CN-122026736-A - Converter, elasticity prediction control method thereof and computer storage medium

CN122026736ACN 122026736 ACN122026736 ACN 122026736ACN-122026736-A

Abstract

The application discloses an elastic prediction control method of a converter, which comprises the steps of constructing an output current increment prediction model, constructing a criterion function, calculating the gradient of the criterion function on a coefficient matrix, enabling the gradient to be zero, obtaining the estimation of the coefficient matrix at the current moment, taking the error between the reference current of the next period and the actual output current of the current period as the target output of the output current increment prediction model, rewriting the output current increment prediction model, obtaining the control law of voltage increment, substituting the control law of voltage increment into the coefficient matrix estimation at the current moment, obtaining the voltage control increment, and selecting the switching state of the converter based on the voltage control increment. The method can realize the predictive control of the converter without parameter dependence, resist false data attack, reduce the switching frequency and improve the robustness of the system.

Inventors

MA JIEN
ZHAO PENGBO
FANG YOUTONG
LIU CHENGHAO
ZHANG ZEYU
QIU LIN
LIU XING

Assignees

浙江大学

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. An elasticity prediction control method of a current transformer, characterized in that the elasticity prediction control method comprises the following steps: Constructing an output current increment prediction model, wherein the output current increment prediction model comprises an inherent dynamic item reflecting the mapping relation between the historical data of the converter system and the output current increment, and an attack item reflecting external abnormality; Constructing a criterion function, wherein the criterion function comprises a tracking term representing an error between an actual output current increment and a predicted output current increment, and a penalty term representing the variation of a coefficient matrix of the inherent dynamic term, and the predicted output current increment is obtained through the output current increment prediction model; Calculating the gradient of the criterion function relative to the coefficient matrix, and enabling the gradient to be zero to obtain the estimation of the coefficient matrix at the current moment; Taking the error between the reference current of the next period and the actual output current of the current period as the target output of the output current increment prediction model, and rewriting the output current increment prediction model to obtain a control law of voltage increment; Substituting the control law of the voltage increment into the coefficient matrix estimation at the current moment to obtain the voltage control increment, and selecting the state of the converter switch based on the voltage control increment.
2. The method of elasticity predictive control of a current transformer according to claim 1, further comprising the steps of: Constructing a neural network model to establish a mapping relation between the attack item and a system state, wherein the system state is the output current of the converter; The state estimator is designed, wherein the state estimator superimposes the output of the neural network model and the output of the output current increment prediction model as prediction items, and introduces feedback correction items based on output current estimation errors, namely the difference between an output current sampling value and a system state estimation value output by the state observer, to realize system state tracking; Based on the state estimator, configuring an update law of a weight coefficient matrix of the neural network model according to a stability criterion, so that the output current estimation error converges to zero under the action of the update law, wherein the update law comprises gradient descent terms related to the output current estimation error; and updating the weight coefficient matrix by using the updating law according to the output current estimation error of the current period, and calculating an attack item estimation value of the next control period by using the updated weight coefficient matrix and the current system state.
3. The method of elasticity prediction control of a current transformer according to claim 2, further comprising: Based on a preset control sampling period, a first-order Euler method is adopted to carry out discretization processing on the state observer and the update law respectively, so that a discretization prediction model of an attack item is generated, and an attack item estimated value of the next control period is calculated by using the discretization prediction model according to the output current estimated error of the current period.
4. A method of controlling a current transformer according to claim 2 or 3, characterized in that the method further comprises: And compensating the voltage control increment based on the attack item estimated value.
5. The method according to claim 2, wherein a system state vector is formed by using a historical output current sequence of the converter, the neural network model adopts a radial basis function neural network model, and an activation function of the neural network model is a gaussian radial basis function taking the system state vector as a variable.
6. The method for predictive control of a motor vehicle according to claim 1, wherein, And adopting an event triggering type to control the updating of the voltage control increment, wherein the event triggering control mechanism comprises: defining event trigger errors The event triggering error For the current tracking error Tracking error with last trigger time The tracking error is the difference between the reference output current and the actual output current; Triggering an update of the voltage control increment when the following condition is satisfied: ; In the formula, Indicating that a preset safety threshold value is to be set, , The tracking error at time k is indicated, For the prediction residual error under ideal control, Representing Ly+1st term coefficient vector in the coefficient matrix estimated at k moment, The vector of gain coefficients is represented and, , Representing the normal number of scaling factors, Representing the regularization constant.
7. The method of claim 1, wherein the control law of voltage increment is expressed by the following formula: ; ; ; In the formula, The vector of gain coefficients is represented and, Representing the normal number of scaling factors, Representing the regularization constant, Representing the historical impact vector(s), Representing Ly+1st term coefficient vector in the coefficient matrix estimated at k moment, The attack item is represented by a code that, Representing the voltage increase.
8. A current transformer, characterized in that the current transformer employs the elasticity prediction control method according to any one of claims 1 to 7.
9. The current transformer according to claim 8, wherein the current transformer is any one of a two-level current transformer, and a modular multi-level current transformer.
10. A computer storage medium storing a processing program which, when executed, performs the elasticity prediction control method according to any one of claims 1 to 7.

Description

Converter, elasticity prediction control method thereof and computer storage medium Technical Field The present application relates to the field of power electronics technologies, and in particular, to a converter, an elasticity prediction control method thereof, and a computer storage medium. Background The grid-connected scale of new energy is continuously expanded, the stability and the electric energy quality of a power system are directly determined by the control performance of a converter serving as a core interface, wherein the finite set model predictive control (FCS-MPC) becomes a research hot spot by virtue of the superior dynamic performance of the FCS-MPC, however, the method is highly dependent on an accurate mathematical model, the problem of reduced control robustness caused by parameter mismatch is faced in practical application, meanwhile, switching loss is not considered in the optimization process of the FCS-MPC, the high-frequency action of a switching device is directly caused, the loss is greatly increased, in addition, the converter is extremely easy to suffer from false data injection attack under the digital interconnection environment, the traditional control strategy lacks an effective defense mechanism, and the system security faces serious threat. Disclosure of Invention The embodiment of the application provides a converter, an elasticity prediction control method thereof and a computer storage medium, the method can realize the prediction control of the converter without parameter dependence, reduce the switching frequency while resisting false data injection attack, and improve the robustness of the system. In a first aspect, the present application provides an elasticity prediction control method for a current transformer, where the elasticity prediction control method includes: Constructing an output current increment prediction model, wherein the output current increment prediction model comprises an inherent dynamic item reflecting the mapping relation between the historical data of the converter system and the output current increment, and an attack item reflecting external abnormality; constructing a criterion function, wherein the criterion function comprises a tracking term for representing errors between actual output current increment and predicted output current increment, and a penalty term for representing variation of a coefficient matrix of an inherent dynamic term, and the predicted output current increment is obtained through an output current increment prediction model; Calculating the gradient of the criterion function on the coefficient matrix, and enabling the gradient to be zero to obtain the estimation of the coefficient matrix at the current moment; taking the error between the reference current of the next period and the actual output current of the current period as the target output of the output current increment prediction model, and rewriting the output current increment prediction model to obtain the control law of voltage increment; Based on the control law of the voltage increment, substituting the control law of the voltage increment into the coefficient matrix estimation at the current moment to obtain the voltage control increment, and selecting the state of the converter switch based on the voltage control increment. In one embodiment, the elasticity prediction control method further includes the steps of: Constructing a neural network model to establish a mapping relation between an attack item and a system state, wherein the system state is the output current of the converter; The state estimator is designed, wherein the state estimator takes the output of the neural network model and the output of the output current increment prediction model as prediction items, and introduces a feedback correction item based on an output current estimation error to realize system state tracking, and the output current estimation error is the difference between an output current sampling value and a system state estimation value output by the state observer; based on the state estimator, configuring an update law of a weight coefficient matrix of the neural network model according to a stability criterion, so that an output current estimation error converges to zero under the action of the update law, wherein the update law comprises a gradient descent term related to the output current estimation error; and according to the output current estimation error of the current period, updating the weight coefficient matrix by using an update law, and calculating an attack item estimation value of the next control period by using the updated weight coefficient matrix and the current system state. In one embodiment, the control method further includes: Based on a preset control sampling period, a first-order Euler method is adopted to carry out discretization processing on a state observer and an update law respectively, so that a discretization prediction mo