CN-122026351-A - Parameter identification method for low-voltage current-limiting link in high-voltage direct-current transmission system

CN122026351ACN 122026351 ACN122026351 ACN 122026351ACN-122026351-A

Abstract

The invention provides a parameter identification method for a low-voltage current-limiting link in a high-voltage direct-current transmission system, which is characterized in that an error feedback mechanism and a heuristic algorithm are utilized to realize high-precision parameter optimization of five key parameters (a starting voltage threshold, an exiting voltage threshold, a maximum current limit, a minimum current limit and acquisition time) of the low-voltage current-limiting link under a single scene, then a multi-scene identification result is fused through a deep learning model, and finally a low-voltage current-limiting link parameter identification model which is robust, universal and capable of accurately reflecting dynamic characteristics of an actual engineering under various disturbances is constructed.

Inventors

ZHANG XUEYOU
LIAO JUN
YUAN HONGDE
WANG YUN
MA HUAN
WANG YUZHE
ZHANG XIAOYU
SUN CHAOPENG
ZHAO HANG
YU JIAN
ZHU HAOYAN
JIAO ZHEN
WANG CHUNYANG
GUO XINGWANG
FAN PEIPEI
YUAN JING
WANG BIN
HE CHENGCHENG
RUAN WEI
Zhang Honggu
ZHU TAO

Assignees

国网安徽省电力有限公司超高压分公司
清华大学

Dates

Publication Date: 20260512
Application Date: 20251205

Claims (10)

1. The parameter identification method for the low-voltage current limiting link in the high-voltage direct-current transmission system is characterized by comprising the following steps of: S1, collecting time sequence data of an inversion side of a high-voltage direct-current transmission system under multi-scene disturbance, wherein the time sequence data comprises alternating voltage, alternating current and a trigger angle, the multi-scene disturbance comprises alternating faults of different types and disturbance of different resistance values under the same fault type, and cleaning the collected data to ensure data integrity and effectiveness; s2, constructing a simulation model comprising fixed arc extinguishing angle control, fixed current control and a low-voltage current limiting link to be identified based on a standard Cigre Benchmark model, wherein parameters of the low-voltage current limiting link comprise a starting voltage threshold, an exiting voltage threshold, a maximum current limit, a minimum current limit and acquisition time; s3, performing multi-objective optimization on the low-voltage current limiting parameters in the simulation model by adopting a particle swarm algorithm, wherein the multi-objective optimization aims at minimizing the root mean square error of the inversion side voltage and the inversion side current output by the simulation model and the measured data, and dynamically adjusting the searching direction of the particle swarm based on the current error value in each iteration until the convergence condition is met; s4, combining the optimal parameters obtained by multi-objective optimization and the time sequence data of the corresponding scene into training samples, inputting the training samples into a convolutional neural network to perform multi-scene parameter mapping relation training, and generating a deep learning model capable of directly outputting low-voltage current limiting link parameters according to new scene time sequence data.
2. The method of claim 1, wherein S1 comprises: s11, adopting a sliding window detection algorithm to identify continuous missing value segments in time sequence data, and filling by a linear interpolation method; S12, identifying abnormal data based on a statistical model, wherein the statistical model eliminates data segments with deviation exceeding a preset threshold by calculating the deviation degree of sliding window variances and historical means of alternating voltage, alternating current and trigger angles.
3. The method of claim 1, wherein S2 comprises: S21, threshold value of starting voltage of low-voltage current-limiting link Is set to 0.9, the exit voltage threshold The initial value of (2) is set to 0.4, maximum current limit Is set to 0.55, the minimum current limit The initial value of (1) is set to 1.0, and the acquisition time is Is set to 0.02; S22, through a preset piecewise function expression Modeling the dynamic characteristics of the low-voltage current limiting link.
4. The method of claim 1, wherein S3 further comprises: S31, defining the objective function of the multi-objective optimization as min(RMSE(U_simulated, U_measured),RMSE (I _simulated, I_measured)); S32, dynamically adjusting inertia weight parameters in a particle speed update formula by calculating the gradient directions of the individual optimal position and the global optimal position of each particle in a particle swarm algorithm.
5. The method of claim 1, wherein S4 comprises: S41, carrying out normalization processing on the input time sequence data, and unifying the numerical ranges of alternating voltage, alternating current and trigger angle to the [0,1] interval; S42, a multi-layer convolutional neural network architecture is adopted, wherein the multi-layer convolutional neural network architecture comprises at least two convolutional layers and a full-connection layer, and the mapping relation between input data and parameter vectors is learned through a nonlinear activation function.
6. The utility model provides a parameter identification device of low voltage current limiting link in HVDC system which characterized in that includes: The system comprises a time sequence data acquisition and cleaning module, a data acquisition module and a data processing module, wherein the time sequence data acquisition and cleaning module is used for acquiring time sequence data of an inversion side of a high-voltage direct-current transmission system under multi-scene disturbance, the time sequence data comprises alternating voltage, alternating current and a trigger angle, the multi-scene disturbance comprises alternating current faults of different types and disturbance scenes with different resistance values under the same fault type, and cleaning is carried out on the time sequence data to remove invalid data and missing values; the simulation model construction module is used for constructing a simulation model comprising fixed arc extinguishing angle control, fixed current control and a low-voltage current limiting link to be identified based on a standard Cigre Benchmark model, wherein parameters of the low-voltage current limiting link comprise a starting voltage threshold, an exiting voltage threshold, a maximum current limit, a minimum current limit and acquisition time; the multi-objective optimization module is used for carrying out multi-objective optimization on the low-voltage current limiting parameters in the simulation model by adopting a particle swarm algorithm, the multi-objective optimization aims at minimizing the root mean square error of the inversion side voltage and the inversion side current which are output by the simulation model and the measured data, and dynamically adjusts the searching direction of the particle swarm based on the current error value in each iteration until the convergence condition is met; And the deep learning model training module is used for combining the optimal parameters obtained by multi-objective optimization and the time sequence data of the corresponding scene into training samples, inputting the training samples into the convolutional neural network to perform multi-scene parameter mapping relation training, and generating a deep learning model capable of directly outputting low-voltage current limiting link parameters according to the new scene time sequence data.
7. The apparatus of claim 6, wherein the timing data acquisition and cleansing module is further to: adopting a sliding window detection algorithm to identify continuous missing value segments in the time sequence data, and filling by a linear interpolation method; And identifying the abnormal data based on a statistical model, wherein the statistical model eliminates the data segment with the deviation exceeding a preset threshold value by calculating the deviation degree of the sliding window variance and the historical mean value of the alternating voltage, the alternating current and the trigger angle.
8. The apparatus of claim 6, wherein the simulation model building module is further to: Threshold of starting voltage for low-voltage current limiting link Is set to 0.9, the exit voltage threshold The initial value of (2) is set to 0.4, maximum current limit Is set to 0.55, the minimum current limit The initial value of (1) is set to 1.0, and the acquisition time is Is set to 0.02; by a preset piecewise function expression Modeling the dynamic characteristics of the low-voltage current limiting link.
9. A computer device comprising a processor and a memory; The processor executes a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to implement the parameter identification method of the low-voltage current limiting link in the hvdc transmission system according to any one of claims 1 to 5.
10. A non-transitory computer readable storage medium having stored thereon a computer program, which when executed by a processor implements a method for parameter identification of a low voltage current limiting link in a hvdc transmission system according to any one of claims 1-5.

Description

Parameter identification method for low-voltage current-limiting link in high-voltage direct-current transmission system Technical Field The invention belongs to the technical field of modeling and control of a high-voltage direct-current transmission system, and particularly relates to a parameter identification method for a low-voltage current limiting link in the high-voltage direct-current transmission system. Background The high-voltage direct current transmission (LCC-HVDC) technology of the grid converter has the advantages of large capacity, low loss, long distance and the like. In order to perform stability analysis, fault diagnosis, control strategy optimization and new system planning and design on a put-into-operation HVDC system, it is an indispensable precondition to build a simulation model capable of accurately reflecting the dynamic characteristics of the actual system. An accurate LCC-HVDC simulation model not only includes one-time systems (e.g. converter transformers, converter valves, dc lines, etc.), but more importantly accurate simulation of the control protection system. In actual engineering, the core logic and parameters that control the protection system are often considered manufacturer's secrets, and the internal details are often "black box" or "gray box" states for the system operator. Therefore, in the modeling process, researchers often adopt standard control links (such as fixed arc extinguishing angle control, fixed current control, low-voltage current limiting links and the like) to equivalently simulate a control module in actual engineering. The low-voltage current limiting link is a key protection functional module for preventing commutation failure of the LCC-HVDC system during fault and recovery and maintaining stable operation of the system, and the accuracy of parameter setting directly determines the dynamic response characteristic of the system under disturbance and is critical to the reliability of simulation results. At present, the parameter identification, especially the identification of VDCOL parameters, of the LCC-HVDC system control link mainly faces the following technical difficulties: 1. The model equivalent error is that the accurate code and logic of the actual engineering controller cannot be known, and structural and parametric errors exist by adopting a standard control link to perform the equivalent. How to make the output response of the equivalent model approach the response of the actual engineering to the maximum is a core challenge. 2. The parameters of VDCOL (such as a start voltage threshold, an exit voltage threshold, a maximum current limit and a minimum current limit, etc.) are tightly coupled with the parameters of other control links (such as a current regulator) in the system, and the parameters are nonlinear links. In the fault transient process, the strong nonlinearity and the coupling make the traditional parameter setting method based on a linear system or a single working point difficult to work, the identification result tends to be in local optimum, and the real dynamic state of the system cannot be globally reflected. 3. Depending on the exact internal information, some existing optimization recognition methods (e.g. based on sensitivity analysis, intelligent optimization algorithms, etc.) are applied to parameter recognition, but they often rely heavily on an accurate knowledge of the internal structure of the controller or need to be able to directly measure the internal variables of the control system as feedback, which is difficult to achieve in the practical engineering context of "black boxes" or "gray boxes". 4. The existing method generally optimizes by taking the error at a certain moment as an objective function in the identification process, and lacks an effective feedback mechanism capable of automatically and real-timely adjusting the parameter searching direction according to the dynamic process error. This results in a slow convergence rate of the identification process, and is sensitive to initial values and noise, and the finally identified parameters do not agree well in the actual system dynamics process (e.g., voltage dip and recovery). Therefore, there is an urgent need in the art for a parameter identification method that does not depend on an accurate model in a controller, can effectively process nonlinearity and parameter coupling of a system, and has adaptive feedback adjustment capability, so as to realize accurate and efficient identification of key control parameters such as a low-voltage current limiting link of an LCC-HVDC system, thereby solving the problem of equivalent modeling of an actual engineering control module and improving accuracy and reliability of a simulation model of the whole HVDC system. Disclosure of Invention The present invention aims to solve at least one of the technical problems in the related art to some extent. Therefore, a first objective of the present invention is t