A microgrid VSG control method based on fractional-order adaptation

Abstract

The present invention provides a microgrid VSG control method based on fractional order adaptation, which comprises the following steps: S101, establishing a circuit topology structure of a voltage sag generator (VSG); S102, modeling a virtual speed regulator and a virtual excitation system of the VSG; S103, determining an initial value of the moment of inertia and an initial value of the damping coefficient of the virtual speed regulator through a machine learning algorithm; S104, designing an adaptive rate based on the initial value of the moment of inertia and the initial value of the damping coefficient of the virtual speed regulator, and generating a VSG control strategy based on fractional order adaptation. The method aims at the moment of inertia and the damping coefficient in the VSG system, and by constructing a fractional order model, establishes a nonlinear relationship between the rate of change of angular frequency and the moment of inertia, the amount of change of angular frequency and the damping coefficient, so as to adaptively adjust the values ​​of the moment of inertia and the damping coefficient in the system, thereby reducing the overshoot of the system and improving the dynamic performance.

Description

Technical Field

[0001] The present invention relates to the technical field of microgrid operation control strategies, and in particular to a microgrid VSG control method based on fractional-order adaptation. Background Art

[0002] my country is rich in marine resources and has many islands, such as solar energy, wind energy and other new energy resources. Therefore, in the tropical island environment, we should make rational use of such clean energy and improve its utilization rate in distributed power generation systems. In addition, the island environment has the disadvantages of serious environmental pollution and bad climate, which will make the reliability and stability of the power supply system worse, which requires the system to have sufficient anti-interference ability. Distributed power sources are often connected to the power grid through power electronic devices, which do not have rotational inertia and damping coefficients. As their proportion in the power grid increases, the corresponding inertia and damping of the system will also decrease. When the system fails, due to the lack of sufficient inertia and damping, the system will usually become unstable or even collapse, which does not meet the conditions that the island microgrid should meet. Therefore, in the island microgrid, the virtual synchronous generator VSG technology can be introduced to provide the system with sufficient inertia and damping coefficients, so that the system has a certain anti-interference ability, and a large number of micro sources can be connected to the grid through inverters in actual use. When multiple micro sources are connected in parallel, in the VSG parallel system, the damping coefficient and rotational inertia will affect the angular velocity when the power fluctuates. Due to the mutual coupling of the two VSGs, when their respective angular frequency increments have a certain deviation, their active power increments will change, thus causing system oscillation.

[0003] In the VSG control system, due to its nonlinearity, variable parameters and susceptibility to external interference, the system controller cannot be simply set as a proportional-integral controller. The traditional PI control algorithm is a method of controlling deviations based on feedback. When external interference occurs, although PI control can ultimately achieve error control of the controlled quantity, it is usually accompanied by large overshoot and oscillation. In order to suppress system oscillations, the present invention designs a microgrid VSG control method based on fractional-order adaptation. This control method improves the system's fit in different scenarios, and can also achieve good optimization effects in complex application scenarios or for complex controlled objects. Summary of the invention

[0004] The purpose of the present invention is to provide a microgrid VSG control method based on fractional-order adaptation to solve the problem of active power oscillation caused by parallel coupling of multiple VSGs. The present invention targets the moment of inertia and damping coefficient in the VSG system, and by constructing a fractional-order model, establishes a nonlinear relationship between the rate of change of angular frequency and the moment of inertia, the change of angular frequency and the damping coefficient, so as to adaptively adjust the values ​​of the moment of inertia and the damping coefficient in the system, thereby reducing the overshoot of the system and improving the dynamic performance.

[0005] To achieve the above-mentioned object of the invention, the present invention provides a microgrid VSG control method based on fractional-order adaptation, the method comprising the following steps:

[0006] S101, establishing a circuit topology structure of a voltage sag generator (VSG);

[0007] S102, modeling a virtual speed regulator and a virtual excitation system of the VSG;

[0008] S103, determining an initial value of the moment of inertia and an initial value of the damping coefficient of the virtual speed regulator through a machine learning algorithm;

[0009] S104. Based on the initial values ​​of the moment of inertia and the damping coefficient of the virtual speed regulator, an adaptive rate is designed to generate a VSG control strategy based on fractional-order adaptation.

[0010] Furthermore, the virtual speed regulator and virtual excitation system of VSG are modeled as follows:

[0011] In the virtual speed regulator, the angle expression of VSG is:

[0012]

[0013] Among them, P ref is the reference rated power, P e is the electromagnetic power, ω n is the rated angular frequency, J is the moment of inertia, and D is the damping coefficient;

[0014] The virtual excitation system of VSG is expressed as:

[0015]

[0016] Where, E is the output voltage amplitude of the inverter bridge arm of VSG; K q is the voltage regulation coefficient of VSG, which reflects the ability of the virtual excitation controller to regulate voltage; Q ref Indicates the setting value of the reference reactive power of VSG; Q indicates the reactive power actually output by VSG; D qIt represents the reactive power regulation coefficient, i.e. the reactive power and voltage droop coefficient; U represents the actual voltage amplitude output by VSG; U ref Indicates the amplitude of the reference voltage output by VSG.

[0017] Furthermore, in step S3, the initial value of the moment of inertia J0 and the initial value of the damping coefficient D0 in the VSG are determined by using a reinforcement learning Q-learning algorithm.

[0018] Furthermore, the initial value of the moment of inertia J0 and the initial value of the damping coefficient D0 in the VSG are determined by the reinforcement learning Q-learning algorithm, which specifically includes the following steps:

[0019] S201, input the number of iterations T, action space A, state space S, exploration rate ∈, decay coefficient γ and update step size α, output action value function Q(s,a), where s is the state and a is the action;

[0020] S202, randomly initialize the action value function Q(s,a), and set the Q value of the terminal state to 0;

[0021] S203, initialize s to the first state of the current sequence;

[0022] S204, using the ∈-greedy method to select the corresponding action a from the state s;

[0023] S205, the agent interacts with the environment, executes the selected action a, and obtains a new state s′ and a corresponding reward R;

[0024] S206. Update the action value function Q(s,a) using the temporal difference method:

[0025] Q(s,a)=(s,a)+[R t +γ*max α Q(s′,α)-Q(s,a)];

[0026] S207, update the state, set s = s';

[0027] S208, if s′ satisfies the stop condition, the current round of iteration is completed, otherwise go to step S204;

[0028] S209. When the action value function Q(s,a) converges, for different VSG systems, select the corresponding action a under different initial value states s to determine the initial value of the moment of inertia and the initial value of the damping coefficient of the VSG system.

[0029] Furthermore, when the initial value of the moment of inertia J0 and the initial value of the damping coefficient D0 in the VSG are determined by the reinforcement learning Q-learning algorithm, the state set of the reinforcement learning Q-learning algorithm is set to the overshoot σ% of the system, and the action set is set to adjust the value of the moment of inertia J and the damping coefficient D, and the reward function is set as shown in the following formula:

[0030]

[0031] Where R t is the reward function, σ% is the overshoot, t r is the rise time.

[0032] Furthermore, the adaptation rate is designed to generate a VSG control strategy based on fractional-order adaptation, which specifically includes the following steps:

[0033] S301, analyzing the rotor angular frequency oscillation curve diagram, and dividing the oscillation period into four time periods: t1 to t2, t2 to t3, t3 to t4, and t4 to t5;

[0034] S302, analyzing the oscillation process in four time periods, and constructing functional relationships between the moment of inertia and the angular frequency change rate, and the damping coefficient and the angular frequency increment in the corresponding time periods;

[0035] S303 , constructing adaptive formulas of the damping coefficient D and the moment of inertia J based on fractional order respectively according to the functional relationship constructed in the previous step.

[0036] Furthermore, the adaptive formula of the damping coefficient D is as follows:

[0037]

[0038] The adaptive formula for the moment of inertia J is as follows:

[0039]

[0040] Among them, J0 and D0 are the moment of inertia and damping coefficient of the system in steady state respectively; K J and K D They represent the adaptive coefficients of the moment of inertia and the damping coefficient respectively; N1 is the adjustment threshold of the angular frequency change, and N2 is the threshold of the angular frequency deviation.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] The present invention provides a microgrid VSG control method based on fractional order adaptation. Compared with the VSG parallel system with fixed parameters, it can eliminate the coupling between systems to a certain extent, reduce the oscillation of active power, and improve the dynamic performance of the system. The method determines the initial values ​​of the moment of inertia and damping coefficient in the adaptive control by the reinforcement learning Q-lerning algorithm, and sets a unique reward function, state set and action set in the algorithm, so that the values ​​of J0 and D0 in different VSG control systems are no longer a fixed value, and the setting of the initial value is more in line with the characteristics of the system itself, which improves the accuracy and rapidity of the adaptive control. Compared with the integer-order adaptive control, the method improves the steady-state and dynamic performance of the parallel system and improves the control accuracy. The controller developed and designed using the fractional-order system theory has better robustness and control accuracy than the controller designed using the traditional integer-order system theory. In addition, the fractional-order adaptive controller is more in line with the actual control system in the current industry and has better implementation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 It is a schematic diagram of the overall flow of a microgrid VSG control method based on fractional-order adaptation provided by an embodiment of the present invention.

[0045] Figure 2 It is a circuit topology diagram of a parallel VSG system provided in an embodiment of the present invention.

[0046] Figure 3 It is a rotor angular frequency oscillation curve diagram provided by an embodiment of the present invention.

[0047] Figure 4 It is an adaptive adjustment curve diagram of the damping coefficient D provided in an embodiment of the present invention.

[0048] Figure 5 4 is an adaptive adjustment curve diagram of the moment of inertia J provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The enumerated embodiments are only used to explain the present invention but not to limit the scope of the present invention.

[0050] Reference Figure 1This embodiment provides a microgrid VSG control method based on fractional-order adaptation, and the method includes the following steps:

[0051] S101. Establish the circuit topology of the voltage sag generator (VSG). Figure 2 The essence of VSG control is to embed the VSG control algorithm into the control algorithm of the inverter, and its core lies in the design of the virtual speed regulator and virtual excitation controller. In this circuit, the distributed power supply and energy storage are regarded as the energy supply on the DC side, and a large capacitor for voltage stabilization is connected in parallel on the DC side, and then connected to the system through the inverter.

[0052] S102. Modeling a virtual speed regulator and a virtual excitation system of the VSG.

[0053] This embodiment designs the virtual speed regulator and virtual excitation system of VSG by analogy with the rotor motion equation and excitation system of synchronous generator. The virtual speed regulator is the core of VGS control. It introduces the moment of inertia j and the damping coefficient D to make the inverter present inertia to the outside, thereby improving the anti-interference performance of the system. There is no actual excitation regulator and excitation power unit in the virtual excitation system. It only regulates reactive power and voltage through the externality of the synchronous generator excitation system.

[0054] In the virtual speed regulator, the angle expression of VSG is:

[0055]

[0056] Among them, P ref is the reference rated power, P e is the electromagnetic power, ω n is the rated angular frequency, J is the moment of inertia, and D is the damping coefficient.

[0057] The virtual excitation system of VSG is expressed as:

[0058]

[0059] Where, E is the output voltage amplitude of the inverter bridge arm of VSG; K q is the voltage regulation coefficient of VSG, which reflects the ability of the virtual excitation controller to regulate voltage; Q ref Indicates the setting value of the reference reactive power of VSG; Q indicates the reactive power actually output by VSG; D q It represents the reactive power regulation coefficient, i.e. the reactive power and voltage droop coefficient; U represents the actual voltage amplitude output by VSG; U ref Indicates the amplitude of the reference voltage output by VSG.

[0060] S103. Determine the initial value of the moment of inertia and the initial value of the damping coefficient of the virtual speed regulator through a machine learning algorithm.

[0061] This embodiment adopts the reinforcement learning Q-learning algorithm to determine the initial value of the moment of inertia J0 and the initial value of the damping coefficient D0 in the VSG, which specifically includes the following steps:

[0062] S201, input the number of iterations T, action space A, state space S, exploration rate ∈, decay coefficient γ and update step size α, output action value function Q(s,a), where s is the state and a is the action;

[0063] S202, randomly initialize the action value function Q(s,a), and set the Q value of the terminal state to 0;

[0064] S203, initialize s to the first state of the current sequence;

[0065] S204, using the ∈-greedy method to select the corresponding action a from the state s;

[0066] S205, the agent interacts with the environment, executes the selected action a, and obtains a new state s′ and a corresponding reward R;

[0067] S206. Update the action value function Q(s,a) using the temporal difference method:

[0068] Q(s,a)=Q(s,a)+[R t +γ*max α Q(s′,α)-Q(s,a)];

[0069] S207, update the state, set s = s';

[0070] S208, if s′ satisfies the stop condition, the current round of iteration is completed, otherwise go to step S204;

[0071] S209. When the network training is completed, that is, the action value function Q(s,a) converges, for different VSG systems, the corresponding action a is selected under different initial value states s to determine the initial value of the moment of inertia and the initial value of the damping coefficient of the VSG system. The initial value of the moment of inertia and the initial value of the damping coefficient are the premise and basis for setting the adaptive rate, and the quality of their setting will directly affect the dynamic performance of the system.

[0072] In the above algorithm, the core is the design and selection of state set, action set and reward function. In VSG control, the state set selects the overshoot σ% of the system, the action set is set to adjust the value of the moment of inertia J and the damping coefficient D, and the reward function is set as shown below:

[0073]

[0074] Where R t is the reward function, σ% is the overshoot, t r is the rise time.

[0075] S104. Based on the initial values ​​of the moment of inertia and the damping coefficient of the virtual speed regulator, an adaptive rate is designed to generate a VSG control strategy based on fractional-order adaptation.

[0076] In the VSG system, the moment of inertia and the damping coefficient will have a certain impact on the dynamic performance of the system. When the damping coefficient is constant, the larger the moment of inertia, the smaller the overshoot of the system. However, when the moment of inertia is too large, the system will lose stability. When the moment of inertia is constant, the larger the set damping coefficient, the smaller the overshoot of the system. In the VSG control strategy based on fractional-order adaptation, the moment of inertia is adjusted by the rate of change of the angular frequency, and the damping coefficient is adjusted by the change of the angular frequency, so as to suppress the active oscillation of the system.

[0077] In this embodiment, an adaptive rate is designed to generate a VSG control strategy based on fractional-order adaptation, which specifically includes the following steps:

[0078] S301, analyze the rotor angular frequency oscillation curve diagram, and divide the oscillation period into four time periods: t1 to t2, t2 to t3, t3 to t4, and t4 to t5. Figure 3 , in each time period, the angular frequency increment and the rate of change of the angular frequency are different. By analyzing the oscillation process, the functional relationship between the moment of inertia and the rate of change of the angular frequency, and the damping coefficient and the angular frequency increment can be constructed respectively.

[0079] S302, analyzing the oscillation process in four time periods, and constructing functional relationships between the moment of inertia and the angular frequency change rate, and the damping coefficient and the angular frequency increment in the corresponding time periods.

[0080] At time t1, due to the increase in system load, the active power outside the system increases, causing oscillation of the angular frequency.

[0081] In the time period t1-t2, the deviation of the system's angular velocity Δω is greater than 0, and the rate of change of the angular velocity dω / dt is also greater than 0, which is equivalent to a positive acceleration motion with reduced acceleration. At this time, the moment of inertia should be increased to reduce the overshoot, so as to prevent the angular frequency offset from being too large and causing the system to become unstable. At the same time, in order to reduce the rate of change of the angular frequency deviation, the damping coefficient should be reduced so that the system reaches the peak faster.

[0082] In the time period t2-t3, the deviation of the angular velocity of the system Δω is greater than 0, and the rate of change of the angular velocity dω / dt is less than 0, which is equivalent to a positive deceleration motion with increasing acceleration. In order to suppress oscillation and make the attenuation of the angular frequency reach stability as soon as possible, the damping coefficient should be appropriately increased at this time. At the same time, if a smaller moment of inertia is given to the system in this positive deceleration stage, this will cause the rotor to return to a stable value with a larger acceleration.

[0083] In the time period t3-t4, the deviation of the angular velocity of the system Δω is less than 0, and the rate of change of the angular velocity dω / dt is less than 0, which is equivalent to a reverse acceleration motion with reduced acceleration. At this time, similar to the time period t1-t2, the moment of inertia should be increased to prevent the angular frequency offset from being too large, causing the system to lose stability. At the same time, in order to reduce the rate of change of the degree deviation, the damping coefficient should be reduced.

[0084] In the time period t4-t5, the deviation of the angular velocity of the system Δω is less than 0, and the rate of change of the angular velocity dω / dt is greater than 0, which is equivalent to a reverse deceleration motion with increasing acceleration. At this time, similar to the time period t2-t3, the moment of inertia should be reduced and the damping coefficient should be increased.

[0085] According to the above analysis, the adaptive law that the moment of inertia J and the damping coefficient D should follow can be obtained as shown in Table 1:

[0086] Table 1 Setting of adaptive rate

[0087]

[0088]

[0089] S303 , constructing adaptive formulas of the damping coefficient D and the moment of inertia J based on fractional order respectively according to the functional relationship constructed in the previous step.

[0090] Reference Figure 4 , Figure 5 , the adaptive formula of the damping coefficient D is as follows:

[0091]

[0092] The adaptive formula for the moment of inertia J is as follows:

[0093]

[0094] Among them, J0 and D0 are the moment of inertia and damping coefficient of the system in steady state respectively; K J and K D They represent the adaptive coefficients of the moment of inertia and the damping coefficient respectively; N1 is the adjustment threshold of the angular frequency change, and N2 is the threshold of the angular frequency deviation.

[0095] In this embodiment, the moment of inertia is first adjusted by the rate of change of the angular frequency change, and the damping coefficient is adjusted by the change of the angular frequency. The N1 and N2 thresholds are used to determine whether to perform adaptive adjustment of the moment of inertia and the damping coefficient. If the threshold is set too small, it will be detrimental to the normal operation of the VSG, but if it is set too large, the accuracy of the adaptive adjustment will be affected. When the rate of change of the angular frequency is relatively small, the adaptive adjustment of the moment of inertia is not performed; and when the change of the angular frequency is relatively small, the adaptive adjustment of the damping coefficient is not performed.

[0096] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A microgrid VSG control method based on fractional order adaptation, characterized in that: The method comprises the following steps: S101, establishing a circuit topology structure of a voltage sag generator (VSG); S102, modeling a virtual speed regulator and a virtual excitation system of the VSG; S103, determining an initial value of the moment of inertia and an initial value of the damping coefficient of the virtual speed regulator through a machine learning algorithm; S104, based on the initial value of the moment of inertia and the initial value of the damping coefficient of the virtual speed regulator, design an adaptive rate and generate a VSG control strategy based on fractional-order adaptation; In step S103, the initial value of the moment of inertia J0 and the initial value of the damping coefficient D0 in the VSG are determined by using a reinforcement learning Q-learning algorithm; The initial value of the moment of inertia J0 and the initial value of the damping coefficient D0 in the VSG are determined by the reinforcement learning Q-learning algorithm, which specifically includes the following steps: S201, input the number of iterations T, action space A, state space S, exploration rate ∈, decay coefficient γ and update step size α, output action value function Q(s,a), where s is the state and a is the action; S202, randomly initialize the action value function Q(s,a), and set the Q value of the terminal state to 0; S203, initialize s to the first state of the current sequence; S204, using the ∈-greedy method to select the corresponding action a from the state s; S205, the agent interacts with the environment, executes the selected action a, and obtains a new state s' and a corresponding reward R; S206. Update the action value function Q(s,a) using the temporal difference method: S207, update the state, set s = s'; S208, if s' satisfies the stop condition, the current round of iteration is completed, otherwise go to step S204; S209. When the action value function Q(s,a) converges, for different VSG systems, select the corresponding action a under different initial value states s to determine the initial value of the moment of inertia and the initial value of the damping coefficient of the VSG system.

2. A fractional-order adaptive microgrid VSG control method according to claim 1, characterized in that: The virtual speed regulator and virtual excitation system of VSG are modeled as follows: In the virtual speed regulator, the angle expression of VSG is: Among them, P ref is the reference rated power, P e is the electromagnetic power, ω n is the rated angular frequency, J is the moment of inertia, and D is the damping coefficient; The virtual excitation system of VSG is expressed as: Where, E is the output voltage amplitude of the inverter bridge arm of VSG; K q is the voltage regulation coefficient of VSG, which reflects the ability of the virtual excitation controller to regulate voltage; Q ref Indicates the setting value of the reference reactive power of VSG; Q indicates the reactive power actually output by VSG; D q It represents the reactive power regulation coefficient, i.e. the reactive power and voltage droop coefficient; U represents the actual voltage amplitude output by VSG; U ref Indicates the amplitude of the reference voltage output by VSG.

3. A microgrid VSG control method based on fractional order adaptation according to claim 1, characterized in that: When the initial value of the moment of inertia J0 and the initial value of the damping coefficient D0 in the VSG are determined by the reinforcement learning Q-learning algorithm, the state set of the reinforcement learning Q-learning algorithm is set to the overshoot σ% of the system, and the action set is set to adjust the value of the moment of inertia J and the damping coefficient D, and the reward function is set as shown in the following formula: Where R t is the reward function, σ% is the overshoot, t r is the rise time.

4. A fractional-order adaptive microgrid VSG control method according to claim 1, characterized in that: Design the adaptive rate and generate the VSG control strategy based on fractional-order adaptation, which specifically includes the following steps: S301, analyzing the rotor angular frequency oscillation curve diagram, and dividing the oscillation period into four time periods: t1 to t2, t2 to t3, t3 to t4, and t4 to t5; S302, analyzing the oscillation process in four time periods, and constructing functional relationships between the moment of inertia and the angular frequency change rate, and the damping coefficient and the angular frequency increment in the corresponding time periods; S303 , constructing adaptive formulas of the damping coefficient D and the moment of inertia J based on fractional order respectively according to the functional relationship constructed in the previous step.

5. A fractional-order adaptive microgrid VSG control method according to claim 4, characterized in that: The adaptive formula of the damping coefficient D is as follows: The adaptive formula for the moment of inertia J is as follows: Among them, J0 and D0 are the moment of inertia and damping coefficient of the system in steady state respectively; K J and K D They represent the adaptive coefficients of the moment of inertia and the damping coefficient respectively; N1 is the adjustment threshold of the angular frequency change, and N2 is the threshold of the angular frequency deviation.

Images