Input incremental optimal control method of a converter
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2026-05-18
- Publication Date
- 2026-06-19
AI Technical Summary
Traditional grid-connected converters are prone to instability under grid disturbances, have limited dynamic support capabilities, and lack adaptability in fixed parameter control strategies, leading to risks such as voltage and current oscillations and subsynchronous oscillations.
The converter adopts an input incremental optimal control method, which calculates active and reactive power by acquiring current and voltage signals, uses a phase-locked loop to acquire frequency signals, constructs an input incremental controller for adaptive adjustment, including offline preparation and online operation stages, uses Hankel matrix and singular value decomposition to determine the system order, uses gradient descent method to optimize controller gain, and generates modulation wave signals for control.
When the grid short-circuit ratio decreases and voltage drops, the stability and dynamic support capability of the converter are enhanced, power oscillations are suppressed, the robustness and dynamic response speed of the system are improved, and the operating performance under complex grid conditions is improved.
Smart Images
Figure CN122246891A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of stable grid-connected operation control technology for converters, and in particular to an optimal input increment control method for converters. Background Technology
[0002] Power electronic converters, as core interface equipment for grid connection of new energy sources (wind power, photovoltaics, etc.), have become key supporting equipment for building new power systems and promoting the green and low-carbon transformation of the energy structure. Currently, wind power, photovoltaics, and other new energy sources mainly achieve large-scale grid connection through traditional grid-connected converters. These converters generally adopt a dual closed-loop control architecture of "power outer loop + current inner loop," relying on phase-locked loops (PLLs) to track the phase and frequency of the grid voltage in real time, thereby achieving synchronous operation with the grid and achieving stable operation under normal grid conditions. However, when the grid encounters disturbances such as reduced short-circuit ratio, voltage dips, and frequency fluctuations, the inherent limitations of this control strategy can induce voltage and current oscillations, subsynchronous oscillations, and even instability and grid disconnection, seriously threatening the safe and stable operation of the power system.
[0003] Furthermore, traditional grid-connected converters employ a dual-closed-loop control strategy with fixed parameters, which are preset based on specific grid operating conditions. However, actual grid operation is susceptible to fluctuations in renewable energy output and load changes, exhibiting dynamic characteristics such as voltage amplitude fluctuations, frequency shifts, and impedance characteristic variations. Fixed-parameter control strategies lack the ability to adapt to dynamic grid changes. When actual grid characteristics deviate from design conditions, model mismatch issues are easily triggered, leading to increased interaction between the converter and the grid, further exacerbating risks such as voltage and current oscillations, subsynchronous oscillations, and grid disconnection. Summary of the Invention
[0004] This invention addresses the problems of traditional grid-connected converters being prone to instability and having limited dynamic support capabilities under grid shocks such as reduced short-circuit ratios and voltage drops. It proposes an input incremental optimal control method for converters, which can adaptively adjust the control output based on real-time electrical quantity measurement data when grid disturbances occur, thereby enhancing the converter's operational stability and voltage / frequency support capabilities during transient processes.
[0005] The technical solution adopted in this invention is:
[0006] The method of the present invention includes the following steps:
[0007] S1. Obtain the output current and output voltage of the converter in the dq coordinate system, and then calculate the active power and reactive power output by the converter.
[0008] S2. Obtain the control frequency signal through a phase-locked loop and perform an integral operation on the control frequency signal to obtain the phase signal;
[0009] S3. The active and reactive power output from the converter are used as inputs to the incremental controller. After processing by the input incremental controller, the d-axis and q-axis current reference values of the inner current loop are obtained.
[0010] S4. Input the d-axis and q-axis current reference values of the inner current loop into the inner current loop processing to generate the modulated wave voltage amplitude signal;
[0011] S5. Generate a drive signal based on the amplitude and phase signals of the modulated wave voltage and control the converter.
[0012] The input incremental controller includes an offline preparation phase and an online operation phase. The offline preparation phase is executed only once before the converter is put into grid-connected operation, and the online operation phase is executed once in each control cycle after the converter is put into grid-connected operation. The offline preparation phase is processed according to the following steps:
[0013] 1) Use the active and reactive power output from the converter as the system output, and the d-axis and q-axis current reference values of the inner current loop as the system input;
[0014] 2) Apply white noise excitation to the converter, collect the system input and system output at different times, and construct the system output sequence and system input sequence in time order. Then, construct the Hankel matrix based on the system output sequence and system input sequence.
[0015] 3) Perform singular value decomposition on the Hankel matrix to obtain multiple singular values, and determine the system order based on the multiple singular values.
[0016] The input incremental controller processes the active and reactive power output of the converter during the online operation phase, specifically according to the following steps:
[0017] S3.1. The active and reactive power output by the converter are used as the system output, and the d-axis and q-axis current reference values of the inner current loop are used as the system input.
[0018] S3.2 Construct the augmented state for the current moment based on the system order, the system input at the previous moment or the preset zero-moment system input, and the active and reactive power output by the converter.
[0019] S3.3 Construct an expression for the system input increment, and then construct a cost function based on the expression for the system input increment and the augmented state at the current time. With minimizing the cost function as the optimization objective, the gradient descent method is used to update the feedback gain and feedforward gain of the controller in the system input increment expression, and the system input increment at the current time is obtained by solving the problem.
[0020] S3.4. Based on the system input of the previous moment or the preset zero-time system input, add the system input increment of the current moment to obtain the system input of the current moment.
[0021] The expressions for the augmented state and the system input increment are specifically set according to the following formulas:
[0022]
[0023]
[0024]
[0025]
[0026]
[0027]
[0028]
[0029] in, This represents the augmented state at time t. express The time is not the minimum controllable state. Let represent the system input at time t-1, m represent the dimension of the system input, r represent the system order, and T represent the permutation matrix. express Time system input-output trajectory Indicates the order is unit array, The dimension is The zero matrix, The dimension is The zero matrix, Represents the selection matrix. The dimension is The zero matrix, The dimension is The zero matrix, For order is As a unit array, Indicates from Time's up The time length is System input, Indicates from Time's up The time length is The system output, This represents the system input increment at time t. and These represent the feedback gain and feedforward gain of the controller, respectively. Reference values representing the augmented state. Indicates the detection of noise signals. This represents the preset reference value of the system output at time t.
[0030] The cost function of the input incremental controller is specifically set according to the following formula:
[0031]
[0032] in, Let E represent the cost function, N represent the expected value, R represent the penalty matrix for the system input increment, Q represent the penalty matrix for the augmented state, and supremum represent the supremum. This represents the augmented state at time t. Reference values representing the augmented state. This represents the system input increment at time t. This represents the transpose of a matrix.
[0033] In step S3.3, the gradient descent method is used to update the feedback gain and feedforward gain of the controller in the system input increment expression, specifically according to the following steps:
[0034] Step 1: Set initial values for the feedback gain and feedforward gain at the initial moment;
[0035] Step 2: In each control cycle, calculate the system input increment at the current moment based on the feedback gain and feedforward gain at the current moment. After running the system, collect the system output at the current moment, and then update the sample covariance matrix based on the system input increment and system output at the current moment.
[0036] Step 3: Based on the updated sample covariance matrix, the current feedback gain and feedforward gain, calculate the feedback parameterization policy matrix and feedforward parameterization policy matrix for the next time step.
[0037] Step 4: Update the feedback parameterized policy matrix and the feedforward parameterized policy matrix for the next time step using the gradient descent method;
[0038] Step 5: Calculate the feedback gain and feedforward gain for the next time step based on the updated feedback parameterized policy matrix and feedforward parameterized policy matrix.
[0039] The offline preparation phase also includes: constructing a covariance-normalized input increment matrix, a covariance-normalized output matrix, a covariance-normalized current augmented state matrix, a covariance-normalized successor augmented state matrix, and a sample covariance matrix based on the collected system inputs and outputs, as initial values for gradient descent solution in the online running phase.
[0040] A computer-readable storage medium storing program data thereon, which, when executed by a processor, implements an optimal input increment control method for a converter.
[0041] A computer device includes a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the steps of an input incremental optimal control method for a converter.
[0042] The beneficial effects of this invention are:
[0043] This invention enables stable grid-connected operation even under conditions of reduced short-circuit ratio and voltage dips. It prevents converter instability due to disturbances, promoting safe and stable system operation and enhancing the converter's ability to adapt to complex and changing operating conditions and grid strength. When typical disturbances such as reduced short-circuit ratio and frequency dips occur in the grid, the converter's input incremental optimal control method can adjust the control quantity in real time, effectively enhancing the system's damping characteristics. This significantly suppresses power oscillations and synchronous oscillations, improving the converter's stability and dynamic support capabilities.
[0044] This invention eliminates the need for a pre-established precise system analytical model, thus fundamentally overcoming the control performance degradation caused by inaccurate model parameters or structural mismatches. It significantly improves robustness under uncertain environments, providing an effective approach to solving the challenges of stable control and model uncertainty in converters under grid disturbances. The algorithm continuously corrects the control output through online optimization, enabling the converter's active and reactive power to track reference values more quickly and accurately, greatly improving its dynamic response speed and reducing settling time. Therefore, this method effectively improves the dynamic performance and operational robustness of converters under complex grid conditions. Attached Figure Description
[0045] Figure 1 This is a flowchart illustrating the method of this embodiment.
[0046] Figure 2 This is a control block diagram of the optimal input increment control method for the converter of the present invention.
[0047] Figure 3 This is a schematic diagram of the converter grid-connected system simulation model used in Examples 1 and 2.
[0048] Figure 4The following are simulation diagrams of the converter when the converter output power increases by a step and the grid short-circuit ratio decreases in Example 1, where [a] represents the simulation diagram of the converter output active power waveform and [b] represents the simulation diagram of the converter output reactive power waveform.
[0049] Figure 5 The diagram shows the converter simulation when the grid voltage drops in Example 2, where [a] represents the converter output reactive power waveform simulation diagram and [b] represents the converter output voltage waveform simulation diagram. 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, the method in this embodiment includes the following steps:
[0053] S1. Obtain the output current and output voltage of the converter in the dq coordinate system, and then calculate the active power and reactive power output by the converter.
[0054] Voltage and current transformers are used to collect the output current and voltage information of the converter. The collected converter output current and voltage are then subjected to Park transforms to obtain the corresponding values. The shaft components are then used to calculate the active and reactive power output of the converter.
[0055] Specifically, this involves obtaining the converter's output current and output voltage information:
[0056] The converter is connected to the infinite power grid via a filter assembly and transmission lines. This single-unit infinite power system is referred to as the system. The filter assembly consists of a filter inductor on the converter side. Filter inductor The equivalent resistance is The filter capacitor is Filter capacitor Series damping resistor is The voltage at the converter's grid connection point is collected using a voltage transformer as the output voltage. The filter inductance on the converter side is collected through a current transformer. The current is used as the output current. .
[0057] Perform Park transform on the output current and output voltage of the converter to obtain the corresponding... Axis components:
[0058]
[0059]
[0060] In the formula, To control the frequency. , and The output current of the converter is respectively The currents of phases A, B, and C, , and The output voltage of the converter are respectively The voltages of phases A, B, and C, , , , These are the converter output current and output voltage, respectively. Axial components.
[0061] Calculate the active and reactive power output of the converter.
[0062]
[0063] In the formula, and These represent the active power and reactive power output by the converter, respectively. and For the output voltage of the converter Axial components, and For the output current of the converter Axial components;
[0064] S2. The control frequency signal is obtained through a phase-locked loop, and the phase signal is obtained by integrating the control frequency signal, which is used as the phase input of the Park transform and the phase signal of the modulated wave voltage.
[0065] The control frequency signal is obtained through a phase-locked loop, and the input phase of the Park transform and the phase signal of the modulated wave voltage are obtained by integrating the control frequency signal; the specific steps include:
[0066] (1) Obtain the control frequency signal through a phase-locked loop.
[0067]
[0068] In the formula, To control the frequency, The rated angular frequency of the power grid. For the output voltage of the converter Axial components, For the output voltage of the converter Reference values for axis components. For the Laplace operator, and These are the proportional and integral coefficients of the phase-locked loop PI controller, respectively.
[0069] (2) Generate the input phase and modulated voltage signal of the Park transform.
[0070] Integrate the control frequency signal:
[0071]
[0072] In the formula, To control the frequency, The input phase and modulated wave voltage phase signals are Park transform input phase and modulated wave voltage phase signals.
[0073] S3. The active and reactive power output from the converter are used as inputs to the incremental controller. After processing by the input incremental controller, the d-axis and q-axis current reference values of the inner current loop are obtained.
[0074] The modulated wave voltage phase signal and the modulated wave voltage amplitude signal are combined to construct a modulated wave voltage signal, and a pulse width modulation signal is generated by a sinusoidal pulse width modulation (SPWM) generator to act on the power switching device of the converter, thereby realizing the optimal control of the converter's input increment.
[0075] An input incremental controller is constructed to generate a reference value signal for the current inner loop, providing a current reference value for the current inner loop control; the specific steps include:
[0076] The input incremental controller is simply referred to as the controller. The controller's inputs and outputs are configured as follows:
[0077] Configure the controller inputs: the active and reactive power outputs of the converter, i.e.:
[0078]
[0079] In the formula, and For the controller in Input at any time, and record , Represents a column vector. and The converter is in The active and reactive power output at any given time. The controller input is simultaneously used as the system output, and the dimension of the system output is... .
[0080] Configure the controller output: current inner loop reference value signal, i.e.
[0081]
[0082] In the formula, and For the controller in Output at any time, and record , Represents a column vector. and for The constant current inner loop reference value signal. The controller output is simultaneously used as the system input, and the dimension of the system input... .
[0083] The system input increment can be calculated from the system input:
[0084]
[0085] In the formula, and for The system inputs an increment at any given time and records it. , Represents a column vector. and for Time system input, and for Time system input.
[0086] The input incremental controller includes an offline preparation phase and an online operation phase. The offline preparation phase is executed only once before the converter is put into grid-connected operation, while the online operation phase is executed once per control cycle after the converter is connected to the grid. The offline preparation phase is processed according to the following steps:
[0087] 1) The active and reactive power output of the converter are used as the system output, and the d-axis and q-axis current reference values of the inner current loop are used as the system input; the system refers to the converter and the power grid it is connected to as a whole.
[0088] 2) Apply white noise excitation to the converter, collect the system input and system output at different times, and construct the system output sequence and system input sequence in time order. Then, construct the Hankel matrix based on the system output sequence and system input sequence.
[0089] 3) Perform singular value decomposition on the Hankel matrix to obtain multiple singular values, and determine the system order based on the multiple singular values.
[0090] Specifically, in the offline phase: the time scale is... from Time to time
[0091] (1) Inject white noise signals into the system through system input to excite the system, collect the system input trajectory and system output trajectory, and construct the Hankel matrix of the system input trajectory and system output trajectory.
[0092] Collect system input trajectory and system output trajectory:
[0093] , ,
[0094] In the formula, Indicates time, This is the upper bound of the system lag. for The time system input trajectory includes from Time's up The time length is System input; for The time system outputs the trajectory, including from Time's up The time length is The system output; for The time-lapse system input-output trajectory includes the time-lapse system input-output trajectory from Time's up The length of time is The system inputs and outputs.
[0095] Construct the Hankel matrices of the system input trajectory and the system output trajectory:
[0096]
[0097] In the formula, Let be the Hankel matrix of the system input trajectory and the system output trajectory, with dimension . , The deadline for offline data collection. It represents the set of real numbers.
[0098] (2) Perform singular value decomposition on the Hankel matrices of the system input trajectory and the system output trajectory to determine the system order.
[0099]
[0100] In the formula, Let be the system order. The value is equal to The number of larger singular values in the set, and satisfying , for The number of smaller singular values. for The front of the left singular matrix A matrix composed of columns for The left singular matrix of the back A matrix composed of columns for Larger singular values A diagonal matrix composed of singular values. for Smaller of the singular values The dimension consisting of singular values as the main diagonal elements is The matrix, for The front of the right singular matrix A matrix composed of rows for The right singular matrix after A matrix composed of rows. This represents the transpose of a matrix.
[0101] (3) Calculate the permutation matrix and construct the non-minimum controllable state and augmented state.
[0102] Calculate the permutation matrix:
[0103]
[0104]
[0105] In the formula, Let be the permutation matrix. For order is unit array, To select a matrix, The dimension is The zero matrix, The dimension is The zero matrix. The dimension is The zero matrix, For order is It is a unit array.
[0106] Constructing a non-minimum controllable state:
[0107]
[0108] In the formula, for The time is not the minimum controllable state. for Time system input-output trajectory.
[0109] Constructing augmenting states:
[0110]
[0111] In the formula, for Constantly expanding state, for The time is not the minimum controllable state. for Time system input.
[0112] (4) Construct the matrix required for online phase control
[0113] Construct the system input incremental data matrix, the system output data matrix, the current augmented state data matrix, and the subsequent augmented state data matrix:
[0114]
[0115] In the formula, for The system inputs an incremental data matrix at each moment. for The system outputs a data matrix at any given time. for The current augmented state data matrix at any given time. for The augmented state data matrix following time step 1, wherein:
[0116]
[0117] Construct the system input incremental-augmented state data matrix:
[0118]
[0119] In the formula, for The time dimension is The system input incremental-augmented state data matrix.
[0120] Construct the sample covariance matrix:
[0121]
[0122] In the formula, for The time dimension is The sample covariance matrix.
[0123] Construct the covariance-normalized input increment matrix, covariance-normalized output matrix, covariance-normalized current augmented state matrix, and covariance-normalized successor augmented state matrix:
[0124]
[0125] In the formula, , , and They are respectively The time-varying input increment matrix is covariance-normalized, the output matrix is covariance-normalized, the current augmented state matrix is covariance-normalized, and the subsequent augmented state matrix is covariance-normalized.
[0126] Online phase: Time scale is from The moment begins
[0127] The controller output is solved using the gradient descent method. To minimize the cost function .
[0128] The input incremental controller processes the active and reactive power output of the converter during the online operation phase, specifically following these steps:
[0129] S3.1. The active and reactive power output by the converter are used as the system output, and the d-axis and q-axis current reference values of the inner current loop are used as the system input.
[0130] S3.2 Construct the augmented state for the current moment based on the system order, the system input at the previous moment or the preset zero-moment system input, and the active and reactive power output by the converter.
[0131] S3.3 Construct an expression for the system input increment, and then construct a cost function based on the expression for the system input increment and the augmented state at the current time. With minimizing the cost function as the optimization objective, the gradient descent method is used to update the feedback gain and feedforward gain of the controller in the system input increment expression, and the system input increment at the current time is obtained by solving the problem.
[0132] S3.4. Based on the system input at the previous moment or the preset zero-time system input, add the system input increment at the current moment to obtain the system input at the current moment, that is, the d-axis and q-axis current reference values of the current inner loop.
[0133] The expressions for augmented state and system input increment are set according to the following formulas:
[0134]
[0135]
[0136]
[0137]
[0138]
[0139]
[0140] in, This represents the augmented state at time t. express The time is not the minimum controllable state. Let represent the system input at time t-1, m represent the dimension of the system input, r represent the system order, and T represent the permutation matrix. express The time-based system input-output trajectory, that is, including the time from Time's up The length of time is The system inputs and outputs, Indicates the order is unit array, The dimension is The zero matrix, The dimension is The zero matrix, Represents the selection matrix. The dimension is The zero matrix, For order is As a unit array, Indicates from Time's up The time length is System input, Indicates from Time's up The time length is The system output, This represents the system input increment at time t. and These represent the feedback gain and feedforward gain of the controller, respectively. This represents a reference value indicating the preset augmentation state. This indicates the detection noise signal, which uses white noise as the detection noise signal.
[0141] The cost function for the input incremental controller is set according to the following formula:
[0142]
[0143] in, Let represent the cost function, E represent the expected value, N represent the total time of the online phase, R represent the penalty matrix for the system input increment, Q represent the penalty matrix for the augmented state, and supremum represent the supremum. This represents the augmented state at time t. This represents a reference value indicating the preset augmentation state. This represents the system input increment at time t. This represents the transpose of a matrix.
[0144] In step S3.3, the gradient descent method is used to update the controller's feedback gain and feedforward gain in the system input increment expression, specifically according to the following steps:
[0145] Step 1: Set initial values for the feedback gain and feedforward gain at the initial moment;
[0146] Step 2: In each control cycle, calculate the system input increment at the current moment based on the feedback gain and feedforward gain at the current moment. After running the system, collect the system output at the current moment, and then update the sample covariance matrix based on the system input increment and system output at the current moment.
[0147] Step 3: Based on the updated sample covariance matrix, the current feedback gain and feedforward gain, calculate the feedback parameterization policy matrix and feedforward parameterization policy matrix for the next time step.
[0148] Step 4: Update the feedback parameterized policy matrix and the feedforward parameterized policy matrix for the next time step using the gradient descent method;
[0149] Step 5: Calculate the feedback gain and feedforward gain for the next time step based on the updated feedback parameterized policy matrix and feedforward parameterized policy matrix.
[0150] The offline preparation phase also includes: constructing a covariance-normalized input increment matrix, a covariance-normalized output matrix, a covariance-normalized current augmented state matrix, a covariance-normalized successor augmented state matrix, and a sample covariance matrix based on the collected system inputs and outputs, as initial values for gradient descent solution in the online running phase.
[0151] Specifically, in The initial feedback gain of the controller is given at any given time. With initial feedforward gain ,from At the moment, the following steps (1)-(4) are executed:
[0152] (1) Input increment through system Computer system input ,Will Input into the system, calculate The non-minimum controllable state, augmented state, and their reference values are updated using the rank-one update method. , , , and .
[0153] Input increment through system Computer system input ,Will Input into the system, collect system input trajectory System output trajectory And construct the system input-output trajectory ,calculate Non-minimum controllable state, augmented state, and their reference values at any given time:
[0154]
[0155] In the formula, for The time is not the minimum controllable state. Let be the permutation matrix. for The system input-output trajectory at time t, for Constantly expanding state, for Time system input, for Constant-time augmented state reference value, This is the reference value for the system output trajectory. Specifically, the reference value for the system output trajectory is...
[0156]
[0157] In the formula, For system output in ( Reference value at that time.
[0158] Build Time increment-augmented state vector
[0159]
[0160] Based on the input increment-augmented state vector Update the matrix using the rank-one update method. , , , and :
[0161]
[0162] (2) According to , and Solve for the feedback parameterized policy matrix and the feedforward parameterized policy matrix.
[0163]
[0164] In the formula, and They represent The feedback parameterized policy matrix and the feedforward parameterized policy matrix at each time step. for The inverse matrix of the sample covariance matrix at time step. and They are Feedback gain and feedforward gain of the timing controller.
[0165] (3) Solve the discrete Lyapunov equation and perform gradient descent to update the feedback parameterized policy matrix and the feedforward parameterized policy matrix.
[0166] Solve the discrete Lyapunov equations: DLyapE1 and DLyapE2
[0167]
[0168] In the formula, and These are the solutions to the Discrete Lyapunov Equations DLyapE1 and DLyapE2, respectively.
[0169] Constructing a matrix , and :
[0170]
[0171] Calculate the gradient of the cost function:
[0172]
[0173] In the formula, and The cost function is related to and gradient, For the system to be in a stable state, The dimension is The unit array.
[0174] Computation of projection operator
[0175]
[0176] In the formula, For projection operators, For order is The unit array.
[0177] The projected gradient descent method is used to update the feedback parameterized policy matrix and the feedforward parameterized policy matrix:
[0178]
[0179] In the formula, and For the updated feedback parameterized policy matrix and feedforward parameterized policy matrix, and Feedback parameterized policy matrix and feedforward parameterized policy matrix Gradient descent updates the step size. and They are respectively and The adaptive factor is updated by gradient descent. Describe the 2-norm of a matrix. This is the projection operator.
[0180] (4) Calculated output
[0181] Update the feedforward gain and feedback gain of the controller
[0182]
[0183] Calculate the output of the controller
[0184]
[0185]
[0186] In the formula, for The system inputs an increment at any time. and They are Feedback gain and feedforward gain of the timing controller. for The output of the timing controller also serves as a reference value for the inner current loop. and .
[0187] Current moment Update to the next sampling time Then, re-execute steps (1)-(4) of the online phase to implement the controller.
[0188] S4. Input the d-axis and q-axis current reference values of the inner current loop into the inner current loop processing to generate the modulated wave voltage amplitude signal;
[0189] Specifically, current inner-loop control is used to control the output current of the converter. The current inner-loop control provides a modulation wave voltage amplitude signal to complete the current inner-loop control. The specific steps include:
[0190] In the current inner loop control, a PI controller is used to control the output current of the converter. Axis components: and The modulation wave voltage amplitude signal is provided by the inner current loop control.
[0191]
[0192] In the formula, This refers to the controller frequency. and These are the modulated wave voltage amplitude signals. shaft and Axial components, and This is the reference value for the inner current loop. and For the output current of the converter shaft and Axial components, and Converter output voltage shaft and Axial components. This is the filter inductor for the converter. and These are the proportional and integral coefficients of the current inner-loop PI controller. and These are the proportional coefficient and time constant of the voltage feedforward circuit. For the Laplace operator.
[0193] Configure reference values for the converter's output active and reactive power, reference value for the q-axis component of the converter's output voltage, grid rated frequency, input incremental controller parameters, PI controller parameters for current inner loop control, and phase-locked loop PI controller parameters to complete the parameter settings for the optimal input incremental control method of the converter. Specific steps include:
[0194] Configure reference values for the active and reactive power output of the converter. and The output voltage of the converter Reference values for axis components Rated frequency value of power grid Input incremental controller parameters: upper bound of system lag Deadline for offline data collection The interval between adjacent moments, i.e., the sampling time. Penalty matrix of augmented state The penalty matrix for system input increment Gradient descent update step size for feedback parameterized policy matrix and feedforward parameterized policy matrix and Parameters of the PI controller for current inner loop control: and The proportional gain and time constant of the voltage feedforward circuit: and Phase-locked loop (PLL) PI controller parameters: and The parameters of the PI controller must be selected to meet the requirements of system stability and speed.
[0195] Among them, the penalty matrix of the augmented state :
[0196]
[0197] In the formula, The penalty matrix for the augmented state. Let be the permutation matrix. The penalty matrix for the system input-output trajectory is set by... It can be calculated .
[0198] The above parameters are configured to complete the parameter settings for the optimal input incremental control method of the converter.
[0199] S5. Generates drive signals based on the modulation wave voltage amplitude and phase signals, and controls the power switching devices of the converter. Specific control design details can be found in [link to relevant documentation]. Figure 2 .
[0200] Figure 2 This is a control block diagram of the optimal input increment control method for the converter of the present invention, which includes a phase-locked loop, Park variation, input increment control, current inner loop, and SPWM modulation. Figure 2 middle For the converter-side filter inductor, This is the equivalent resistance of the filter inductor on the converter side. For filtering capacitors, For damping resistor, and For transmission line inductance and resistance. This is the output voltage of the converter. For the converter output current, This is the grid voltage. , and It is a three-phase modulated wave voltage signal. This is the DC power supply voltage. and The output active power and reactive power of the converter. This represents the shift operator, and PCC is the converter grid connection point.
[0201] A modulated voltage signal is constructed by combining the obtained modulated voltage phase signal and modulated voltage amplitude signal, and a pulse width modulation signal is generated by an SPWM generator and applied to the power switching devices of the converter to achieve optimal input incremental control of the converter. Specific steps include:
[0202] (1) Generation of modulated wave voltage signal:
[0203] For modulated wave voltage phase signal and modulated wave voltage amplitude signal , Performing the inverse Park transform yields the modulated voltage signal:
[0204]
[0205] In the formula, To control the frequency signal, , and These are the A-phase, B-phase, and C-phase components of the modulated wave voltage signal. and These are respectively the modulated wave voltage signals shaft and Axial components.
[0206] (2) Generation of pulse width modulation signal:
[0207] The obtained modulated wave voltage signal , and The pulse width modulation signal is obtained through an SPWM generator.
[0208] By applying pulse width modulation signals to the power switching devices of the converter, optimal control of the converter's input increment can be achieved.
[0209] Example 1:
[0210] Reference Figure 3 This is the first embodiment of the present invention, which is a converter grid-connected simulation system. It employs both Grid Following (GFL) control and the Input Incremental Optimal Control (InIOC) method of the present invention as the converter control strategy. The relevant simulation parameter settings are shown in Table 1.
[0211] Table 1. Relevant parameters of the converter grid-connected system in the simulation verification of the embodiment.
[0212] parameter numerical values <![CDATA[Voltage reference U B / V]]> 380 <![CDATA[Power reference S B / kVA]]> 1.5 <![CDATA[DC power supply voltage U dc / V]]> 320 Rated frequency of power grid f / Hz 50 <![CDATA[Filter inductor L1 / mH]]> 15.3 <![CDATA[Equivalent resistance of filtering inductor R1 / mΩ]]> 9.6 <![CDATA[Filter capacitor C f / uF]]> 2 <![CDATA[Damping resistor R d / mΩ]]> 9.6
[0213] Reference values for converter output active and reactive power and The output voltage of the converter Reference values for axis components Rated frequency value of power grid Input incremental controller parameters: upper bound of system lag Deadline for offline data collection Sampling time The penalty matrix of the system input-output trajectory The penalty matrix for system input increment ,in, Represents a diagonal matrix. Indicates the order is The identity matrix, Represents the Kronecker product. The gradient descent update step size is performed on the feedback parameterized policy matrix and the feedforward parameterized policy matrix. and Parameters of the PI controller for current inner loop control: and The proportional gain and time constant of the voltage feedforward circuit: and Phase-locked loop (PLL) PI controller parameters: and .
[0214] In this embodiment, the short-circuit ratio (SCR) of the converter grid-connected system is: The converter's active power output jumps from 0 p.u. to 1.0 p.u. in 0.1 seconds, and the grid short-circuit ratio changes from [previous value] to [previous value] in 0.5 seconds. Decrease to Run the simulation and record the active and reactive power output of the converter, such as... Figure 4 As shown.
[0215] Through observation Figure 4 The results show that the InIOC control method significantly outperforms the traditional GFL when the converter experiences active power step disturbances and grid strength (SCR) decreases. Its active and reactive power response curves are smoother, and it can quickly and without overshoot track the command value after a 0.1s power surge, exhibiting rapid dynamic response. A key advantage is seen after a 0.5s decrease in the system short-circuit ratio: InIOC effectively suppresses power oscillations caused by grid weakening, and the curves quickly return to stability; while GFL exhibits continuous and large-amplitude active and reactive power oscillations, resulting in a significant decrease in system stability. This indicates that InIOC enhances the converter's robustness to grid impedance changes and significantly improves the converter's grid-connected operation stability under grid disturbances.
[0216] Example 2:
[0217] To further verify that the input incremental optimal control method of the converter in this invention can adapt to grid changes and suppress oscillations, the grid voltage was set to drop by 0.1 pu at 0.2 s. The simulation model was run and the reactive power output and output voltage of the converter were recorded as follows: Figure 5 As shown.
[0218] Through observation Figure 5 The results show that when the grid voltage drops by 0.2 seconds, the InIOC control strategy exhibits a significant advantage over the GFL. The reactive power output of the InIOC remains highly stable throughout the process, almost unaffected by voltage disturbances; while the reactive power of the GFL shows a significant decrease, indicating that its output passively changes with the grid voltage and lacks sufficient support capability. This highlights the active voltage support and strong robustness of the InIOC: it can quickly detect voltage drops and dynamically adjust reactive power to help restore the grid voltage. Therefore, the InIOC can better maintain grid connection point voltage stability during voltage disturbances, improving the converter's uninterrupted operation capability and voltage recovery speed, making it more suitable for grid-connected environments with high voltage stability requirements.
[0219] The input incremental optimal control method for converters does not require a precise system mathematical model. It adaptively adjusts control strategy parameters and structure by sensing the dynamic operating states of the grid, such as voltage amplitude, frequency, and impedance, in real time, effectively overcoming the inherent limitations of traditional fixed-parameter control architectures. This method can dynamically correct model mismatch deviations caused by changes in grid operating conditions, suppress oscillations such as voltage and current oscillations and subsynchronous oscillations, and achieve real-time dynamic adaptation of the converter control strategy to the grid operating state. Its application can significantly improve the safe and stable operation of renewable energy grid-connected systems and provide technical support for the large-scale consumption of renewable energy, possessing significant engineering practical value.
[0220] The incremental input optimal control method for converters aims to address the core challenges faced by traditional grid-following control under grid faults or disturbances. Firstly, this invention optimizes control commands online in real-time through algorithms, actively suppressing power oscillations caused by shocks such as decreased grid short-circuit ratios and frequency fluctuations, thereby enhancing the converter's synchronization stability and damping capability and preventing system instability. Secondly, this method does not rely on precise physical mathematical models for controller design. It dynamically adjusts the strategy by continuously tracking actual system operating data, thus effectively overcoming the control performance degradation caused by changes in system parameters, inaccurate models, or unmodeled dynamics, significantly improving control robustness and adaptability under uncertain operating conditions.
[0221] The above detailed embodiments illustrate the technical solution and beneficial effects of the present invention. 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 method for optimal input increment control of a converter, characterized in that, The method includes the following steps: S1. Obtain the output current and output voltage of the converter in the dq coordinate system, and then calculate the active power and reactive power output by the converter. S2. Obtain the control frequency signal through a phase-locked loop and perform an integral operation on the control frequency signal to obtain the phase signal; S3. The active and reactive power output from the converter are used as inputs to the incremental controller. After processing by the input incremental controller, the d-axis and q-axis current reference values of the inner current loop are obtained. S4. Input the d-axis and q-axis current reference values of the inner current loop into the inner current loop processing to generate the modulated wave voltage amplitude signal; S5. Generate a drive signal based on the amplitude and phase signals of the modulated wave voltage and control the converter.
2. The optimal input increment control method for a converter according to claim 1, characterized in that: The input incremental controller includes an offline preparation phase and an online operation phase. The offline preparation phase is executed only once before the converter is put into grid-connected operation, and the online operation phase is executed once in each control cycle after the converter is put into grid-connected operation. The offline preparation phase is processed according to the following steps: 1) Use the active and reactive power output from the converter as the system output, and the d-axis and q-axis current reference values of the inner current loop as the system input; 2) Apply white noise excitation to the converter, collect the system input and system output at different times, and construct the system output sequence and system input sequence in time order. Then, construct the Hankel matrix based on the system output sequence and system input sequence. 3) Perform singular value decomposition on the Hankel matrix to obtain multiple singular values, and determine the system order based on the multiple singular values.
3. The optimal input increment control method for a converter according to claim 2, characterized in that: The input incremental controller processes the active and reactive power output of the converter during the online operation phase, specifically according to the following steps: S3.
1. The active and reactive power output by the converter are used as the system output, and the d-axis and q-axis current reference values of the inner current loop are used as the system input. S3.2 Construct the augmented state for the current moment based on the system order, the system input at the previous moment or the preset zero-moment system input, and the active and reactive power output by the converter. S3.3 Construct an expression for the system input increment, and then construct a cost function based on the expression for the system input increment and the augmented state at the current time. With minimizing the cost function as the optimization objective, the gradient descent method is used to update the feedback gain and feedforward gain of the controller in the system input increment expression, and the system input increment at the current time is obtained by solving the problem. S3.
4. Based on the system input at the previous moment or the preset zero-time system input, add the system input increment at the current moment to obtain the system input at the current moment.
4. The optimal input increment control method for a converter according to claim 1, characterized in that: The expressions for the augmented state and the system input increment are specifically set according to the following formulas: in, This represents the augmented state at time t. express The time is not the minimum controllable state. Let represent the system input at time t-1, m represent the dimension of the system input, r represent the system order, and T represent the permutation matrix. express Time system input-output trajectory Indicates the order is unit array, The dimension is The zero matrix, The dimension is The zero matrix, Represents the selection matrix. The dimension is The zero matrix, The dimension is The zero matrix, For order is As a unit array, Indicates from Time's up The time length is System input, Indicates from Time's up The time length is The system output, This represents the system input increment at time t. and These represent the feedback gain and feedforward gain of the controller, respectively. Reference values representing the augmented state. Indicates the detection of noise signals. This represents the preset reference value of the system output at time t.
5. The optimal input increment control method for a converter according to claim 1, characterized in that: The cost function of the input incremental controller is specifically set according to the following formula: in, Let E represent the cost function, N represent the expected value, R represent the penalty matrix for the system input increment, Q represent the penalty matrix for the augmented state, and supremum represent the supremum. This represents the augmented state at time t. Reference values representing the augmented state. This represents the system input increment at time t. This represents the transpose of a matrix.
6. The optimal input increment control method for a converter according to claim 1, characterized in that: In step S3.3, the gradient descent method is used to update the feedback gain and feedforward gain of the controller in the system input increment expression, specifically according to the following steps: Step 1: Set initial values for the feedback gain and feedforward gain at the initial moment; Step 2: In each control cycle, calculate the system input increment at the current moment based on the feedback gain and feedforward gain at the current moment. After running the system, collect the system output at the current moment, and then update the sample covariance matrix based on the system input increment and system output at the current moment. Step 3: Based on the updated sample covariance matrix, the current feedback gain and feedforward gain, calculate the feedback parameterization policy matrix and feedforward parameterization policy matrix for the next time step. Step 4: Update the feedback parameterized policy matrix and the feedforward parameterized policy matrix for the next time step using the gradient descent method; Step 5: Calculate the feedback gain and feedforward gain for the next time step based on the updated feedback parameterized policy matrix and feedforward parameterized policy matrix.
7. The optimal input increment control method for a converter according to claim 1, characterized in that: The offline preparation phase also includes: constructing a covariance-normalized input increment matrix, a covariance-normalized output matrix, a covariance-normalized current augmented state matrix, a covariance-normalized successor augmented state matrix, and a sample covariance matrix based on the collected system inputs and outputs, as initial values for gradient descent solution in the online running phase.
8. A computer-readable storage medium storing program data thereon, characterized in that, When the program data is executed by the processor, the method as described in any one of claims 1-7 is implemented.
9. A computer device, characterized in that, It includes a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the steps of the input incremental optimal control method for the converter as described in any one of claims 1 to 7.