Inertia and damping adaptive VSG control method based on neural network optimization

Abstract

The invention discloses an inertia and damping adaptive VSG control method based on neural network optimization. The method comprises the steps that S1, a VSG rotor motion equation is obtained through equivalent conversion; s2, introducing a speed regulator into the active power-frequency control loop to obtain a VSG reactive power-voltage control equation; s3, performing circuit model equivalence on the VSG grid-connected system to obtain an output active power equation, and obtaining a relation between inertia and damping; s4, analyzing power angle characteristics and angular frequency oscillation under a disturbance condition, and summarizing an inertia self-adaptive mechanism; and S5, establishing an artificial neural network, analyzing and predicting frequency and voltage changes of the power grid, and adjusting weight values to realize self-adaptive adjustment of inertia and damping. According to the inertia and damping adaptive VSG control method based on neural network optimization provided by the invention, the inertia and damping of the system can be cooperatively adjusted, the frequency is stabilized, the overshoot is reduced, and the method has faster response time in the aspect of recovering to a stable state.

Description

Technical Field

[0001] The present invention belongs to the technical field of virtual synchronous generator control, and in particular relates to an inertia and damping adaptive VSG control method based on neural network optimization. Background Art

[0002] As the energy structure transforms towards low-carbon and clean energy, the proportion of renewable energy generation in the power system continues to increase, which gradually reduces the inertia of the power system. Virtual Synchronous Generator (VSG) is an inverter-based power conversion system that simulates the operating characteristics of traditional synchronous generators and is used to achieve the interconnection of distributed power resources and power networks. The virtual synchronous generator reduces the frequency deviation by adaptively controlling the moment of inertia and damping coefficient, allowing the system to return to a stable state faster and improve the frequency response of the power system.

[0003] The main technical difficulties faced by adaptive control at present include the following aspects:

[0004] (1) Control algorithm complexity

[0005] The design of adaptive control algorithms is often complex and involves a lot of calculations and optimization processes. How to reduce algorithm complexity, reduce computing resource consumption, and improve real-time performance while ensuring control effects is a problem that needs to be solved.

[0006] (2) Dynamic adaptability and accuracy

[0007] In a complex and ever-changing power grid environment, how to ensure that the adaptive control strategy can respond quickly and accurately to changes in the power grid state, especially when the frequency fluctuates greatly or the system suffers severe disturbances, and maintain high-precision inertia and damping coefficient adjustment remains a major challenge.

[0008] (3) Stability and robustness

[0009] When designing an adaptive control strategy, it is very important to ensure system stability and improve robustness to model uncertainty, external disturbances, and parameter changes. How to work stably and reliably under various operating conditions is a major challenge. Summary of the invention

[0010] The purpose of the present invention is to provide an inertia and damping adaptive VSG control method based on neural network optimization, which can coordinately adjust the inertia and damping of the system, effectively stabilize the frequency, and significantly reduce the overshoot. In addition, the method has a faster response time in returning to a stable state.

[0011] The technical solution adopted by the present invention is an inertia and damping adaptive VSG control method based on neural network optimization, and the specific steps are as follows:

[0012] S1. Based on the traditional two-order model of synchronous generator, the pole pair number of synchronous generator is assumed to be 1, and the VSG rotor motion equation is obtained through equivalent transformation;

[0013] S2. Introduce a speed regulator into the active power-frequency control loop, control the active power output of the generator by frequency modulation, and obtain the VSG reactive power-voltage control equation based on the VSG active frequency regulation equation;

[0014] S3. Establish a VSG power frequency small signal model, perform circuit model equivalence on the VSG grid-connected system, obtain the VSG output active power equation, combine the VSG rotor motion equation and the VSG active frequency regulation equation, obtain the approximate expression of the VSG active loop closed-loop transfer function, and then derive the functional relationship between inertia and damping;

[0015] S4. According to the functional relationship between inertia and damping, combined with the power angle characteristic curve and the angular frequency oscillation curve, the power angle characteristics and angular frequency oscillation of VSG under disturbance are comprehensively considered, the selection principle of inertia and damping is obtained, and the inertia adaptive mechanism based on power difference is summarized;

[0016] S5. Establish an artificial neural network to monitor and model the frequency and voltage changes of the power grid, adjust the weight value after evaluating the prediction results, and realize adaptive adjustment of inertia and damping.

[0017] The present invention is also characterized in that:

[0018] The specific process of S1 is:

[0019] The rotor motion equation of a traditional synchronous generator is based on Newton's second law, where torque equals mass times acceleration, and the equation is as follows:

[0020]

[0021] In the above formula, T m , T e and T d They are respectively expressed as the mechanical, electromagnetic and damping torques of the synchronous generator, all in N·m;

[0022] The relationship between torque and power can be expressed as: Substituting into the equation, and the damping torque in the VSG is 0, the rotor motion equation of the VSG is obtained as follows:

[0023]

[0024] Where J represents the inertia of the synchronous generator, in kg·m 2 ;P m , P e Respectively represent mechanical power and electromagnetic power, the unit is W; D represents the damping corresponding to the damping torque, the unit is N·m·s / rad; Δω=ω-ω0, ΔP=P m -P e is the actual output angular velocity of VSG, ω0 is the rated angular velocity, both in rad / s; δ is the power angle of the synchronous generator.

[0025] The specific method of S2 is:

[0026] The VSG active frequency regulation equation is:

[0027] P m =P ref -k p (ω-ω0) (2)

[0028] Where P ref is the given active power; K p is the primary frequency modulation coefficient. Combining (1) and (2) to simplify, we can get:

[0029]

[0030] Where n = 1 / (Dω0+k p ), τ=Jω0 / (Dω0+k p ),

[0031] By measuring the actual frequency of the grid connection point, the actual frequency is compared with the rated frequency to obtain the frequency deviation, simulating the behavior of the mechanical speed governor of the synchronous generator, inputting the frequency deviation Δf, and outputting the expected active power change ΔP. The active power change ΔP output by the speed governor model is used as one of the inputs of the active-frequency control loop, and is added or subtracted from the existing active power reference value to form a new active power instruction, which will guide the inverter to actively adjust its output active power according to the change of grid frequency so that the VSG has primary frequency regulation capability, and at the same time, the VSG reactive-voltage control equation is obtained:

[0032]

[0033] In the formula, U, U n are the actual output voltage and rated voltage of VSG respectively, E is the no-load electromotive force of VSG, Q ref is the reactive power setting value, Q e is the actual output reactive power value, k q is the reactive power droop coefficient of the system, and K represents the integral gain.

[0034] The specific methods of S3 are:

[0035] Assume that E∠δ and U∠0 are the output voltage phasor on the inverter side and the voltage phasor at the PCC point, respectively. The resistance and reactance in the equivalent circuit are represented by R V and X V It means that under high voltage grid, the resistance on the line can be ignored and can be approximately regarded as pure inductive. Assuming U≈E, sinδ≈δ, the VSG output active power equation is:

[0036]

[0037] By combining equations (1) and (2), the approximate expression of the VSG active closed-loop transfer function can be obtained as shown below:

[0038]

[0039] The virtual synchronous generator power frequency small signal model is established. According to the natural oscillation angular frequency ω of formula (6), n and the damping ratio ξ are:

[0040]

[0041] The above formula is the functional relationship between inertia and damping.

[0042] The specific method of S4 is:

[0043] The VSG power angle characteristic curve and VSG angular frequency oscillation curve are analyzed. Based on the analysis, the selection principles of inertia and damping are summarized. Through the correlation between inertia J and power difference ΔP, an adaptive control strategy of rotation inertia based on power difference is proposed, as follows:

[0044]

[0045] In the formula, Δω=ω-ω0, ΔP=P m -P e , J0 and D0 are the virtual moment of inertia and virtual damping of VSG during normal operation, respectively; K J and K D are the adjustment coefficients of virtual moment of inertia and virtual damping respectively; T J and T D They are the adjustment thresholds of virtual moment of inertia and virtual damping respectively.

[0046] The specific method of S5 is:

[0047] S5.1. Establish an artificial neural network, where X1 and X2 represent the input signals of the neurons, and W1 and W2 represent the weight values ​​of the input signals. The weight values ​​are changed in real time through the training algorithm to achieve the best prediction effect of the neural network;

[0048] S5.2, the output of the neuron z = g(X1, X2), where z is a nonlinear function of X1 and X2, the input vector x = (X1, X2); the output layer output vector y0 = (y1, y2); the expected output vector d0 = (d1, d2), through the single neuron network structure diagram, the change in weight Δw is obtained i (k):

[0049]

[0050] Where μ is the learning rate, and the final weight after sorting is:

[0051]

[0052] S5.3. For the VSG system, the network parameters are optimized. Δω and ΔP are used as the neural network input. K J and K D is the weight, and the expected output vector value is d1=ω=314rad / s、d2=P=P set , P set is the initial power;

[0053] Substitute the input X1 and X2 into formula (10) to calculate the correction of the weight, and then get the latest weight K J and K D , use this weight to substitute into equations (8) and (9) to calculate the new inertia J and damping D. At this time, the VSG frequency and voltage are the output of the neural network, and ω=314rad / s, P=P set Substitute the expected output vector into the global error formula:

[0054]

[0055] The error is calculated with the output. When the error meets the acceptable conditions, the algorithm ends and the latest weights W1 and W2, i.e. K, are accepted. J and K D , when the error function meets the set requirements, stop training, otherwise select the next set of samples for learning.

[0056] The beneficial effects of the present invention are:

[0057] (1) The present invention provides an inertia and damping adaptive VSG control method based on neural network optimization. By analyzing the power angle characteristic curve and angular frequency oscillation curve of the VSG, the relationship between the system power difference and the angular frequency change rate is revealed, and an inertia and damping adaptive control strategy based on the power difference is proposed to avoid the phenomenon of amplifying noise and deteriorating the power loop caused by the angular frequency change rate, effectively stabilize the system frequency, and significantly reduce the overshoot;

[0058] (2) The present invention provides an inertia and damping adaptive VSG control method based on neural network optimization. According to the artificial neural network algorithm, the prediction of frequency and voltage changes of the power grid is optimized, which effectively improves the frequency and power stability capability of the system when facing disturbances, and reduces the effectiveness of frequency disturbance and accelerating frequency recovery when encountering faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a topological structure diagram of a virtual synchronous generator based on the inertia and damping adaptive VSG control method optimized by a neural network of the present invention;

[0060] Figure 2 It is a structural block diagram of VSG active and reactive control of the inertia and damping adaptive VSG control method based on neural network optimization of the present invention;

[0061] Figure 3 It is a VSG grid-connected equivalent circuit diagram of the inertia and damping adaptive VSG control method based on neural network optimization of the present invention;

[0062] Figure 4 It is a model diagram of a virtual synchronous generator power frequency small signal of the inertia and damping adaptive VSG control method based on neural network optimization of the present invention;

[0063] Figure 5 It is a power angle characteristic curve diagram of the inertia and damping adaptive VSG control method based on neural network optimization of the present invention;

[0064] Figure 6 It is an angular frequency oscillation curve diagram of the inertia and damping adaptive VSG control method based on neural network optimization of the present invention;

[0065] Figure 7 It is a single neuron network structure diagram of the inertia and damping adaptive VSG control method based on neural network optimization of the present invention;

[0066] Figure 8 It is a neural network control flow chart adopted in the present invention. DETAILED DESCRIPTION

[0067] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0068] The inertia and damping adaptive VSG control method based on neural network optimization of the present invention has the following specific steps:

[0069] S1. Based on the traditional two-order model of synchronous generator, the pole pair number of synchronous generator is assumed to be 1, and the VSG rotor motion equation is obtained through equivalent transformation.

[0070] like Figure 1 As shown in the figure, VSG analyzes the electromagnetic characteristics and mechanical torque characteristics of the synchronous generator, uses the control algorithm to simulate the generator output characteristics, realizes the purpose of the grid-connected inverter simulating the synchronous generator, and obtains the virtual synchronous generator topology. The specific process is:

[0071] The rotor motion equation of a traditional synchronous generator is based on Newton's second law, where torque equals mass times acceleration, and the equation is as follows:

[0072]

[0073] In the above formula, T m , T e and T d They are respectively expressed as the mechanical, electromagnetic and damping torques of the synchronous generator, all in N·m;

[0074] The relationship between torque and power can be expressed as: Substituting into the equation, and the damping torque in the VSG is 0, the rotor motion equation of the VSG is obtained as follows:

[0075]

[0076] Where J represents the inertia of the synchronous generator, in kg·m 2 ;P m , P e Respectively represent mechanical power and electromagnetic power, the unit is W; D represents the damping corresponding to the damping torque, the unit is N·m·s / rad; Δω=ω-ω0, ΔP=P m -P e is the actual output angular velocity of VSG, ω0 is the rated angular velocity, both in rad / s; δ is the power angle of the synchronous generator.

[0077] S2. In order to make the microgrid inverter have the primary frequency regulation capability of the synchronous generator, a speed regulator is introduced into the active-frequency control loop. The function of the speed regulator is to control the speed of the prime mover of the generator for primary frequency regulation, thereby indirectly controlling the active power output of the generator. The VSG reactive-voltage control equation is obtained based on the VSG active frequency regulation equation (speed regulator control equation). The specific method is:

[0078] The VSG active frequency regulation equation is:

[0079] P m =P ref -k p (ω-ω0) (2)

[0080] Where P ref is the given active power; K p is the primary frequency modulation coefficient. Combining (1) and (2) to simplify, we can get:

[0081]

[0082] Where n = 1 / (Dω0+k p ), τ=Jω0 / (Dω0+k p ),

[0083] By measuring the actual frequency of the grid connection point, the actual frequency is compared with the rated frequency to obtain the frequency deviation, simulating the behavior of the mechanical speed governor of the synchronous generator, inputting the frequency deviation Δf, and outputting the expected active power change ΔP. The active power change ΔP output by the speed governor model is used as one of the inputs of the active-frequency control loop, and is added or subtracted from the existing active power reference value to form a new active power instruction, which will guide the inverter to actively adjust its output active power according to the change of grid frequency so that the VSG has primary frequency regulation capability, and at the same time, the VSG reactive-voltage control equation is obtained:

[0084]

[0085] In the formula, U, U n are the actual output voltage and rated voltage of VSG respectively, E is the no-load electromotive force of VSG, Q ref is the reactive power setting value, Q e is the actual output reactive power value, k q is the reactive power droop coefficient of the system, and K represents the integral gain.

[0086] The reactive voltage equation of the original VSG is:

[0087] Q e =Q ref +k q (U n -U),

[0088] However, in order to prevent voltage jumps during the voltage regulation process, the VSG is improved in the excitation regulation link and an inertia link is introduced. Make its dynamic response process closer to the actual synchronous generator, and add U nThis is because the control equation adjusts the reactive output of the VSG according to the deviation between the current voltage and the rated voltage. When the voltage deviates from the rated value, the VSG can help pull the voltage back to the rated value by increasing or decreasing its reactive output, such as Figure 2 As shown, Figure 2 The control block diagram below is the improved VSG excitation regulator control diagram.

[0089] S3. Establish a VSG power frequency small signal model and perform circuit model equivalence on the VSG grid-connected system. The equivalent circuit diagram is as follows: Figure 3 As shown, the VSG output active power equation is obtained. Combining the VSG rotor motion equation and the VSG active frequency regulation equation, the approximate expression of the VSG active loop closed-loop transfer function is obtained, and then the functional relationship between inertia and damping is obtained. The specific method is:

[0090] Assume that E∠δ and U∠0 are the output voltage phasor on the inverter side and the voltage phasor at the PCC point, respectively. The resistance and reactance in the equivalent circuit are represented by R V and X V It means that under high voltage grid, the resistance on the line can be ignored and can be approximately regarded as pure inductive. Assuming U≈E, sinδ≈δ, the VSG output active power equation is:

[0091]

[0092] Among them, the method to obtain formula (5) is:

[0093] Because under high-voltage power grid, due to the small line impedance, the line impedance voltage drop can be ignored compared to the output voltage, and the inductance value of the line is much greater than its resistance value, then δ=0, R=0 and the original formula is:

[0094]

[0095] When δ=0,R=0, but sinθ=1,sinδ≈δ,

[0096] Then we get:

[0097]

[0098] By combining equation (1) and equation (2), the approximate expression of the closed-loop transfer function of the VSG active loop can be obtained as shown in the following equation:

[0099]

[0100] Establish a virtual synchronous generator power frequency small signal model, such as Figure 4 As shown, according to the natural oscillation angular frequency ω of formula (6)n and the damping ratio ξ are:

[0101]

[0102] The above formula is the functional relationship between inertia and damping.

[0103] Formula (7) is to analyze the change principle of inertia and damping together with the power angle characteristic and the angular frequency oscillation curve. Formula (7) clarifies the relationship between inertia and damping. For example, in the control system, there is usually a certain damping ratio, but the over-damped system responds slowly, so the under-damped system is often used (usually ζ is 0.4~0.8). Therefore, it is necessary to keep the damping ratio unchanged. When the inertia increases, the damping needs to be increased according to the formula. When the inertia decreases, the damping needs to be adjusted down.

[0104] S4. According to the functional relationship between inertia and damping, combined with the power angle characteristic curve and the angular frequency oscillation curve, the power angle characteristics and angular frequency oscillation of VSG under disturbance are comprehensively considered, the selection principles of inertia and damping are obtained, and the inertia adaptive mechanism based on power difference is summarized.

[0105] Based on the above formula (7), the relationship between inertia and damping can be clarified. Then, combined with the analysis of the power angle characteristic curve and the angular frequency oscillation curve, the relationship between frequency and inertia can be confirmed according to the obtained natural angular frequency formula, thereby obtaining a selection principle for inertia. According to the obtained damping ratio formula, because the system needs to be stable in an underdamped state, the damping ratio remains unchanged. The inertia changes according to the frequency, and the damping also changes according to the change of inertia. The specific method of S4 is:

[0106] like Figure 5 , 6 As shown in the figure, the VSG power angle characteristic curve and the VSG angular frequency oscillation curve are analyzed. When the active power of the given virtual synchronous generator increases from P1 to P2, the stable operating point of the system switches from point A to point B. During the switching process, the power and frequency changes of the system show the characteristics of attenuated oscillation, which is similar to the oscillation process of the synchronous generator. For the convenience of analysis, the attenuated oscillation process is divided into 4 parts: t1-t2, t2-t3, t3-t4, t4-t5. Due to the different characteristics of the power and power angle changes in each interval, the expected inertia and damping values ​​are also different.

[0107] In the interval t1-t2, the system is disturbed and transitions from point A to point B. Due to the input mechanical power P of the VSG m Always greater than the electromagnetic power P output by the system eThe rotor angular velocity of VSG is higher than the grid angular velocity, and the difference between the two gradually increases. The rate of change of angular velocity dω / dt is greater than 0. When it reaches point B, dω / dt=0. Therefore, in the interval t1-t2, the virtual inertia J must be increased to slow down the rapid rise of angular frequency, and the increase of Δω>0 is limited by increasing damping D. In the interval t2-t3, the system runs from point B to point C. At this time, the input power is always less than the output electromagnetic power. Therefore, the rate of change of the virtual rotor angular velocity dω / dt<0, the system begins to decelerate, and the angular frequency ω gradually decreases. However, in this interval Δω>0, in order to quickly restore the angular velocity to a stable value, the inertia J should be appropriately reduced. In addition, increasing the damping D will reduce dω / dt, so the damping D should be appropriately reduced to prevent the rapid recovery of the angular velocity due to excessive damping D. For the intervals t3-t4 and t4-t5, the analysis and selection of inertia and damping are similar to those in the intervals t1-t2 and t2-t3.

[0108] Based on the above analysis, the selection principles of inertia and damping are summarized as follows: Figure 7 As shown in the figure, by analyzing the correlation between the inertia J and the power difference ΔP, an adaptive control strategy for the moment of inertia based on the power difference is proposed, as follows:

[0109]

[0110] In the formula, Δω=ω-ω0, ΔP=P m -P e , J0 and D0 are the virtual moment of inertia and virtual damping of VSG during normal operation, respectively; K J and K D are the adjustment coefficients of virtual moment of inertia and virtual damping respectively; T J and T D They are the adjustment thresholds of virtual moment of inertia and virtual damping respectively.

[0111] Table 1 is a diagram showing the principle of selecting inertia J and damping D in the present invention;

[0112]

[0113] S5, such as Figure 7 As shown, an artificial neural network is established to monitor and model the frequency and voltage changes of the power grid, and the weight value is adjusted after evaluating the prediction results to achieve adaptive adjustment of inertia and damping. The specific method is as follows:

[0114] S5.1. Establish an artificial neural network, where X1 and X2 represent the input signals of the neurons, and W1 and W2 represent the weight values ​​of the input signals. The weight values ​​are changed in real time through the training algorithm to achieve the best prediction effect of the neural network;

[0115] S5.2, the output of the neuron z = g(X1, X2), where z is a nonlinear function of X1 and X2, the input vector x = (X1, X2); the output layer output vector y0 = (y1, y2); the expected output vector d0 = (d1, d2), through the single neuron network structure diagram, the change in weight Δw is obtained i (k):

[0116]

[0117] Where μ is the learning rate, and the final weight after sorting is:

[0118]

[0119] S5.3, if Figure 8 As shown in the figure, the VSG system is optimized through network parameters, Δω and ΔP are used as neural network inputs, K J and K D is the weight, and the expected output vector value is d1=ω=314rad / s、d2=P=P set , P set is the initial power;

[0120] Substitute the input X1 and X2 into formula (10) to calculate the correction of the weight, and then get the latest weight K J and K D , use this weight to substitute into equations (8) and (9) to calculate the new inertia J and damping D. At this time, the VSG frequency and voltage are the output of the neural network, and ω=314rad / s, P=P set Substitute the expected output vector into the global error formula:

[0121]

[0122] The error is calculated with the output. When the error meets the acceptable conditions, the algorithm ends and the latest weights W1 and W2, i.e. K, are accepted. J and K D , when the error function meets the set requirements, stop training, otherwise select the next set of samples for learning.

[0123] Example 1

[0124] 1. The inertia and damping adaptive VSG control method based on neural network optimization, the specific steps are as follows:

[0125] S1. Based on the traditional two-order model of synchronous generator, the pole pair number of synchronous generator is assumed to be 1, and the VSG rotor motion equation is obtained through equivalent transformation;

[0126] S2. Introduce a speed regulator into the active power-frequency control loop, control the active power output of the generator by frequency modulation, and obtain the VSG reactive power-voltage control equation based on the VSG active frequency regulation equation;

[0127] S3. Establish a VSG power frequency small signal model, perform circuit model equivalence on the VSG grid-connected system, obtain the VSG output active power equation, combine the VSG rotor motion equation, VSG active frequency regulation equation and VSG reactive-voltage control equation, obtain the approximate expression of the VSG active loop closed-loop transfer function, and then derive the functional relationship between inertia and damping;

[0128] S4. Analyze the power angle characteristics and angular frequency oscillation of VSG under disturbance conditions, clarify the selection principles of inertia and damping, and further analyze and summarize the inertia adaptation mechanism based on power difference;

[0129] S5. Establish an artificial neural network to analyze the operation status of the power grid, analyze and predict the frequency and voltage changes of the power grid, adjust the weight value after analyzing the predicted value, and realize adaptive adjustment of inertia and damping.

[0130] Example 2

[0131] 1. The inertia and damping adaptive VSG control method based on neural network optimization, the specific steps are as follows:

[0132] S1. Based on the traditional two-order model of synchronous generator, the pole pair number of synchronous generator is assumed to be 1, and the VSG rotor motion equation is obtained through equivalent transformation. The specific process is as follows:

[0133] The rotor motion equation of a traditional synchronous generator is based on Newton's second law, where torque equals mass times acceleration, and the equation is as follows:

[0134]

[0135] In the above formula, T m , T e and T d They are respectively expressed as the mechanical, electromagnetic and damping torques of the synchronous generator, all in N·m;

[0136] The relationship between torque and power can be expressed as: Substituting into the equation, and the damping torque in the VSG is 0, the rotor motion equation of the VSG is obtained as follows:

[0137]

[0138] Where J represents the inertia of the synchronous generator, in kg·m 2 ;P m , P eRespectively represent mechanical power and electromagnetic power, the unit is W; D represents the damping corresponding to the damping torque, the unit is N·m·s / rad; Δω=ω-ω0, ΔP=P m -P e is the actual output angular velocity of VSG, ω0 is the rated angular velocity, both in rad / s; δ is the power angle of the synchronous generator.

[0139] S2. In order to make the microgrid inverter have the primary frequency regulation capability of the synchronous generator, a speed regulator is introduced into the active-frequency control loop. The function of the speed regulator is to control the speed of the prime mover of the generator for primary frequency regulation, thereby indirectly controlling the active power output of the generator. The VSG reactive-voltage control equation is obtained based on the VSG active frequency regulation equation (speed regulator control equation). The specific method is:

[0140] The VSG active frequency regulation equation is:

[0141] P m =P ref -k p (ω-ω0) (2)

[0142] Where P ref is the given active power; K p is the primary frequency modulation coefficient. Combining (1) and (2) to simplify, we can get:

[0143]

[0144] Where n = 1 / (Dω0+k p ), τ=Jω0 / (Dω0+k p ),

[0145] By measuring the actual frequency of the grid connection point, the actual frequency is compared with the rated frequency to obtain the frequency deviation, simulating the behavior of the mechanical speed governor of the synchronous generator, inputting the frequency deviation Δf, and outputting the expected active power change ΔP. The active power change ΔP output by the speed governor model is used as one of the inputs of the active-frequency control loop, and is added or subtracted from the existing active power reference value to form a new active power instruction, which will guide the inverter to actively adjust its output active power according to the change of grid frequency so that the VSG has primary frequency regulation capability, and at the same time, the VSG reactive-voltage control equation is obtained:

[0146]

[0147] In the formula, U, U n are the actual output voltage and rated voltage of VSG respectively, E is the no-load electromotive force of VSG, Q ref is the reactive power setting value, Q e is the actual output reactive power value, kq is the reactive power droop coefficient of the system, and K represents the integral gain.

[0148] The reactive voltage equation of the original VSG is:

[0149] Q e =Q ref +k q (U n -U),

[0150] However, in order to prevent voltage jumps during the voltage regulation process, the VSG is improved in the excitation regulation link and an inertia link is introduced. Make its dynamic response process closer to the actual synchronous generator, and add U n This is because the control equation adjusts the reactive output of the VSG according to the deviation between the current voltage and the rated voltage. When the voltage deviates from the rated value, the VSG can help pull the voltage back to the rated value by increasing or decreasing its reactive output, such as Figure 2 As shown, Figure 2 The control block diagram below is the improved VSG excitation regulator control diagram.

[0151] S3. Establish a VSG power frequency small signal model and perform circuit model equivalence on the VSG grid-connected system. The equivalent circuit diagram is as follows: Figure 3 As shown in the figure, the VSG output active power equation is obtained. Combining the VSG rotor motion equation, the VSG active frequency regulation equation and the VSG reactive-voltage control equation, the approximate expression of the VSG active loop closed-loop transfer function is obtained, and then the functional relationship between inertia and damping is obtained. The specific method is:

[0152] Assume that E∠δ and U∠0 are the output voltage phasor on the inverter side and the voltage phasor at the PCC point, respectively. The resistance and reactance in the equivalent circuit are represented by R V and X V It means that under high voltage grid, the resistance on the line can be ignored and can be approximately regarded as pure inductive. Assuming U≈E, sinδ≈δ, the VSG output active power equation is:

[0153]

[0154] Among them, the method to obtain formula (5) is:

[0155] Because under high-voltage power grid, due to the small line impedance, the line impedance voltage drop can be ignored compared to the output voltage, and the inductance value of the line is much greater than its resistance value, then δ=0, R=0 and the original formula is:

[0156]

[0157] When δ=0,R=0, but sinθ=1,sinδ≈δ,

[0158] Then we get:

[0159]

[0160] By combining equation (1) and equation (2), the approximate expression of the closed-loop transfer function of the VSG active loop can be obtained as shown in the following equation:

[0161]

[0162] Establish a virtual synchronous generator power frequency small signal model, such as Figure 4 As shown, according to the natural oscillation angular frequency ω of formula (6) n and the damping ratio ξ are:

[0163]

[0164] The above formula is the functional relationship between inertia and damping.

[0165] Formula 7 clarifies the relationship between inertia and damping. For example, there is usually a certain damping ratio in the control system, but the over-damped system responds slowly, so the under-damped system is often used (usually ζ is 0.4 to 0.8). Therefore, it is necessary to keep the damping ratio unchanged. When the inertia increases, the damping needs to be increased according to the formula. When the inertia decreases, the damping needs to be adjusted down.

[0166] S4. Analyze the power angle characteristics and angular frequency oscillation of VSG under disturbance conditions, clarify the selection principles of inertia and damping, and further analyze and summarize the inertia adaptation mechanism based on power difference;

[0167] S5. Establish an artificial neural network to analyze the operation status of the power grid, analyze and predict the frequency and voltage changes of the power grid, adjust the weight value after analyzing the predicted value, and realize adaptive adjustment of inertia and damping.

[0168] Example 3

[0169] 1. The inertia and damping adaptive VSG control method based on neural network optimization, the specific steps are as follows:

[0170] S1. Based on the traditional two-order model of synchronous generator, the pole pair number of synchronous generator is assumed to be 1, and the VSG rotor motion equation is obtained through equivalent transformation;

[0171] S2. Introduce a speed regulator into the active power-frequency control loop, control the active power output of the generator by frequency modulation, and obtain the VSG reactive power-voltage control equation based on the VSG active frequency regulation equation;

[0172] S3. Establish a VSG power frequency small signal model, perform circuit model equivalence on the VSG grid-connected system, obtain the VSG output active power equation, combine the VSG rotor motion equation, VSG active frequency regulation equation and VSG reactive-voltage control equation, obtain the approximate expression of the VSG active loop closed-loop transfer function, and then derive the functional relationship between inertia and damping;

[0173] S4. Analyze the power angle characteristics and angular frequency oscillation of VSG under disturbance conditions, clarify the selection principles of inertia and damping, and further analyze and summarize the inertia adaptive mechanism based on power difference. The specific method is:

[0174] like Figure 5 , 6 As shown in the figure, the VSG power angle characteristic curve and the VSG angular frequency oscillation curve are analyzed. When the active power of the given virtual synchronous generator increases from P1 to P2, the stable operating point of the system switches from point A to point B. During the switching process, the power and frequency changes of the system show the characteristics of attenuated oscillation, which is similar to the oscillation process of the synchronous generator. For the convenience of analysis, the attenuated oscillation process is divided into 4 parts: t1-t2, t2-t3, t3-t4, t4-t5. Due to the different characteristics of the power and power angle changes in each interval, the expected inertia and damping values ​​are also different.

[0175] In the interval t1-t2, the system is disturbed and transitions from point A to point B. Due to the input mechanical power P of the VSG m Always greater than the electromagnetic power P output by the system e The rotor angular velocity of VSG is higher than the grid angular velocity, and the difference between the two gradually increases. The rate of change of angular velocity dω / dt is greater than 0. When it reaches point B, dω / dt=0. Therefore, in the interval t1-t2, the virtual inertia J must be increased to slow down the rapid rise of angular frequency, and the increase of Δω>0 is limited by increasing damping D. In the interval t2-t3, the system runs from point B to point C. At this time, the input power is always less than the output electromagnetic power. Therefore, the rate of change of the virtual rotor angular velocity dω / dt<0, the system begins to decelerate, and the angular frequency ω gradually decreases. However, in this interval Δω>0, in order to quickly restore the angular velocity to a stable value, the inertia J should be appropriately reduced. In addition, increasing the damping D will reduce dω / dt, so the damping D should be appropriately reduced to prevent the rapid recovery of the angular velocity due to excessive damping D. For the intervals t3-t4 and t4-t5, the analysis and selection of inertia and damping are similar to those in the intervals t1-t2 and t2-t3.

[0176] Based on the above analysis, the selection principles of inertia and damping are summarized as follows: Figure 7As shown in the figure, by analyzing the correlation between the inertia J and the power difference ΔP, an adaptive control strategy for the moment of inertia based on the power difference is proposed, as follows:

[0177]

[0178] In the formula, Δω=ω-ω0, ΔP=P m -P e , J0 and D0 are the virtual moment of inertia and virtual damping of VSG during normal operation, respectively; K J and K D are the adjustment coefficients of virtual moment of inertia and virtual damping respectively; T J and T D They are the adjustment thresholds of virtual moment of inertia and virtual damping respectively.

[0179] Table 1 is a diagram showing the principle of selecting inertia J and damping D in the present invention;

[0180]

[0181] S5, such as Figure 7 As shown, an artificial neural network is established to analyze the operation status of the power grid, analyze and predict the frequency and voltage changes of the power grid, adjust the weight value after analyzing the predicted value, and realize the adaptive adjustment of inertia and damping. The specific method is:

[0182] S5.1. Establish an artificial neural network, where X1 and X2 represent the input signals of the neurons, and W1 and W2 represent the weight values ​​of the input signals. The weight values ​​are changed in real time through the training algorithm to achieve the best prediction effect of the neural network;

[0183] S5.2, the output of the neuron z = g(X1, X2), where z is a nonlinear function of X1 and X2, the input vector x = (X1, X2); the output layer output vector y0 = (y1, y2); the expected output vector d0 = (d1, d2), through the single neuron network structure diagram, the change in weight Δw is obtained i (k):

[0184]

[0185] Where μ is the learning rate, and the final weight after sorting is:

[0186]

[0187] S5.3, if Figure 8 As shown in the figure, the VSG system is optimized through network parameters, Δω and ΔP are used as neural network inputs, K J and K Dis the weight, and the expected output vector value is d1=ω=314rad / s、d2=P=P set , P set is the initial power;

[0188] Substitute the input X1 and X2 into formula (10) to calculate the correction of the weight, and then get the latest weight K J and K D , use this weight to substitute into equations (8) and (9) to calculate the new inertia J and damping D. At this time, the VSG frequency and voltage are the output of the neural network, and ω=314rad / s, P=P set Substitute the expected output vector into the global error formula:

[0189]

[0190] The error is calculated with the output. When the error meets the acceptable conditions, the algorithm ends and the latest weights W1 and W2, i.e. K, are accepted. J and K D , when the error function meets the set requirements, stop training, otherwise select the next set of samples for learning.

Claims

1. The inertia and damping adaptive VSG control method based on neural network optimization is characterized by: The specific steps are as follows: S1. Based on the traditional two-order model of synchronous generator, the pole pair number of synchronous generator is assumed to be 1, and the VSG rotor motion equation is obtained through equivalent transformation; S2. Introduce a speed regulator into the active power-frequency control loop, control the active power output of the generator by frequency modulation, and obtain the VSG reactive power-voltage control equation based on the VSG active frequency regulation equation; S3. Establish a VSG power frequency small signal model, perform circuit model equivalence on the VSG grid-connected system, obtain the VSG output active power equation, combine the VSG rotor motion equation and the VSG active frequency regulation equation, obtain the approximate expression of the VSG active loop closed-loop transfer function, and then derive the functional relationship between inertia and damping; S4. According to the functional relationship between inertia and damping, combined with the power angle characteristic curve and the angular frequency oscillation curve, the power angle characteristics and angular frequency oscillation of VSG under disturbance are comprehensively considered, the selection principle of inertia and damping is obtained, and the inertia adaptive mechanism based on power difference is summarized; S5. Establish an artificial neural network to monitor and model the frequency and voltage changes of the power grid, adjust the weight value after evaluating the prediction results, and realize adaptive adjustment of inertia and damping.

2. The inertia and damping adaptive VSG control method based on neural network optimization according to claim 1 is characterized in that: The specific process of S1 is as follows: The rotor motion equation of a traditional synchronous generator is based on Newton's second law, where torque equals mass times acceleration, and the equation is as follows: In the above formula, T m , T e and T d They are respectively expressed as the mechanical, electromagnetic and damping torques of the synchronous generator, all in N·m; The relationship between torque and power can be expressed as: Substituting into the equation, and the damping torque in the VSG is 0, the rotor motion equation of the VSG is obtained as follows: Where J represents the inertia of the synchronous generator, in kg·m 2 ;P m , P e Respectively represent mechanical power and electromagnetic power, the unit is W; D represents the damping corresponding to the damping torque, the unit is N·m·s / rad; Δω=ω-ω0, ΔP=P m -P e is the actual output angular velocity of VSG, ω0 is the rated angular velocity, both in rad / s; δ is the power angle of the synchronous generator.

3. The inertia and damping adaptive VSG control method based on neural network optimization according to claim 1 is characterized in that: The specific method of S2 is: The VSG active frequency regulation equation is: P m =P ref -k p (ω-ω0) (2) Where P ref is the given active power; K p is the primary frequency modulation coefficient. Combining (1) and (2) to simplify, we can get: where n = 1 / (Dω0 + k p ), τ = Jω0 / (Dω0 + k p ), By measuring the actual frequency of the grid connection point, the actual frequency is compared with the rated frequency to obtain the frequency deviation, simulating the behavior of the mechanical speed governor of the synchronous generator, inputting the frequency deviation Δf, and outputting the expected active power change ΔP. The active power change ΔP output by the speed governor model is used as one of the inputs of the active-frequency control loop, and is added or subtracted from the existing active power reference value to form a new active power instruction, which will guide the inverter to actively adjust its output active power according to the change of grid frequency so that the VSG has primary frequency regulation capability, and at the same time, the VSG reactive-voltage control equation is obtained: In the formula, U, U n are the actual output voltage and rated voltage of VSG respectively, E is the no-load electromotive force of VSG, Q ref is the reactive power setting value, Q e is the actual output reactive power value, k q is the reactive droop coefficient of the system, and K represents the integral gain.

4. The inertia and damping adaptive VSG control method based on neural network optimization according to claim 1 is characterized in that: The specific method of S3 is: Assume that E∠δ and U∠0 are the output voltage phasor on the inverter side and the voltage phasor at the PCC point, respectively. The resistance and reactance in the equivalent circuit are represented by R V and X V It means that under high voltage grid, the resistance on the line can be ignored and can be approximately regarded as pure inductive. Assuming U≈E, sinδ≈δ, the VSG output active power equation is: By combining equations (1) and (2), the approximate expression of the VSG active closed-loop transfer function can be obtained as shown below: The virtual synchronous generator power frequency small signal model is established. According to the natural oscillation angular frequency ω of formula (6), n and the damping ratio ξ are: The above formula is the functional relationship between inertia and damping.

5. The inertia and damping adaptive VSG control method based on neural network optimization according to claim 1 is characterized in that: The specific method of S4 is: The VSG power angle characteristic curve and VSG angular frequency oscillation curve are analyzed. Based on the analysis, the selection principles of inertia and damping are summarized. Through the correlation between inertia J and power difference ΔP, an adaptive control strategy of rotation inertia based on power difference is proposed, as follows: In the formula, Δω=ω-ω0, ΔP=P m -P e , J0 and D0 are the virtual moment of inertia and virtual damping of VSG during normal operation, respectively; K J and K D are the adjustment coefficients of virtual moment of inertia and virtual damping respectively; T J and T D They are the adjustment thresholds of virtual moment of inertia and virtual damping respectively.

6. The inertia and damping adaptive VSG control method based on neural network optimization according to claim 1 is characterized in that: The specific method of S5 is: S5.

1. Establish an artificial neural network, where X1 and X2 represent the input signals of the neurons, and W1 and W2 represent the weight values ​​of the input signals. The weight values ​​are changed in real time through the training algorithm to achieve the best prediction effect of the neural network; S5.2, the output of the neuron z = g(X1, X2), where z is a nonlinear function of X1 and X2, the input vector x = (X1, X2); the output layer output vector y0 = (y1, y2); the expected output vector d0 = (d1, d2), through the single neuron network structure diagram, the change in weight Δw is obtained i (k): Where μ is the learning rate, and the final weight after sorting is: S5.

3. For the VSG system, the network parameters are optimized. Δω and ΔP are used as the neural network input. K J and K D is the weight, and the expected output vector value is d1=ω=314rad / s、d2=P=P set , P set is the initial power; Substitute the input X1 and X2 into formula (10) to calculate the correction of the weight, and then get the latest weight K J and K D , use this weight to substitute into equations (8) and (9) to calculate the new inertia J and damping D. At this time, the VSG frequency and voltage are the output of the neural network, and ω=314rad / s, P=P set Substitute the expected output vector into the global error formula: The error is calculated with the output. When the error meets the acceptable conditions, the algorithm ends and the latest weights W1 and W2, i.e. K, are accepted. J and K D , when the error function meets the set requirements, stop training, otherwise select the next set of samples for learning.

Images