the virtual inertia and damping coefficient collaborative adaptive target generation algorithm, includes the following sub-steps:

[0104] Step 3.1: Set weighting coefficients

[0105] Based on the intelligent decision-making capabilities of reinforcement learning algorithms, the state spaces of virtual inertia controllers and damping coefficient controllers are designed. Weight coefficients are used to balance the system's emphasis on dynamic response and stability. The state spaces of the two controllers are as follows:

[0106]

[0107] In the formula, S J S is the state space of the virtual inertia controller. D Let be the state space of the damping coefficient controller, and 'a' be the weighting coefficient, where 'a' ∈ [0.1, 1]. When 'a' = 0.1, the system prioritizes stronger response capability, and when 'a' = 1, the system prioritizes stability.

[0108] Step 3.2: Inertia Target Generation

[0109] In this embodiment, S is set J The discrete set is (0,1)Hz, [1,3)Hz, [3,5)Hz, [5,8)Hz, [8,10)Hz, [10,15)Hz, [15,+∞)Hz; the action set A of the controller is... J The output of the controller should be the real-time set of scheduling instructions ΔJ, and the action space can be defined as {0, 0.3, 0.6, 0.9, 1.2, 1.5, 1.8, 2.1, 2.3, 2.5, 2.8}.

[0110] The reward and punishment function is set as follows:

[0111]

[0112] In the formula: r J μ1 represents the frequency reward; μ2, μ3, μ4, μ5, and μ6 are the weights corresponding to the reward function of each control region.

[0113] In this embodiment, μ1, μ2, μ3, μ4, μ5 and μ6 are respectively taken as -3, -8, -12, -15, -20 and -25.

[0114] Therefore, the inertia target is as follows:

[0115] J target =J0+△J

[0116] Step 3.3: Damping Target Generation

[0117] In this embodiment, S is set D The discrete set is (-∞, -12), [-8, -12), [-5, -0.5), [-0.5, 0.5), [0.5, 8), [8, 12), [12, +∞); the action set A of the controller.D The output of the controller should be the real-time set of scheduling instructions ΔD, and the action space can be defined as {-10, -8, -5, -3, -1, 0, 1, 3, 5, 8, 10}.

[0118] The reward and punishment function is set as follows:

[0119]

[0120] In the formula: r D The frequency reward is δ1, δ2, δ3, δ4, δ5 and δ6 are the weights corresponding to the reward function of each control region, and δ1, δ2, δ3, δ4, δ5 and δ6 are taken as -3, -8, -12, -15, -20 and -25 respectively.

[0121] Therefore, the damping target is as follows:

[0122] D target =D0 + △D;

[0123] In the formula, ΔD is the damping coefficient controller output value, D0 is the rated damping coefficient, and D target This represents the target value for the damping coefficient of the virtual synchronous generator.

[0124] Furthermore, the accuracy of the target generation algorithm in this embodiment is verified using the following simulation experiments.

[0125] First, based on step 1, a virtual synchronous generator model is established to determine the active power regulation and reactive power regulation schemes. Based on reinforcement learning control, the basic control structure is as follows: Figure 2 As shown, it is divided into active power regulation, reactive power regulation, and signal generation.

[0126] Then, based on the model, the values ​​of the rated virtual inertia and rated damping coefficient are determined using small-signal analysis. Specifically, in this embodiment, the rated virtual inertia is selected as 0.2 kg·m. 2 The rated damping coefficient is 10 N·m·s·rad. -1 .

[0127] Finally, based on step 3, the state space, action space, and reward function of the virtual inertia controller and damping coefficient controller are designed, and the target values ​​of inertia and damping are generated by the deviation of the system frequency and its rate of change.

[0128] Adjusting the value of the weighting coefficient 'a' while keeping other parameters constant, the result is as follows: Figure 3 As shown.

[0129] It can be observed that when a = 0.1, the system prioritizes a faster response speed and has the shortest settling time, but also the largest overshoot. When a = 0.5, the system prioritizes both stability and response speed, resulting in both a short settling time and a small overshoot. When a = 1, the system prioritizes better stability, resulting in the smallest overshoot, but also a longer settling time.

[0130] When the system encounters a strong load disturbance, the load active power increases from the initial 2kW to 12kW. At this point, by comparing the target values ​​of system inertia and damping generated by traditional virtual synchronous machine control and virtual synchronous machine generator parameter adaptive control, the system active power and frequency under these three different control methods are compared. Figure 4 As shown.

[0131] Simulation results demonstrate that the target generation algorithm designed in this invention can change the system's emphasis on stability and response speed by altering the weight coefficients. Furthermore, the comparative results also verify the superiority of the method designed in this paper.

[0132] In summary, the multi-parameter cooperative adaptive control method for virtual synchronous generators based on reinforcement learning designed in this invention enables the grid-connected inverter to cooperatively and adaptively generate inertia and damping when disturbances occur, which helps to improve the grid-connected stability and dynamic response speed of new energy systems.

[0133] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A cooperative adaptive control method for virtual synchronous generators based on reinforcement learning, comprising the following steps: Step 1: Construct a mathematical model of the virtual synchronous engine; Step 2: Through small-signal model analysis, determine the influence of virtual inertia and damping coefficient on the characteristics of virtual synchronous generator, and propose a method for selecting rated virtual inertia and rated damping coefficient; Step 3: Based on reinforcement learning algorithms, design a collaborative adaptive target generation algorithm for virtual inertia and damping coefficients to generate optimal values ​​of inertia and damping under different states; Step 3, the virtual inertia and damping coefficient collaborative adaptive target generation algorithm, includes the following sub-steps: Step 3.1: Set weight coefficients: Based on the intelligent decision-making capability of reinforcement learning algorithm, design the state space of virtual inertia controller and damping coefficient controller, and use weight coefficients to balance the system's emphasis on dynamic response capability and stability. The state spaces of the virtual inertia controller and the damping coefficient controller are: （10） In the formula, For rotational inertia, The damping coefficient is... Let ω be the angular velocity of the virtual synchronous generator, and dω and dt be the differentials of the angular velocity and time of the virtual synchronous generator, respectively. For angular velocity deviation; S J S is the state space of the virtual inertia controller. D Let be the state space of the damping coefficient controller, and 'a' be the weighting coefficient. [0.1,1]; When a=0.1, the system prioritizes stronger response capability; when a=1, the system prioritizes stronger stability. Step 3.2, Inertia Target Generation: Design a suitable action space and reward function for the state space of the virtual inertia controller, so as to select appropriate actions in different state spaces, and generate an inertia target by combining the rated virtual inertia mentioned above; Step 3.3, Damping Target Generation: Design a suitable action space and reward function for the state space of the damping coefficient controller, so as to select appropriate actions in different state spaces, and generate the damping target by combining the rated virtual inertia mentioned above.

2. The virtual synchronous generator cooperative adaptive control method based on reinforcement learning according to claim 1, characterized in that, Step 1, which involves constructing the mathematical model of the virtual synchronous generator, includes the following sub-steps: Step 1.1: Determine the correspondence between the virtual synchronous generator model and the synchronous generator model, and establish the mathematical model of the virtual synchronous generator; Step 1.2, based on the mechanical torque T m The adjustment is used to regulate the active power output of the virtual synchronous generator model, thereby implementing the inverter's active power regulation command. Step 1.3: Based on the adjustment of the excitation, the reactive power output of the virtual synchronous generator model is adjusted to obtain the potential voltage of the virtual synchronous generator.

3. The method of claim 2, wherein the method is characterized by, In step 1.1, the determination of the correspondence between the virtual synchronous generator model and the synchronous generator model... The mechanical equations of the virtual synchronous generator are as follows: （1） In the formula, ω0 is the angular velocity of the power grid synchronization, and T m T e and T D These represent the mechanical torque, electromagnetic torque, and damping torque, respectively; δ is the power angle of the virtual synchronous generator; and dδ is the differential of the power angle of the virtual synchronous generator. where T e The calculation formula is: （2） In the formula, To output electromagnetic power, , , These are the three-phase output potentials. , , These are the three-phase output currents.

4. The virtual synchronous generator cooperative adaptive control method based on reinforcement learning according to claim 3, characterized in that, The active power output adjustment method described in step 1.2 is achieved by adjusting the mechanical torque T. m The active power output of the virtual synchronous generator is controlled by adjusting the parameters, as shown in the following expression: （3） In the formula, T ref The rated mechanical torque is T, where ΔT is the mechanical torque deviation. ref The expressions for ΔT are: （4） In the formula, P ref K is the rated active power. ω This is the adjustment coefficient; The reactive power output adjustment described in step 1.3 is achieved by adjusting the potential E to regulate the reactive power output, as shown in the following formula: （5） In the formula, E0 is the no-load potential, U ref U and U are the rated and actual values ​​of the virtual synchronous machine terminal voltage, respectively; Q is the reactive power regulation coefficient. ref Q is the rated reactive power; Q is the instantaneous reactive power, expressed as: （6） In the formula, , , These are the three-phase node voltages; Therefore, the potential-voltage vector of the virtual synchronous generator can be obtained as follows: （7） wherein, is the phase of the virtual synchronous generator.

5. The method of claim 4, wherein the method is characterized by, Step 2, the small signal model analysis, includes the following sub-steps: Step 2.1: Construct a second-order transfer function based on the small-signal model analysis method and the model of a virtual synchronous generator; Step 2.2: Calculate the rated virtual inertia based on the transfer function; Step 2.3: Calculate the rated damping coefficient based on the transfer function.

6. The virtual synchronous generator cooperative adaptive control method based on reinforcement learning according to claim 5, characterized in that, The second-order transfer function mentioned in step 2.1 is as follows: （8） Where E is the voltage amplitude of the virtual synchronous generator, U is the terminal voltage of the virtual synchronous generator, Z is the impedance of the filter circuit of the virtual synchronous generator, and s is the complex variable in the transfer function. It is a second-order transfer function. The function is the output and input of a second-order transfer function.

7. The method of claim 6, wherein the method is characterized by, The natural oscillation frequency is determined by referring to the synchronous generator, and the rated virtual inertia J0 mentioned in step 2.2 and the rated damping coefficient D0 mentioned in step 2.3 are obtained by the following formula (9); （9） In the formula, ω n Let J be the natural oscillation angular frequency, and the J obtained by solving is the rated virtual inertia J0; ξ is the damping coefficient. Based on the preset ξ value, D obtained by solving according to the above formula (9) is the rated damping coefficient D0 mentioned in step 2.

3.

8. The virtual synchronous generator coordinated adaptive control method based on reinforcement learning according to claim 1, characterized in that, The inertia target generation method described in step 3.2 is as follows: Set the state space S of the virtual inertia controller. J The discrete set, and the action set A of the controller. J The A J The output of the controller is the real-time set of scheduling instructions ∆J, defined as A. J Action space; The current state space of the virtual inertia controller is determined, and the action that yields the maximum reward value in the current state space is selected using the reward function. The inertia target is as follows: （11） In the formula, For rated inertia, For the output value of the virtual inertia controller, The target value for the virtual inertia of the virtual synchronous generator; The damping target generation method described in step 3.3 is as follows: Set the state space S of the damping coefficient controller. D The discrete set, and the action set A of the controller. D The A D The output of the controller is the real-time set of scheduling instructions ∆D, defined as A. D The action space; Determine the current state space of the damping coefficient controller, and select the action that yields the maximum reward value in the current state space using the reward function. The damping objective is as follows: （12） In the formula, This is the output value of the damping coefficient controller. The rated damping coefficient, This represents the target value for the damping coefficient of the virtual synchronous generator.

Images