Grid-connected control method for improved RBF (Radial Basis Function) networking type converter with active support capability

Abstract

The invention relates to the technical field of intelligent power grids, in particular to a grid-connected control method for an improved RBF (Radial Basis Function) network construction type converter with active support capability. The method comprises the following steps: firstly, taking a group of initial values in a VSG system virtual inertia and equivalent damping coefficient parameter value range; determining an initial angular frequency according to the initial value and a swing equation; a VSG secondary control system is improved, an improved phase self-tracking pre-synchronization method without a phase-locked loop is added, a voltage deviation signal is responded, a frequency phase is compensated, and the rotational inertia and the damping coefficient of the system are adjusted in real time to respond to the change of the system state. According to the neural network, the virtual inertia, the damping coefficient and the angular velocity compensation amount are adaptively adjusted, and whether parameters accord with criteria is compared, so that the input parameters are corrected, the influence of disturbance on the system frequency and the output active power is reduced, the dynamic response process of the system is improved, and the stability of the system is improved. And system output power distribution errors and overshoot caused by a rotor motion equation are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grids, and in particular to a grid-connected control method for an improved RBF grid-forming converter with active support capability. Background Art

[0002] In recent years, intermittent energy sources such as photovoltaic and wind power have been connected to the grid in the form of current sources, which has an increasingly greater impact on the grid. To this end, many scholars have proposed virtual synchronous generators, which make the external interface of renewable energy power generation systems have the characteristics of synchronous generators, opening up a new way for large-scale access of distributed power sources to the grid. When the capacity of distributed power sources is small, they can supply power nearby, which is called a microgrid; when the capacity of distributed power sources can not only meet the power demand of local loads, but also deliver a certain amount of energy to the grid, it is called grid-connected operation. Under the background of the "dual carbon" goals advocated by my country, distributed power sources usually need to be connected to the grid, which involves the stability problem of synchronous operation with the grid. Like traditional synchronous generators, VSGs also need to be synchronized with the grid before they are connected to the grid. Direct grid-connected operation of distributed power sources will generate huge instantaneous impact currents on the grid and damage related devices. In order to ensure good grid-connected quality, reduce unnecessary impacts, and improve system stability, the converter needs to be pre-synchronized before grid connection.

[0003] The energy storage inverter system has two operating states: off-grid and grid-connected. According to the operating needs and the fault conditions of the power grid, the VSG needs to switch flexibly between the two operating states. However, due to the different control methods of the two operating states, the inverter output voltage and the grid voltage may be different in frequency, amplitude and phase when switching from the off-grid state to the grid-connected state, which is likely to cause impacts during the grid-connected process. The impact current generated will affect the normal operation of electronic devices, and in severe cases may cause damage, ultimately leading to a decrease in system stability and grid-connected failure. Therefore, how to smoothly switch the inverter system from the off-grid state to the grid-connected state without generating severe impacts and fluctuations and ensuring the stable operation of the system is a very important issue. Adjusting the PCS output voltage characteristics to be consistent with the grid and eliminating phase deviations are the key to ensuring stable grid-connected operation.

[0004] In pre-synchronous control, Virtual Synchronous Generator (VSG) technology is the most widely used. Traditional pre-synchronous control uses a phase-locked loop to collect amplitude, frequency and phase information on the grid side, and continuously adjusts the VSG output voltage with reference to the phase-locked loop output information. The grid connection point voltage is synchronized with the grid voltage through a fixed voltage difference of the line impedance, and the voltage difference and phase angle difference are respectively connected to the Q-U controller and P-f controller through the PI controller to achieve seamless switching from off-grid to grid-connected and achieve friendly grid connection.

[0005] likeFigure 1 As shown, the traditional inverter grid-connected synchronization process only improves the VSG active control loop, which includes the grid voltage u g Coincident with the d-axis of the dq-axis, the phase is θ g , with angular velocity ω g The inverter terminal voltage u has a phase of ω and rotates at an angular velocity ω. The phase difference between the grid voltage and the inverter output voltage is Δθ. When the grid-connected inverter switches from off-grid to grid-connected operation, the inverter output voltage is adjusted to have the same phase as the grid voltage by adjusting the angular velocity of the inverter output voltage. When the grid voltage and the inverter output voltage coincide, u q = 0. By controlling u q =0 can make the two phases consistent.

[0006] However, the introduction of a phase-locked loop will affect the accuracy and speed of pre-synchronization control, the system dynamic process is complex, and the frequency fluctuation amplitude is too large during the process, which has a great impact on the stability of the system. In addition, after the control link is out of operation, the output power jump phenomenon occurs, which cannot meet the active support requirements of new energy grid connection and the system stability requirements.

[0007] As a powerful nonlinear mapping tool, radial basis function (RBF) neural network has shown broad application prospects in the field of system modeling and control; RBF neural network has good generalization ability and the advantage of approximating any nonlinear function. Therefore, the use of RBF neural network can solve the nonlinear problem between system parameters while ensuring fast response. Furthermore, RBF neural network can be designed to adaptively adjust the moment of inertia and damping coefficient of VSG and grid-connected compensation value and applied to fractional-order VSG controller.

[0008] Therefore, it is an urgent problem for technical personnel in this field to propose an improved RBF grid-connected converter grid-connected control method with active support capability to suppress dynamic oscillation and power overshoot of grid-connected active power under disturbance, reduce overshoot, ensure smooth switching of off-grid / grid-connected mode, improve the robustness and stability of grid-connected control, and reduce phase error. Summary of the invention

[0009] The present invention provides an improved RBF grid-connected converter control method with active support capability, which can respond to voltage deviation signals and compensate for frequency phase through an improved phase self-tracking pre-synchronization method without a phase-locked loop, and adjust the system's moment of inertia and damping coefficient in real time to respond to changes in system status, thereby reducing the impact of disturbances on system frequency and output active power, improving the system's dynamic response process, reducing the system output power distribution error and overshoot caused by the rotor motion equation, improving the system's grid-connected performance and stability, and effectively controlling the grid frequency and improving the grid power quality. The specific technical solutions are as follows:

[0010] An improved RBF grid-connected converter control method with active support capability includes the following process:

[0011] S100: The grid-connected inverter operates in an island state, the active power loop of the virtual synchronous generator obtains the voltage phase θ, and uses its initial damping and virtual inertia parameters as the input of the RBF neural network, and compares the rotor motion equation in real time to control and adjust the control error;

[0012] The rotor motion equation is as follows:

[0013]

[0014] Where J and D represent the moment of inertia and damping coefficient respectively; Pm represents mechanical power; Pe represents electromagnetic power; T m and T e They are the mechanical torque and electromagnetic torque of the synchronous generator respectively;

[0015] S200: The angle θ obtained by the virtual synchronous generator through the active power loop is compared with the electrical angle Δθ of the filter phase shift L Add together to get the modulation wave electrical angle θ of the grid-connected inverter N ;

[0016] S300: By sampling the current and voltage on the grid side, the output voltage phase of the grid-connected inverter is adjusted; the grid voltage is converted into u by 3s / 2s ga and u gb , the virtual synchronous generator active power loop output phase angle θ is input into 2s / 2r to obtain the inverter d-axis and q-axis components u gq 、u gd , and the grid-side reference value After comparison, the obtained difference is PI regulated, and the Δω output by the PI regulator is compensated to the active power loop of the VSG;

[0017] S400: voltage synchronization adjustment; specific process: the grid voltage of the energy storage converter is converted by 3s / 2s, and then u is obtained through the voltage amplitude calculation module.gm ; Taking the grid voltage amplitude u gm For a given, the inverter output voltage amplitude u m For feedback, the obtained difference is PI regulated, and the output Δu of the PI regulator is compensated to the voltage given by the reactive power loop of the virtual synchronous generator, so that the output voltage amplitude of the inverter can be adjusted, and finally the output voltage amplitude of the inverter is made to track the voltage amplitude of the grid, completing the synchronization of the inverter and the grid side voltage;

[0018] Step 500: Determine the initial phase angle and angular frequency according to the initial value G2(s) and the rotor motion equation; and according to the angular velocity compensation and damping inertia parameters output in the improved pre-synchronization control; determine whether to update the weights by the neural network evaluation function E(k), and then adaptively adjust the virtual inertia J, damping coefficient D and angular velocity deviation Δω according to the neural network, and update the grid-connected angular velocity deviation to make it more suitable for the grid-side value;

[0019] The initial value G2(s) is determined by the following formula:

[0020]

[0021] S600: Determine whether the three fall within a given value range. If so, obtain the value at the next moment. Calculate the equivalent compensation angular momentum by the following formula, return the virtual inertia J, damping coefficient D and phase adjustment feedback to the VSG control equation. Finally, when the grid voltage is consistent with the grid-connected inverter output voltage frequency, electrical angle and amplitude, close the PCC to complete the grid connection.

[0022] According to the expression of active closed-loop transfer function, the fixed angular frequency ω of the VSG system is obtained n and the damping ratio ξ,

[0023]

[0024] In the formula, K ω is the VSG adjustment coefficient, E is the voltage electromotive force;

[0025] The adjustment time of the system based on the damping ratio and frequency disturbance is:

[0026] By adjusting the time, it is determined whether the damping and inertia iterated by RBF meet the system stability requirements.

[0027] Preferably, the RBF neural network parameters are as follows:

[0028] Neural Network Machine:

[0029]

[0030] Where i = 1, 2, ... 5, bi is the element in the width vector b of the Gaussian basis function, b=[b1,…,b5] T ;

[0031] is the center point vector value of the i-th hidden layer neuron;

[0032] The RBF neural network used uses input variables x(1)=ω, x(2)=dω / dt, and adopts a neural network structure of 2-5-3 with input layer, hidden layer and output layer. The output layer is:

[0033]

[0034] In the formula, Represents the input neural network weights; the output of the output layer represents the damping coefficient D, virtual inertia J and angular velocity compensation; and then the corresponding output function can be expressed as:

[0035]

[0036] in, is the output function of the output layer;

[0037] The evaluation function of this neural network is:

[0038]

[0039] The gradient descent method is used to adjust the weights in the hidden layer, and an inertia link is added to improve the optimization speed of the algorithm.

[0040]

[0041] Where ω represents the actual angular velocity of the VSG, ω0 represents the rated angular velocity of the power grid; η and α represent the learning rate and inertia coefficient of the network, respectively.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention provides an improved RBF grid-connected converter control method with active support capability, which can respond to voltage deviation signals and compensate for frequency phase through an improved phase self-tracking pre-synchronization method without a phase-locked loop, and adjust the system's moment of inertia and damping coefficient in real time to respond to changes in system status, thereby reducing the impact of disturbances on system frequency and output active power, improving the system's dynamic response process, reducing the system output power distribution error and overshoot caused by the rotor motion equation, and improving the system's grid-connected performance and stability. The control strategy can effectively track the phase and seamlessly switch between off-grid and grid-connected, thereby improving the grid's ability to accept renewable energy power generation units. In order to achieve the above objectives, the present invention adopts the following technical solutions.

[0044] The control method of the present invention has a simple control structure, is easy to implement, and can solve the system output power angle error caused by interference. It improves the grid-connected control accuracy, actively supports the system by improving damping and inertia through RBF, accelerates the system control response speed, effectively solves the instability problem caused by grid connection in the power grid, and improves the stability of the new power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The invention is an inverter grid-connected synchronization process which only improves the VSG active control loop in the prior art.

[0046] Figure 2 The present invention provides an improved RBF grid-connected converter grid-connected control structure topology diagram with active support capability.

[0047] Figure 3 It is a schematic diagram of the RBF neural network structure of the present invention.

[0048] Figure 4 It is a schematic diagram of the working principle of the RBF neural network of the present invention.

[0049] Figure 5 It is a block diagram of the non-phase-locked loop pre-synchronization control based on VSG of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Figure 2 The invention provides a grid-connected control structure topology diagram of an improved RBF grid-connected converter with active support capability, which comprises a coordinate conversion part, a voltage and current dual closed-loop control part, an active power control part, a virtual excitation adjustment part and an RBF secondary control part in sequence; wherein the initial voltage value of the VSG rotor motion equation is input through the coordinate change part, and the VSG synchronizes the phase and amplitude of the inverter and the grid side through the grid-active power control loop and the grid-excitation adjustment loop, and a neural network controller and a grid-connected pre-synchronization control are used in the grid-connected control system to cooperate, collect the initial value information of damping inertia and angular momentum and bring it into the evaluation function iteration, couple the damping coefficient, virtual inertia and phase adjustment feedback amount with parameter relationship, and then output and feed back to the VSG equation, perform voltage and current dual closed-loop control and PWM modulation, and complete synchronous grid-connected control.

[0052] Figure 4 This is the working principle diagram of the RBF neural network. First, a set of initial values ​​within the range of the system virtual inertia J and equivalent damping coefficient D0 parameters are randomly selected; then the initial angular frequency ω is determined based on the initial values ​​and the swing equation, and the evaluation function E(k) is used to determine whether to update the weights; then the values ​​of the virtual inertia J, damping coefficient D and damping compensation ΔD are adaptively adjusted according to the neural network to see whether the three fall within the given value range, and they are compared with the corresponding maximum and minimum values, and the logical value is output to obtain the value at the next moment; by the formula: The equivalent damping coefficient D0 is calculated, and the virtual inertia J and the equivalent damping coefficient D0 are returned to the swing equation to complete the next closed-loop control.

[0053] Figure 5 It is a non-phase-locked loop pre-synchronization control block diagram based on VSG. The phase part includes: by sampling the current and voltage on the grid side, adjusting the output voltage phase of the grid-connected inverter: converting the grid voltage through 3s / 2s to obtain u ga and u gb Then, it is input into 2s / 2r transformation with the virtual synchronous generator active power loop output phase angle θ to obtain the inverter d-axis and q-axis components u gq 、u gd and the grid-side reference value After comparison, the difference is PI-regulated, and the Δω output by the PI regulator is compensated to the active power loop of the VSG; the amplitude part includes: the grid voltage of the energy storage converter is transformed by 3s / 2s, and then u is obtained by the voltage amplitude calculation module. gm ; Taking the grid voltage amplitude u gm For a given, the inverter output voltage amplitude u m The difference is fed back by PI regulation, and the output Δu of the PI regulator is compensated to the voltage given by the reactive power loop of the virtual synchronous generator, so that the output voltage amplitude of the inverter can be adjusted, and finally the output voltage amplitude of the inverter can track the voltage amplitude of the grid; finally, the grid-connected phase angle θ M and grid-connected amplitude E m Output dual closed-loop control, and finally perform PWM modulation to complete the closed loop.

[0054] The control process is as follows:

[0055] S100: The grid-connected inverter operates in an island state, the active power loop of the virtual synchronous generator obtains the voltage phase θ, and uses its initial damping and virtual inertia parameters as the input of the RBF neural network, and compares the rotor motion equation in real time to control and adjust the control error;

[0056] The rotor motion equation is as follows:

[0057]

[0058] Where J and D represent the moment of inertia and damping coefficient respectively; Pm represents mechanical power; Pe represents electromagnetic power; T m and T e They are the mechanical torque and electromagnetic torque of the synchronous generator respectively;

[0059] The RBF neural network parameters are as follows:

[0060] Neural Network Machine:

[0061]

[0062] Where i = 1, 2, ... 5, b i is the element in the width vector b of the Gaussian basis function, b=[b1,…,b5] T ;

[0063] is the center point vector value of the i-th hidden layer neuron;

[0064] The RBF neural network used uses input variables x(1)=ω, x(2)=dω / dt, and adopts a neural network structure of 2-5-3 with input layer, hidden layer and output layer. The output layer is:

[0065]

[0066] In the formula, Represents the input neural network weights; the output of the output layer represents the damping coefficient D, virtual inertia J and angular velocity compensation; and then the corresponding output function can be expressed as:

[0067]

[0068] in, is the output function of the output layer;

[0069] The evaluation function of this neural network is:

[0070]

[0071] The gradient descent method is used to adjust the weights in the hidden layer, and an inertia link is added to improve the optimization speed of the algorithm.

[0072]

[0073] Where ω represents the actual angular velocity of the VSG, ω0 represents the rated angular velocity of the power grid; η and α represent the learning rate and inertia coefficient of the network respectively;

[0074] S200: The angle θ obtained by the virtual synchronous generator through the active power loop is compared with the electrical angle Δθ of the filter phase shift L Add together to get the modulation wave electrical angle θ of the grid-connected inverter N ; At this time, the electrical angle of the inverter output voltage u after filtering by the LC filter is also θ, which is the same as the angle obtained by the virtual synchronous generator through the active power loop;

[0075] S300: By sampling the current and voltage on the grid side, the output voltage phase of the grid-connected inverter is adjusted; the grid voltage is converted into u by 3s / 2s ga and u gb , the virtual synchronous generator active power loop output phase angle θ is input into 2s / 2r to obtain the inverter d-axis and q-axis components u gq 、u gd , and the grid-side reference value After comparison, the obtained difference is PI regulated, and the Δω output by the PI regulator is compensated to the active power loop of the VSG;

[0076] S400: voltage synchronization adjustment; specific process: the grid voltage of the energy storage converter is converted by 3s / 2s, and then u is obtained through the voltage amplitude calculation module. gm ; Taking the grid voltage amplitude u gm For a given, the inverter output voltage amplitude u m For feedback, the obtained difference is PI regulated, and the output Δu of the PI regulator is compensated to the voltage given by the reactive power loop of the virtual synchronous generator, so that the output voltage amplitude of the inverter can be adjusted, and finally the output voltage amplitude of the inverter is made to track the voltage amplitude of the grid, completing the synchronization of the inverter and the grid side voltage;

[0077] Step 500: Determine the initial phase angle and angular frequency according to the initial value G2(s) and the rotor motion equation; and according to the angular velocity compensation and damping inertia parameters output in the improved pre-synchronization control; determine whether to update the weights by the neural network evaluation function E(k), and then adaptively adjust the virtual inertia J, damping coefficient D and angular velocity deviation Δω according to the neural network, and update the grid-connected angular velocity deviation to make it more suitable for the grid-side value;

[0078] The initial value G2(s) is determined by the following formula:

[0079]

[0080] S600: Determine whether the three fall within a given value range. If so, obtain the value at the next moment. Calculate the equivalent compensation angular momentum by the following formula, return the virtual inertia J, damping coefficient D and phase adjustment feedback to the VSG control equation. Finally, when the grid voltage is consistent with the grid-connected inverter output voltage frequency, electrical angle and amplitude, close the PCC to complete the grid connection.

[0081] According to the expression of active closed-loop transfer function, the fixed angular frequency ω of the VSG system is obtained n and the damping ratio ξ,

[0082]

[0083] In the formula, K ω is the VSG adjustment coefficient, E is the voltage electromotive force;

[0084] The adjustment time of the system based on the damping ratio and frequency disturbance is: By adjusting the time, it is determined whether the damping and inertia iterated by RBF meet the system stability requirements.

Claims

1. An improved RBF grid-connected converter control method with active support capability, characterized in that: The process includes the following: S100: The grid-connected inverter operates in an island state, the active power loop of the virtual synchronous generator obtains the voltage phase θ, and uses its initial damping and virtual inertia parameters as the input of the RBF neural network, and compares the rotor motion equation in real time to control and adjust the control error; The rotor motion equation is as follows: Where J and D represent the moment of inertia and damping coefficient respectively; Pm represents mechanical power; Pe represents electromagnetic power; T m and T e They are the mechanical torque and electromagnetic torque of the synchronous generator respectively; S200: The angle θ obtained by the virtual synchronous generator through the active power loop is compared with the electrical angle Δθ of the filter phase shift L Add together to get the modulation wave electrical angle θ of the grid-connected inverter N ; S300: By sampling the current and voltage on the grid side, the output voltage phase of the grid-connected inverter is adjusted; the grid voltage is converted into u by 3s / 2s ga and u gb , the virtual synchronous generator active power loop output phase angle θ is input into 2s / 2r to obtain the inverter d-axis and q-axis components u gq 、u gd , and the grid-side reference value After comparison, the obtained difference is PI regulated, and the Δω output by the PI regulator is compensated to the active power loop of the VSG; S400: voltage synchronization adjustment; specific process: the grid voltage of the energy storage converter is converted by 3s / 2s, and then u is obtained through the voltage amplitude calculation module. gm ; Taking the grid voltage amplitude u gm For a given, the inverter output voltage amplitude u m For feedback, the obtained difference is PI regulated, and the output Δu of the PI regulator is compensated to the voltage given by the reactive power loop of the virtual synchronous generator, so that the output voltage amplitude of the inverter can be adjusted, and finally the output voltage amplitude of the inverter is made to track the voltage amplitude of the grid, completing the synchronization of the inverter and the grid side voltage; Step 500: Determine the initial phase angle and angular frequency according to the initial value G2(s) and the rotor motion equation; And according to the angular velocity compensation and damping inertia parameters output in the improved pre-synchronization control; the neural network evaluation function E(k) determines whether to update the weight, and then the virtual inertia J, damping coefficient D and angular velocity deviation Δω are adaptively adjusted according to the neural network to update the grid-connected angular velocity deviation to make it more suitable for the grid-side value; The initial value G2(s) is determined by the following formula: S600: Determine whether the three fall within a given value range. If so, obtain the value at the next moment. Calculate the equivalent compensation angular momentum by the following formula, return the virtual inertia J, damping coefficient D and phase adjustment feedback to the VSG control equation. Finally, when the grid voltage is consistent with the grid-connected inverter output voltage frequency, electrical angle and amplitude, close the PCC to complete the grid connection. According to the expression of active closed-loop transfer function, the fixed angular frequency ω of the VSG system is obtained n and the damping ratio ξ, In the formula, K ω is the VSG adjustment coefficient, E is the voltage electromotive force; The adjustment time of the system based on the damping ratio and frequency disturbance is: By adjusting the time, it is determined whether the damping and inertia iterated by RBF meet the system stability requirements.

2. According to the improved RBF grid-connected converter control method with active support capability as described in claim 1, it is characterized in that: The RBF neural network parameters are as follows: Neural Network Machine: Where i = 1, 2, ... 5, b i is the element in the width vector b of the Gaussian basis function, b=[b1,…,b5] T ; is the center point vector value of the i-th hidden layer neuron; The RBF neural network used uses input variables x(1)=ω, x(2)=dω / dt, and adopts a neural network structure of 2-5-3 with input layer, hidden layer and output layer. The output layer is: In the formula, Represents the input neural network weights; the output of the output layer represents the damping coefficient D, virtual inertia J and angular velocity compensation; and then the corresponding output function can be expressed as: in, is the output function of the output layer; The evaluation function of this neural network is: The gradient descent method is used to adjust the weights in the hidden layer, and an inertia link is added to improve the optimization speed of the algorithm. Where ω represents the actual angular velocity of the VSG, ω0 represents the rated angular velocity of the power grid; η and α represent the learning rate and inertia coefficient of the network, respectively.

Images