Network construction type SVG control method based on data-driven optimal self-synchronization

Abstract

The invention discloses a network construction type SVG (static var generator) control method based on data-driven optimal self-synchronization. Obtaining the capacitor voltage of the direct current side of the SVG and the actual voltage and the actual current of the grid-connected point; constructing a DeePC controller input according to the actual voltage of the grid-connected point, the SVG direct current side capacitor voltage and a preset reference value, and obtaining a frequency deviation through the processing of the DeePC controller; adding the frequency deviation and a power grid rated frequency to obtain a controller frequency, and integrating to obtain a modulation wave phase signal; according to the SVG output voltage reference value, the actual voltage is sequentially processed by a voltage outer loop, current amplitude limiting and current inner loop to obtain a modulation wave voltage amplitude signal; and processing the modulated wave phase signal to obtain a modulated wave signal. According to the method, the control sequence is optimized online by directly utilizing the voltage data collected in real time, the mismatching risk of a traditional model is avoided, and a reactive power support solution with better dynamic response and autonomous toughness is provided for a new energy power grid.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of stable grid-connected operation control of a converter, and particularly relates to a grid-forming SVG control method based on data-driven optimal self-synchronization. BACKGROUND

[0002] A static var generator (SVG) is a core device of a flexible AC power transmission system, which dynamically adjusts reactive power through a voltage source converter (VSC) to improve the stability of a power system. The static var generator (SVG) can be generally divided into two types: a grid-following type and a grid-forming type. The grid-following SVG is difficult to provide strong support similar to the grid-forming device due to the restriction of its synchronization mechanism, and the voltage and frequency support capacity is relatively weak. Although the grid-forming SVG can provide stable grid support, it needs to be configured with a super capacitor or an energy storage device on the DC side, which significantly increases the system cost. At the same time, due to the real-time changes of the grid operating state, the traditional control method has the risk of model mismatch, which may induce system oscillation, and therefore, the SVG also has the disadvantage of poor adaptability to the grid.

[0003] The traditional synchronization unit usually relies on a preset bandwidth closed-loop tracking (such as a synchronous reference phase-locked loop), and its design is based on the assumption of ideal grid conditions, but the actual grid has complex working conditions such as impedance variation, background harmonic disturbance and asymmetric fault. The fixed bandwidth leads to a contradiction between dynamic performance and disturbance rejection: increasing the bandwidth can accelerate the tracking of the synchronization signal, but it will amplify the influence of grid voltage distortion and noise, causing phase jitter or even loss of lock; reducing the bandwidth can enhance the steady-state accuracy, but the transient response is slow, which is difficult to meet the demand for fast frequency support under high proportion of new energy access. More importantly, the traditional synchronization unit lacks the sensing ability of the grid impedance characteristics. When the inverter is connected to a weak grid (high grid impedance), the interaction between the phase-locked loop (PLL) loop and the grid impedance will induce subsynchronous oscillation, and the fixed parameters cannot be adjusted online to suppress the stability risk. In addition, during the grid fault (such as voltage sag), the traditional unit may produce frequency offset error or synchronization failure due to the lack of adaptive phase-locked strategy, which leads to inaccurate action of the subsequent current control loop and aggravates the transient instability of the system. In comparison, the optimal adaptive synchronization unit realizes a generational breakthrough through online parameter identification and dynamic optimization control. Its core advantage lies in real-time sensing of grid dynamic characteristics. The adaptive synchronization unit solves the systematic defects of the traditional synchronization unit in dynamic response, impedance adaptability and fault resilience, and provides key technical support for new power system core devices such as grid-forming converters.

[0004] The traditional proportional-integral (PI) based converter control method has systematic limitations in dealing with modern power grids with high penetration of renewable energy. The core deficiency lies in the dependence on accurate mathematical models: the controller design is based on the linear time invariant (LTI) model of the controlled object, while the actual grid-connected inverter operates in a dynamically changing grid environment, influenced by multiple factors such as grid impedance fluctuations, adjacent power electronic equipment coupling, and nonlinear loads, making it difficult to establish an accurate model. In engineering practice, PI parameters are usually set by frequency domain analysis or trial and error method under the simplified single-machine infinite system model, and when the actual grid dynamics (especially in high power electronic penetration scenarios) deviate from this idealized model, serious model mismatch problems will occur, leading to controller performance degradation or even instability, such as the widely observed subsynchronous oscillation in wind farms. In theory, incorporating the real grid model into the control design can solve the mismatch problem. However, the real grid has high dimensionality, variability, and many uncertainties, and its accurate model is usually unknown. Although robust control can be used to deal with grid uncertainties and unknown dynamics, it is extremely challenging to design a fixed controller that is robust to all potential scenarios. In contrast, data-driven predictive control (DeePC) provides a fundamental solution through behavior system theory. Its core advantage is model-free optimization: it directly uses real-time collected voltage, current, and other input / output data to represent the closed-loop system dynamics, avoiding the dependence on prior mathematical models and eliminating the risk of model mismatch from the root. In view of the high cost of traditional SVG, the lack of grid support ability, poor adaptability, and model mismatch of synchronous units, it is urgent to develop optimal adaptive synchronous units based on data-driven control.

[0005] The traditional grid-connected SVG control paradigm cannot meet the autonomous support capabilities required by new power systems. Its grid support capability is relatively weak: when a severe grid fault occurs (such as a voltage drop to 0.2 pu), although the grid-connected control will instruct the converter to increase reactive power output to support the voltage, the PLL itself is at risk of instability or unlocking under abnormal operating conditions, leading to deterioration of control performance. More importantly, its grid-connected characteristic means it cannot autonomously establish and maintain voltage stability without a stable grid voltage reference, unlike grid-connected converters, thus limiting its support capability and reliability under extreme conditions; transient response lag: the fixed-parameter current closed-loop control has a long response delay, and in weak grids, it is prone to interaction with grid impedance, inducing subsynchronous oscillations. Furthermore, the dynamic support performance of the grid-connected SVG is greatly affected by its own control parameters. While grid-connected SVGs can provide stable grid support, they require the configuration of supercapacitors or energy storage devices on the DC side, significantly increasing system costs. Both grid-following SVG and grid-building SVG have the disadvantage of poor adaptability to the power grid. When the real-time operating conditions of the power grid change, traditional control may suffer from model mismatch, which may lead to system instability.

[0006] The present invention provides a data-driven network-based SVG with optimal self-synchronization, achieving both optimal self-synchronization and network construction functions. It employs behavioral systems theory to construct data-driven predictive control (DeePC), directly utilizing real-time acquired voltage data to optimize the control sequence online, thus avoiding the mismatch risk of traditional models. This integrated "sensing-decision-network construction" paradigm provides a better dynamic response and autonomous resilience reactive power support solution for high-proportion renewable energy power grids. Simultaneously, this model-free approach minimizes dependence on physical storage components, improving system performance and preventing oscillations. Summary of the Invention

[0007] To address the shortcomings of traditional methods in dealing with the variability and numerous uncertainties of real power grids, this invention provides a data-driven optimal self-synchronization-based network-type SVG control method.

[0008] The technical solution adopted in this invention is:

[0009] This invention includes the following steps:

[0010] S1. Obtain the capacitor voltage on the DC side of the SVG and the actual voltage and current of the SVG grid connection point in the dq coordinate system;

[0011] S2. Construct the input of the DeePC controller based on the actual voltage of the SVG grid connection point in the dq coordinate system, the DC-side capacitor voltage of the SVG, and the preset reference value of the DC-side capacitor voltage of the SVG. The frequency deviation is obtained after processing by the DeePC controller.

[0012] S3. Add the frequency deviation to the preset rated frequency of the power grid to obtain the controller frequency, and integrate the controller frequency to obtain the modulation wave phase signal;

[0013] S4. Subtract the actual voltage from the preset SVG output voltage reference value and input it to the voltage outer loop processing to obtain the current reference value. Then, perform current limiting on the current reference value output by the voltage outer loop to obtain the current reference value of the current inner loop. Subtract the actual current from the current reference value of the current inner loop to obtain the current error, and then input it to the current inner loop processing to obtain the modulated wave voltage amplitude signal.

[0014] S5. The modulation wave signal is obtained by processing the modulation wave phase signal and the modulation wave voltage amplitude signal, and then the driving signal is generated based on the modulation wave signal to realize the control of the SVG.

[0015] The DeePC controller in S2 processes the data according to the following steps:

[0016] S2.1. The difference between the square of the DC-side capacitor voltage of the SVG and the square of the preset reference value of the DC-side capacitor voltage of the SVG is used to obtain the square deviation of the DC-side capacitor voltage. The square deviation of the DC-side capacitor voltage and the q-axis component of the actual voltage at the grid connection point of the SVG are used as the system output, and the frequency deviation is used as the system input.

[0017] S2.2 Apply white noise excitation to the SVG system, use the amplitude of the white noise signal as the frequency deviation, collect the square deviation of the DC side capacitor voltage, the q-axis component and frequency deviation of the actual voltage at the SVG grid connection point for a first preset time length, and then construct the system input sequence and system output sequence in time sequence;

[0018] S2.3. Construct the Hankel matrices of the system input sequence and the system output sequence respectively, and determine whether the Hankel matrix of the system input sequence satisfies the full row rank condition:

[0019] If satisfied, then the Hankel matrices of the system input sequence and the system output sequence are divided into blocks to obtain the block-based Hankel matrices.

[0020] If the condition is not met, proceed to step S2.2;

[0021] S2.4 Collect historical system input data and historical system output data for a second preset time length before the current moment, and then construct the current input sequence and the current output sequence;

[0022] S2.5 Construct an optimization function. Based on the current input sequence, the current output sequence, and the block-based Hankel matrix, construct constraints for the optimization function. Solve the optimization function to obtain the frequency deviation sequence. Select the first value in the frequency deviation sequence as the frequency deviation at the current time.

[0023] S2.3 specifically involves: constructing the system input sequence and system output sequence based on the system input sequence and system output sequence, respectively. Row Hankel matrix: Determine if the Hankel matrix of the system input sequence satisfies the full row rank condition.

[0024] If satisfied, then perform a block partitioning operation, prepending the Hankel matrices of the system input sequence and the system output sequence. The first row is used as the historical input and historical output data matrix, and the last N rows of the Hankel matrix of the system input sequence and system output sequence are used as the future input and future output data matrix, respectively.

[0025] If the condition is not met, proceed to step S2.2.

[0026] The optimization function is set according to the following formula:

[0027]

[0028] in, and Let A and B represent the system's predicted input and predicted output, respectively. and feasible domain, As decision variables, and Let these represent the slack variables of the system's input and output, respectively. As the reference output trajectory of the system, , and They are respectively , and The scaling factor, For regularization terms, Let R represent the weighted L2 norm squared of the predicted input, where R is a predefined positive definite matrix. Let Q represent the weighted L2 norm squared difference between the predicted output and the system's reference output, where Q is a pre-defined positive semi-definite matrix. and Let L2 and L2 represent the squared values ​​of the slack variables of the system's input and output, respectively.

[0029] The constraints of the optimization function are set according to the following formula:

[0030]

[0031] in, , , and These represent data matrices representing historical inputs, historical outputs, future inputs, and future outputs, respectively, with g representing the decision variables. and These represent the current input sequence and the current output sequence, respectively. and Let these represent the slack variables of the system's input and output, respectively. and These represent the system's predicted input and predicted output, respectively.

[0032] The DeePC controller employs a rolling optimization method. After the block-based Hankel matrix is ​​initially constructed, it remains fixed. The frequency deviation at each corresponding time step is obtained by updating the current input sequence and the current output sequence at each time step and solving the optimization function.

[0033] The current limiting in step S4 is handled according to the following formula:

[0034]

[0035] in, and These represent the d-axis and q-axis components of the current reference value output by the outer voltage loop, respectively. and This represents the d-axis and q-axis components of the current reference value of the inner current loop. Indicates the current reference limit.

[0036] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the data-driven optimal self-synchronization-based network-type SVG control method.

[0037] A computer-readable storage medium storing a computer program for executing the data-driven optimal self-synchronization-based network-type SVG control method.

[0038] This invention embeds a data-driven predictive (DeePC) controller into an SVG (Static Var Generator), forming a data-driven grid-connected SVG (DDG-SVG) that achieves both grid connection and optimal adaptive synchronization. The DDG-SVG can autonomously establish and support grid voltage and frequency, providing inertial response and voltage stabilization without relying on additional energy storage devices. Furthermore, the DDG-SVG achieves optimal self-synchronization through a data-driven approach, eliminating the need for an explicit system model. This allows for dynamic response to system parameter changes, ensuring continuous and reliable converter operation over time. This model-free control method significantly reduces reliance on physical energy storage components while enhancing system dynamics and anti-interference capabilities, effectively suppressing oscillations, and improving the robustness of the SVG under varying grid conditions.

[0039] The beneficial effects of this invention are:

[0040] This invention can achieve optimal self-synchronization and network construction functions, while minimizing dependence on physical storage components, improving system performance and preventing oscillations. It overcomes the limitations of traditional model-based systems, which is conducive to the safe and stable operation of the system and also improves the ability to adapt to complex and changing operating conditions and power grid intensity.

[0041] This invention employs behavioral systems theory to construct a data-driven predictive control (DeePC) system, directly utilizing real-time acquired voltage data to optimize the control sequence online, thus avoiding the mismatch risks of traditional models. This integrated "sensing-decision-grid construction" paradigm provides a better dynamic response and autonomous resilience reactive power support solution for high-proportion renewable energy power grids. Simultaneously, this model-free approach minimizes dependence on physical storage components, improving system performance and preventing oscillations. The DeePC controller uses input and output data to learn system behavior and complete control actions without requiring an explicit model. This SVG with grid construction capabilities eliminates hardware dependencies, achieves optimal adaptive synchronization, and overcomes the limitations of traditional model-based systems. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating the method of the present invention.

[0043] Figure 2 This is a schematic diagram of the data-driven mesh-type SVG control structure of the present invention.

[0044] Figure 3 This is a schematic diagram of the SVG grid-connected system simulation model for Example 1.

[0045] Figure 4The simulation waveforms for the infinite bus voltage drop in Example 1 are shown in (a), where (a) represents the DC capacitor voltage waveform of the SVG grid-connected system, (b) represents the SVG output voltage waveform, (c) represents the SVG output active power waveform, and (d) represents the SVG output reactive power waveform.

[0046] Figure 5 The simulation waveforms for the infinite bus phase step in Example 1 are shown in (a) and (b) respectively. (c) represents the DC capacitor voltage waveform of the SVG grid-connected system, (d) represents the SVG output voltage waveform, and (c) represents the SVG output active power waveform.

[0047] Figure 6 The simulation waveforms of the short-circuit ratio of the simulation verification system in Example 1 are shown below. (a) represents the DC capacitor voltage waveform of the SVG grid-connected system, (b) represents the SVG output voltage waveform, (c) represents the SVG output active power waveform, and (d) represents the SVG output reactive power waveform.

[0048] Figure 7 This is a schematic diagram of the simulation model of the dual-machine grid-connected system in Example 2.

[0049] Figure 8 The following are simulation waveforms for the infinite bus voltage variation in Example 2, where (a) represents the DC capacitor voltage waveform of the SVG, (b) represents the output voltage waveform of the SVG, (c) represents the active power waveform of the SVG output, and (d) represents the active power waveform of the converter with grid control. Detailed Implementation

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of this invention.

[0052] like Figure 1 As shown, this embodiment includes the following steps:

[0053] S1. Obtain the capacitor voltage on the DC side of the SVG and the actual voltage and current of the SVG grid connection point in the dq coordinate system;

[0054] Specifically, it acquires the SVG output voltage and SVG DC-side capacitor voltage information, obtains the controller frequency through the DeePC controller, and uses the controller frequency integral as the phase input for the Park change and the phase signal of the modulation wave; it also acquires the SVG output current information and performs Park change to obtain the corresponding... Axial components;

[0055] Specifically, obtain the DC-side capacitor voltage of the SVG. The voltage at the SVG grid connection point is collected by a voltage transformer and used as the SVG output voltage. .right By performing a Park transformation, the output voltage of the SVG is obtained. Axis components:

[0056]

[0057] In the formula, For controller frequency, , and SVG output voltage The voltages of phases A, B, and C, and These are the output voltages of the SVG. Axial components.

[0058] S2. Construct the input of the DeePC controller based on the actual voltage of the SVG grid connection point, the DC-side capacitor voltage of the SVG, and the preset reference value of the DC-side capacitor voltage of the SVG. The frequency deviation is obtained after processing by the DeePC controller.

[0059] S3. Add the frequency deviation to the preset rated frequency of the power grid to obtain the controller frequency, and integrate the controller frequency to obtain the modulation wave phase signal;

[0060] S4. Subtract the actual voltage from the preset SVG output voltage reference value and input it to the voltage outer loop processing to obtain the current reference value. Then, perform current limiting on the current reference value output by the voltage outer loop to obtain the current reference value of the current inner loop. Subtract the actual current from the current reference value of the current inner loop to obtain the current error, and then input it to the current inner loop processing to obtain the modulated wave voltage amplitude signal.

[0061] S5. The modulation wave signal is obtained by processing the modulation wave phase signal and the modulation wave voltage amplitude signal, and then the driving signal is generated based on the modulation wave signal to realize the control of the SVG.

[0062] Specifically, a modulation wave signal is constructed by combining the obtained modulation wave phase signal and modulation wave voltage amplitude signal, thereby generating an SVPWM pulse modulation wave signal that acts on the SVG to complete the control of the data-driven mesh SVG. The specific steps include:

[0063] Consider the modulated wave phase signal obtained in 1). The modulated wave voltage amplitude signal obtained from (and 4) , Perform an inverse Park transformation to obtain a three-phase modulated wave signal in the abc coordinate system:

[0064]

[0065] In the formula, , and The modulated wave signal is in the abc coordinate system. and These are the modulated wave signals. shaft and Axial components.

[0066] The three-phase modulated wave signal is applied to the SVPWM generator to generate the pulse signal used to control the SVG, thus completing the control of the data-driven mesh SVG.

[0067] The DeePC controller in S2 is processed according to the following steps:

[0068] S2.1. The difference between the square of the DC-side capacitor voltage of the SVG and the square of the preset reference value of the DC-side capacitor voltage of the SVG is used to obtain the square deviation of the DC-side capacitor voltage. The square deviation of the DC-side capacitor voltage and the q-axis component of the actual voltage at the grid connection point of the SVG are used as the system output, and the frequency deviation is used as the system input.

[0069] Specifically, the difference between the square of the actual value and the square of the reference value of the DC-side capacitor voltage of the SVG is calculated and the SVG output voltage is... The axis component serves as the output of the SVG grid-connected system (referred to as the system) and is defined as the input of the DeePC controller, i.e.

[0070]

[0071] In the formula, and They are respectively At this moment, the two inputs of the DeePC controller, This is the actual value of the DC-side capacitor voltage of the SVG. This is a reference value for the DC-side capacitor voltage of the SVG. For SVG output voltage Axis components. Let the number of system outputs be . .

[0072] The frequency deviation is taken as the input of the system and defined as the output of the DeePC controller, i.e.

[0073]

[0074] in, For the DeePC controller in Output at any moment Let the frequency deviation be denoted as . Let the number of system inputs be . .

[0075] S2.2 Apply white noise excitation to the SVG system, use the amplitude of the white noise signal as the frequency deviation, collect the square deviation of the DC side capacitor voltage, the q-axis component and frequency deviation of the actual voltage at the SVG grid connection point for a first preset time length, and then construct the system input sequence and system output sequence in time sequence;

[0076] Specifically, let as well as satisfy , The number of system state variables. To predict the length, The lengths of the system's most recent input and output trajectories. The number of system state variables. Represents a set of integers. (Through...) Injecting white noise perturbation signals into the SVG allows for the measurement of timing lengths from the system. Input trajectory and output trajectory ,in , , Let represent the set of real numbers. (Note: The original text contains some formatting errors and inconsistencies. A more accurate , ,but and It can be represented as:

[0077]

[0078] S2.3 Construct the system input sequence and system output sequence respectively based on the system input sequence and system output sequence. Row Hankel matrix: Determine if the Hankel matrix of the system input sequence satisfies the full row rank condition.

[0079] If satisfied, then the Hankel matrices of the system input sequence and the system output sequence are divided into blocks to obtain the block-based Hankel matrices.

[0080] If the condition is not met, proceed to step S2.2;

[0081] Specifically, using and Build The order Hankel matrix, i.e.

[0082]

[0083]

[0084] White noise perturbation signal is used to ensure the input signal yes The order of continuous incentives, i.e. of The order Hankel matrix is ​​of full row rank:

[0085]

[0086] In the formula, When the matrix is ​​of full row rank, the input signal yes It provides continuous motivation.

[0087] Divide the Hankel matrix into and ,in , , , ,and .

[0088] S2.4 Collect historical system input data and historical system output data for a second preset time length before the current moment, and then construct the current input sequence and the current output sequence;

[0089] Specifically, in At any given time, the collection length is The system's most recent input trajectory and output trajectory, i.e., the system's most recent input trajectory:

[0090]

[0091] The system's most recent output trajectory:

[0092]

[0093] Construct a matrix using the system's most recent input and output trajectories. ,Right now

[0094]

[0095] In the formula, the symbol express Frequency deviation at time, express The square of the DC-side capacitor voltage at time t. express The q-axis component of the SVG output voltage at time t.

[0096] S2.5 Construct an optimization function. Based on the current input sequence, the current output sequence, and the block-based Hankel matrix, construct constraints for the optimization function. Solve the optimization function to obtain a frequency deviation sequence with a predicted time domain length of N. Select the first value in the frequency deviation sequence as the frequency deviation at the current time.

[0097] S2.3 specifically involves constructing the system input sequence and system output sequence based on the system input sequence and system output sequence, respectively. Row Hankel matrix: Determine if the Hankel matrix of the system input sequence satisfies the full row rank condition.

[0098] If satisfied, then perform a block partitioning operation, prepending the Hankel matrices of the system input sequence and the system output sequence. The first row is used as the historical input and historical output data matrix, respectively. The last N rows of the Hankel matrix of the system input sequence and system output sequence are used as the future input and future output data matrices, respectively. The block-based Hankel matrix is ​​composed of the historical input, future input, historical output, and future output data matrices. The preset length of the historical system input data and historical system output data is the second preset time length in S2.4, and N represents the preset prediction time domain length.

[0099] If the condition is not met, proceed to step S2.2.

[0100] The optimization function should be set according to the following formula:

[0101]

[0102] in, and Let A and B represent the system's predicted input and predicted output, respectively. and feasible domain, As decision variables, and Let these represent the slack variables of the system's input and output, respectively. As the reference output trajectory of the system, , and They are respectively , and The scaling factor, For regularization terms, Let R represent the weighted L2 norm squared of the predicted input, where R is a predefined positive definite matrix. Let Q represent the weighted L2 norm squared difference between the predicted output and the system's reference output, where Q is a pre-defined positive semi-definite matrix. and Let L2 and L2 represent the squared values ​​of the slack variables of the system's input and output, respectively.

[0103] For vectors quadratic form , =u or yr, P is the cost matrix, P=R or Q, R and Q are positive definite and positive semi definite matrices respectively; The second norm of vector S The square of S = or , = ; , and They are respectively , and The scaling factor.

[0104] The constraints of the optimization function are set according to the following formula:

[0105]

[0106] in, , , and These represent data matrices representing historical inputs, historical outputs, future inputs, and future outputs, respectively, with g representing the decision variables. and These represent the current input sequence and the current output sequence, respectively. and Let these represent the slack variables of the system's input and output, respectively. y and y represent the system's input and output, respectively.

[0107] The optimal control sequence can be obtained by solving the optimization function. .

[0108] in, It is based on and Constructed The order Hankel matrix, where . It is a decision variable. and It is the system's most recent input and output vector, while and It predicts the trajectory. For predicting length. Slack variables. , (by , (Penalties) are used to relax data consistency. And... It is by Scaling regularization term (usually The square of the Euclidean norm Cost matrix and These are positive definite matrices and semi-positive definite matrices, respectively, used to penalize control error and tracking error. This is the reference value for the output trajectory. Input and output constraints are determined by... and definition.

[0109] The DeePC controller employs a rolling optimization method. After the Hankel matrix is ​​constructed in blocks, it remains fixed. The frequency deviation at each corresponding time step is obtained by updating the current input sequence and the current output sequence at each time step and solving the optimization function.

[0110] The current limiting in step S4 is handled according to the following formula:

[0111]

[0112] in, and These represent the d-axis and q-axis components of the current reference value output by the outer voltage loop, respectively. and This represents the d-axis and q-axis components of the current reference value of the inner current loop. Indicates the current reference limit.

[0113] exist At all times Input into the system, where , To control the time domain and calculate the controller frequency:

[0114]

[0115] In the formula, For controller frequency, The rated frequency of the power grid. This represents the frequency deviation.

[0116] 4) Replace with Time, collect again The time length is The system's most recent input and output trajectories are used to reconstruct the system. The optimization problem is then resolved, and the controller frequency is recalculated. In other words, optimal self-synchronization of the SVG is achieved through a rolling optimization approach.

[0117] Integrating the controller frequency yields the phase input of the Park variation and the phase signal of the modulated wave:

[0118]

[0119] In the formula, For controller frequency, The phase input is Park's variation and the phase signal of the modulated wave is also included.

[0120] Step S4 is as follows:

[0121] The SVG uses an L-type filter for filtering. In a three-phase abc coordinate system, the current of the filter inductor on the SVG side is collected by a current transformer as the output current of the SVG. The SVG output current is then transformed to the dq coordinate system using the following Park transformations. , .

[0122]

[0123] In the formula, For controller frequency, , and The output current of the SVG is respectively The currents of phase A, phase B, and phase C.

[0124] Using the output current and voltage of the SVG as controlled variables, PI controllers are designed to control the inner current loop and outer voltage loop respectively. The output of the outer voltage loop, after current limiting, is used as the reference value for the inner current loop, thus completing the overall control loop of the outer voltage loop and the inner current loop. The specific steps include:

[0125] 1) Design a PI controller to control the inner current loop:

[0126] The SVG uses an L-type filter, and the filter inductor on the SVG side is... Filter inductor The equivalent resistance is The line inductance is The SVG output current is The SVG output voltage is The SVG modulated wave signal is According to the circuit law KVL, the filter inductance on the SVG side is obtained. Electrical model:

[0127]

[0128] Performing Park transformations yields Model in coordinate system:

[0129]

[0130] By performing a Laplace transform, we obtain the frequency domain model:

[0131]

[0132] It can be seen that the SVG output current is Coupling exists in the coordinate system, which can be resolved by adding a quantity that is the opposite of the coupling term to the controller. To facilitate the design of a PI controller, coupling can be added to the controller to achieve axis decoupling. shaft current Add coupling to the output of the PI controller ;exist shaft current Add coupling to the output of the PI controller This allows for the decoupling of the dq axis.

[0133] 2) Design a PI controller to control the outer voltage loop:

[0134] Voltage outer loop control output is

[0135]

[0136] In the formula, and The voltage outer loop control output is used. and These are the output voltages of the SVG. shaft and Reference value for the axis. and For SVG output voltage shaft and Axial components, and These are the proportional and integral coefficients of the voltage outer loop PI controller, respectively.

[0137] 3) The reference value for the inner current loop is obtained by current limiting the voltage outer loop control output:

[0138]

[0139] In the formula, and This is the reference value for the inner current loop. For current reference limiting, when the current reference amplitude is less than (or equal to) When the current reference amplitude given by the outer voltage loop is greater than a certain value, the limiting strategy does not work. At that time, the actual current reference amplitude given to the inner current loop is limited to... Furthermore, priority is given to increasing the d-axis current to meet the needs of active power transmission.

[0140] By using the voltage outer loop control output, after current limiting, as the reference value for the current inner loop, the overall control loop of the voltage outer loop and the current inner loop can be completed.

[0141] Set the SVG output voltage reference value, DC side capacitor voltage reference value, grid rated frequency value, DeePC controller parameters, and PI controller parameters to establish a data-driven network-type SVG. Obtain the modulated wave voltage amplitude signal through current inner loop control. Specific steps include:

[0142] 1) Set the SVG output voltage reference value and DC side capacitor voltage reference value Rated frequency value of power grid DeePC controller parameters: The timing lengths of the input and output trajectories are measured from the system by injecting white noise perturbation signals into the SVG. The time series length of the system's most recent input and output trajectories. Predicted length Control Time Domain Reference for system output trajectory Cost matrix and Regularization weights , and .

[0143] 2) Design the PI controller parameters for the voltage outer loop and current inner loop to meet the system stability and speed requirements:

[0144] The transfer function of the PI controller is:

[0145]

[0146] Two parameters of the PI controller and The selection of the phase margin (or frequency) primarily considers system stability and dynamic response. For system stability, the phase margin of the open-loop transfer function should be greater than 0°, and a certain stability margin is still required after the system stabilizes. The bandwidth of the closed-loop transfer function determines the speed of the system response. For a stable negative feedback system, the bandwidth of the closed-loop transfer function is greater than the cutoff frequency of the open-loop transfer function, but the difference is not significant, and they can be used interchangeably in engineering.

[0147] Selected based on the system's rapid response requirements. The PI controller provides a small phase hysteresis selection. Simultaneously, the bandwidth of the inner current loop is designed to be approximately equal to the switching frequency. of The voltage outer loop bandwidth is approximately equal to the current inner loop bandwidth. .

[0148] 3) The modulated wave voltage amplitude signal obtained through the inner current loop control is:

[0149]

[0150] In the formula, and These are the modulated wave voltage amplitude signals. shaft and Axial components. and These are the proportional and integral coefficients of the current inner-loop PI controller. and This is the reference value for the inner current loop. and For SVG output current shaft and Axial components. This is the filter inductor for the SVG side. and For SVG output voltage shaft and Axial components, This refers to the controller frequency.

[0151] like Figure 1As shown, the process begins by measuring the SVG output voltage and obtaining the DC-side capacitor voltage information using a voltage transformer. The controller frequency is then acquired via a DeePC controller, serving as the phase input for Park changes and the phase signal for the modulation wave, providing a foundation for subsequent control. The SVG output current is measured using a current transformer, and the corresponding dq-axis components are obtained through Park changes. Using the SVG output current and voltage as controlled variables, PI controllers are designed to control the inner current loop and outer voltage loop respectively. The voltage outer loop control output, after current limiting, serves as the reference value for the inner current loop, completing the overall control loop for both the outer and inner current loops. The SVG output voltage reference value, DC-side capacitor voltage reference value, grid rated frequency, PI controller parameters, and DeePC controller parameters are set to establish a data-driven networked SVG. The modulation wave voltage amplitude signal is obtained through the inner current loop control. Combining the obtained modulation wave phase signal and modulation wave voltage amplitude signal, a modulation wave signal is constructed, generating an SVPWM pulse modulation wave signal acting on the SVG to complete the control of the data-driven networked SVG. Specific details of each control design can be found in [link to relevant documentation]. Figure 2 .

[0152] Figure 2 This is a control block diagram of the data-driven mesh-type SVG of the present invention. Figure 2 middle For SVG side filter inductor, For line inductance, For SVG output voltage, For SVG output current, For modulated wave voltage, This is the DC-side capacitor voltage of the SVG.

[0153] Example 1:

[0154] Reference Figure 3 This is the first embodiment of the present invention, which is an SVG stand-alone grid-connected simulation system. Five different control strategies are employed to control the SVG: Grid-following control (GFL), Grid-building control (GFM), Virtual Synchronization Control (ViSynC), DC capacitor voltage synchronization control (DDG1), and hybrid synchronization control of DC capacitor voltage and SVG output voltage (DDG2). DDG2 refers to the data-driven grid-building SVG with optimal self-synchronization functionality. A simulation model of the SVG stand-alone grid-connected system is established based on the voltage and current control strategies, and the relevant simulation parameters are set as shown in Table 1.

[0155] Table 1. Relevant parameters of the SVG grid-connected system in the simulation verification of the embodiment.

[0156]

[0157] Among them, the SVG output voltage reference value and DC side capacitor voltage reference value Rated frequency value of power grid The parameters of the current loop PI controller are: , The voltage loop PI controller parameters are as follows: , DeePC control parameters: The timing lengths of the input and output trajectories are measured from the system by injecting white noise disturbance signals into the SVG. The time series length of the system's most recent input and output trajectories. Predicted length Control Time Domain Reference for system output trajectory , Represents column vectors and cost matrices with elements of 0 and 0. and Regularization weights , and .in, Indicates the order is The identity matrix, when controlled by DDG1. When using DDG2 control, express The matrix is ​​a diagonal matrix with diagonal elements of 2000 and 250.

[0158] To operate the SVG in grid-connected mode, set the voltage amplitude of the infinite bus to decrease by 0.05 pu at t = 0.2 s. Run the simulation and record the DC capacitor voltage, SVG output voltage, SVG output active power, and SVG output reactive power of the SVG grid-connected system. Figure 4 As shown.

[0159] Set the phase of the infinite bus to jump by 20° at t = 0.2 s, run the simulation, and record the DC capacitor voltage, SVG output voltage, SVG output active power, and SVG output reactive power of the SVG grid-connected system as follows: Figure 5 As shown.

[0160] Set the system short-circuit ratio (SCR) to decrease from 1.2 to 1.0 at t = 0.2 s. Run the simulation and record the DC capacitor voltage, SVG output voltage, SVG output active power, and SVG output reactive power of the SVG grid-connected system as follows: Figure 6 As shown.

[0161] Through observation Figure 4The results show that the data-driven grid-type SVG (DDG2) can achieve grid construction and optimal self-synchronization under steady-state conditions. Furthermore, it can actively increase its reactive power output when the grid voltage drops, thus providing voltage support for the system. Firstly, in terms of DC capacitor voltage response, DDG2 exhibits smaller fluctuations and faster recovery, indicating good DC-side stability and more effectively suppressing fluctuations caused by voltage spikes. Secondly, during AC voltage amplitude recovery, the system controlled by DDG2 shows smaller overshoot and a smoother transition without significant oscillations. In contrast, GFL, GFM, and ViSynC all exhibit varying degrees of overshoot or oscillations. While DDG1 control shows some improvement over traditional methods, it is still not as smooth as DDG2's response. Regarding power response, DDG2 demonstrates faster recovery and stronger robustness in both active and reactive power regulation. Its reactive power output exhibits smaller fluctuations and faster recovery during transient processes, helping to support system voltage more quickly and improve the power quality of the grid. It is evident that DDG2 exhibits superior dynamic stability, faster response speed, and better operational reliability during voltage surges, making it suitable for applications with high requirements for voltage and power quality.

[0162] Through observation Figure 5 The results show that after a sudden phase change in the infinite bus voltage, DDG2 exhibits less fluctuation and smoother regulation during the DC-side voltage recovery process, without significant overshoot or continuous oscillation, demonstrating excellent anti-interference capability and voltage stability. In AC voltage and power response, the DDG2-controlled system also demonstrates faster tracking speed and stronger robustness, with smoother changes in active and reactive power, and a significantly shorter time to re-establish equilibrium compared to GFL, GFM, and ViSynC control methods. Especially compared to DDG1 control, DDG2, with its hybrid synchronization mechanism, better coordinates the dynamic relationship between voltage and internal potential, effectively suppressing power and voltage chattering and improving the system's transient stability under phase disturbances. Overall, DDG2 control demonstrates superior dynamic quality, less regulation oscillation, and more reliable synchronization performance in typical grid disturbances such as phase steps, making it suitable for applications sensitive to system frequency and phase changes under high-proportion renewable energy integration, and possessing significant engineering application value.

[0163] Through observation Figure 6The results show that in a weak grid disturbance scenario where the system short-circuit ratio (SCR) drops from 1.2 to 1.0, the DDG2 control strategy exhibits significantly better dynamic performance than other control methods. When the system strength suddenly decreases at t=0.2s, DDG2 shows the smallest fluctuation in the DC-side voltage response, with a rapid and smooth recovery process and no sustained low-frequency oscillations, demonstrating extremely strong voltage support capability and DC link stability. On the AC side, the DDG2-controlled port voltage drop is the mildest and can quickly recover to near the reference value, effectively mitigating the risk of voltage instability under weak grid conditions. Compared to the significant active power oscillations and slow recovery issues of GFL and GFM controls, and the large overshoot introduced by the ViSynC strategy, DDG2 demonstrates superior robustness in both active and reactive power regulation: its power response transition is smooth, it re-establishes power balance the fastest, and it exhibits almost no steady-state error. Compared with DDG1, DDG2 further optimizes the synchronization stability of the system under low SCR conditions with its hybrid synchronization mechanism, better suppresses the coupled oscillation of power and voltage, and thus provides more reliable reactive power support and more stable operating characteristics when the grid strength changes abruptly.

[0164] Example 2:

[0165] To further verify that the data-driven grid-type SVG with optimal self-synchronization function of this invention can provide voltage support for traditional grid-type converters, a dual-machine test system for simulation was built, such as... Figure 7 As shown, the SVG is connected in parallel with a grid-controlled converter. The voltage amplitude of the infinite bus is set to decrease by 0.05 pu at t=0.2s and recover after 0.02s. The simulation model is run, and the DC capacitor voltage of the SVG, the grid connection point voltage, the active power output of the SVG, and the active power output of the grid-controlled converter are recorded as follows: Figure 8 As shown.

[0166] Through observation Figure 8The results show that during a brief voltage dip (0.05 pu, lasting 0.02 seconds) on the infinite bus, the DDG2 control strategy exhibits superior dynamic performance, especially in multi-device coupled operation environments. In the DC capacitor voltage response, the SVG controlled by DDG2 not only exhibits minimal fluctuations but also quickly stabilizes after voltage recovery, without the sustained oscillations or deviations seen under control strategies like GFL, demonstrating stronger DC-side anti-interference capabilities and voltage autonomy. Regarding AC voltage support, DDG2 can compensate for system voltage dips more quickly, effectively suppressing bus voltage fluctuations and improving power quality at the PCC point. Particularly noteworthy is the smooth dynamic process and minimal overshoot of the SVG controlled by DDG2 in active power response, and its faster recovery to steady state after fault clearance. It also positively impacts the parallel-operated GFL converter: the active power oscillations of the GFL are significantly reduced and recovery is faster, indicating that DDG2 enhances the stability of multi-power electronic equipment operation by improving the overall dynamic characteristics of the system. Compared to traditional control methods such as GFL and GFM, which exhibit slow response, power oscillation, and slow recovery under fault conditions, DDG2 effectively coordinates internal and external dynamics through its hybrid synchronization mechanism, maintaining good synchronization and power regulation capabilities even under strong disturbances. In summary, DDG2 control demonstrates superior dynamic stiffness, faster recovery speed, and better operational adaptability in dual-machine systems. It not only improves the performance of the SVG itself but also enhances the operational reliability of other power electronic devices in the system, making it highly suitable for multi-device coordinated operation scenarios in highly electronic power grids.

[0167] The specific embodiments described above illustrate the technical solution and beneficial effects of the present invention in detail. It should be understood that the above description is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A data-driven optimal self-synchronization-based network-type SVG control method, characterized in that, Includes the following steps: S1. Obtain the capacitor voltage on the DC side of the SVG and the actual voltage and current of the SVG grid connection point in the dq coordinate system; S2. Construct the input of the DeePC controller based on the actual voltage of the SVG grid connection point in the dq coordinate system, the DC-side capacitor voltage of the SVG, and the preset reference value of the DC-side capacitor voltage of the SVG. The frequency deviation is obtained after processing by the DeePC controller. S3. Add the frequency deviation to the preset rated frequency of the power grid to obtain the controller frequency, and integrate the controller frequency to obtain the modulation wave phase signal; S4. Subtract the actual voltage from the preset SVG output voltage reference value and input it to the voltage outer loop processing to obtain the current reference value. Then, perform current limiting on the current reference value output by the voltage outer loop to obtain the current reference value of the current inner loop. Subtract the actual current from the current reference value of the current inner loop to obtain the current error, and then input it to the current inner loop processing to obtain the modulated wave voltage amplitude signal. S5. The modulation wave signal is obtained by processing the modulation wave phase signal and the modulation wave voltage amplitude signal, and then the driving signal is generated based on the modulation wave signal to realize the control of the SVG.

2. The data-driven optimal self-synchronization-based network-type SVG control method according to claim 1, characterized in that: The DeePC controller in S2 processes the data according to the following steps: S2.

1. The difference between the square of the DC-side capacitor voltage of the SVG and the square of the preset reference value of the DC-side capacitor voltage of the SVG is used to obtain the square deviation of the DC-side capacitor voltage. The square deviation of the DC-side capacitor voltage and the q-axis component of the actual voltage at the grid connection point of the SVG are used as the system output, and the frequency deviation is used as the system input. S2.2 Apply white noise excitation to the SVG system, use the amplitude of the white noise signal as the frequency deviation, collect the square deviation of the DC side capacitor voltage, the q-axis component and frequency deviation of the actual voltage at the SVG grid connection point for a first preset time length, and then construct the system input sequence and system output sequence in time sequence; S2.

3. Construct the Hankel matrices of the system input sequence and the system output sequence respectively, and determine whether the Hankel matrix of the system input sequence satisfies the full row rank condition: If satisfied, then the Hankel matrices of the system input sequence and the system output sequence are divided into blocks to obtain the block-based Hankel matrices. If the condition is not met, proceed to step S2.2; S2.4 Collect historical system input data and historical system output data for a second preset time length before the current moment, and then construct the current input sequence and the current output sequence; S2.5 Construct an optimization function. Based on the current input sequence, the current output sequence, and the block-based Hankel matrix, construct constraints for the optimization function. Solve the optimization function to obtain the frequency deviation sequence. Select the first value in the frequency deviation sequence as the frequency deviation at the current time.

3. The data-driven optimal self-synchronization-based network-type SVG control method according to claim 2, characterized in that: S2.3 specifically involves: constructing the system input sequence and system output sequence based on the system input sequence and system output sequence, respectively. Row Hankel matrix: Determine if the Hankel matrix of the system input sequence satisfies the full row rank condition. If satisfied, then perform a block partitioning operation, prepending the Hankel matrices of the system input sequence and the system output sequence. The first row is used as the historical input and historical output data matrix, and the last N rows of the Hankel matrix of the system input sequence and system output sequence are used as the future input and future output data matrix, respectively. If the condition is not met, proceed to step S2.

2.

4. The data-driven optimal self-synchronization-based network-type SVG control method according to claim 3, characterized in that: The optimization function is set according to the following formula: in, and Let A and B represent the system's predicted input and predicted output, respectively. and feasible domain, As decision variables, and Let these represent the slack variables of the system's input and output, respectively. As the reference output trajectory of the system, , and They are respectively , and The scaling factor, For regularization terms, Let R represent the weighted L2 norm squared of the predicted input, where R is a predefined positive definite matrix. Let Q represent the weighted L2 norm squared difference between the predicted output and the system's reference output, where Q is a pre-defined positive semi-definite matrix. and Let L2 and L2 represent the squared values ​​of the slack variables of the system's input and output, respectively.

5. The data-driven optimal self-synchronization-based network-type SVG control method according to claim 1, characterized in that: The constraints of the optimization function are set according to the following formula: in, , , and These represent data matrices representing historical inputs, historical outputs, future inputs, and future outputs, respectively, with g representing the decision variables. and These represent the current input sequence and the current output sequence, respectively. and Let these represent the slack variables of the system's input and output, respectively. and These represent the system's predicted input and predicted output, respectively.

6. The data-driven optimal self-synchronization-based network-type SVG control method according to claim 3, characterized in that: The DeePC controller employs a rolling optimization method. After the block-based Hankel matrix is ​​initially constructed, it remains fixed. The frequency deviation at each corresponding time step is obtained by updating the current input sequence and the current output sequence at each time step and solving the optimization function.

7. The data-driven optimal self-synchronization-based network-type SVG control method according to claim 1, characterized in that: The current limiting in step S4 is handled according to the following formula: in, and These represent the d-axis and q-axis components of the current reference value output by the outer voltage loop, respectively. and This represents the d-axis and q-axis components of the current reference value of the inner current loop. Indicates the current reference limit.

Images