Load current prediction method, device and equipment for short-circuit simulation and medium
By constructing an equivalent model and training it using a surrogate model, the applicable range is determined, which solves the problems of insufficient accuracy and high complexity in short-circuit current calculation and achieves efficient short-circuit current prediction.
Patent Information
- Application Number
- CN202511554411.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-29
- Publication Date
- 2026-01-27
AI Technical Summary
Existing short-circuit current calculation methods are difficult to accurately reflect the impact of distributed generation, resulting in insufficient calculation accuracy, high complexity, and long calculation time. Improving the efficiency of short-circuit current prediction while ensuring accuracy has become an urgent problem to be solved.
An equivalent model is constructed and trained using a surrogate model to determine the applicable range of the equivalent model. The surrogate model is then used to fit the equipment characteristic parameters to update the equivalent model. Finally, a load mechanism model is combined to predict short-circuit current outside the applicable range, thereby improving prediction efficiency.
By determining the update of the equivalent model through surrogate models and training methods, the efficiency of parameter optimization is improved, ensuring that the efficiency of short-circuit current prediction is improved while the accuracy of current prediction meets the requirements.
Smart Images
Figure CN121413541A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of power system simulation modeling technology, specifically relating to a load current prediction method, prediction device, computer equipment, and storage medium for short-circuit simulation. Background Technology
[0002] Currently, with the rapid development of renewable energy and the widespread application of distributed power sources in distribution networks, the power grid topology of power systems is becoming increasingly complex, and the types of loads in power systems are constantly increasing, including various industrial loads, commercial loads, residential loads, and electric vehicle charging loads, and the diversity of load characteristics is also increasing.
[0003] Existing short-circuit current calculation methods are usually based on steady-state models of networks, which make it difficult to accurately reflect the impact of distributed generation. Moreover, existing methods often ignore the fluctuations and dynamic changes in the output of distributed generation and the diversity and dynamic changes in load characteristics, resulting in insufficient accuracy in short-circuit current calculations, as well as high computational complexity and long computation time.
[0004] To address the aforementioned issues, existing technologies propose using load mechanism models for short-circuit current prediction to improve the accuracy of short-circuit current prediction. This allows for the analysis of the safety and stability of the power grid under various operating conditions based on accurate short-circuit current prediction. Load mechanism models are typically established based on a deep understanding of the physical principles, electrical characteristics, and operating mechanisms of load equipment, and can clearly describe the behavior of the load under different operating conditions.
[0005] However, due to its inherent complexity, the load mechanism model still suffers from high computational complexity and long computation time, resulting in poor efficiency in short-circuit current prediction. Therefore, improving the efficiency of short-circuit current prediction while ensuring its accuracy has become an urgent problem to be solved. Summary of the Invention
[0006] To address the shortcomings of existing technologies, this invention provides a load current prediction method, prediction device, computer equipment, and storage medium for short-circuit simulation.
[0007] To solve one or more or all of the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A load current prediction method for short-circuit simulation includes: constructing an equivalent model based on a load mechanism model; analyzing the equivalent model to obtain equipment characteristic parameters and dynamic operating parameters; constructing a surrogate model, wherein the surrogate model takes the equipment characteristic parameters and dynamic operating parameters as input and the output parameters of the equivalent model as output; training the surrogate model based on preset samples and preset labels; fitting the equipment characteristic parameters of the equivalent model using the surrogate model and updating the equivalent model; determining the applicable range of the updated equivalent model; when the short-circuit point is within the applicable range, using the updated equivalent model to predict the short-circuit current; otherwise, using the load mechanism model to predict the short-circuit current.
[0008] Furthermore, the method for training the surrogate model includes: randomly generating initial values for the input parameters based on physical constraints and preset boundaries; concatenating the initial values corresponding to the input parameters into a sample vector, using the sample vector as a preset sample; calculating the sample output value of the equivalent model based on the initial values corresponding to the input parameters and the equivalent model, using the sample output value as a preset label; inputting the preset sample into the surrogate model for prediction to obtain the sample current; calculating the first training loss based on the sample current, the preset label, and the mean squared error loss function; training the surrogate model based on the first training loss until the first training loss converges, thus obtaining the trained surrogate model.
[0009] Furthermore, the equipment characteristic parameters include the impedance value of the constant impedance load, the power value of the constant power load, the internal impedance of the induction motor, the current value of the constant current load, the first attenuation coefficient, and the second attenuation coefficient; the dynamic operating parameters include the system voltage, the peak value of the internal potential, the current time point, the short circuit time point, and the recovery time point.
[0010] Furthermore, the method for fitting the device characteristic parameters of the equivalent model using a surrogate model includes: obtaining reference information and reference current when a short circuit occurs at a reference location; assigning values to dynamic operating parameters based on the reference information to obtain reference values corresponding to each dynamic operating parameter; generating corresponding values for each device characteristic parameter, forming a reference input vector from the reference values corresponding to the reference information and the generated values corresponding to each device characteristic parameter; inputting the reference input vector into the trained surrogate model to obtain the reference predicted current; calculating the second training loss based on the reference predicted current, the reference current, and a preset loss function; updating the generated values of the device characteristic parameters in the reference input vector based on the second training loss until the second training loss converges, thus obtaining the reference values corresponding to each device characteristic parameter.
[0011] Furthermore, the method for determining the applicable range of the updated equivalent model includes: obtaining the current prediction errors corresponding to multiple verification locations; clustering the current prediction errors using a preset clustering method to obtain an error cluster set; determining the error cluster set with the smallest mean error as the first target set based on the mean error of each error cluster set; clustering the verification locations corresponding to each current prediction error in the first target set using a preset clustering method to obtain a location cluster set; determining the location cluster set with the largest number of verification locations as the second target set based on the number of verification locations contained in each location cluster set; and determining the applicable range of the equivalent model based on the maximum and minimum values of each verification location in the second target set.
[0012] Furthermore, it also includes: validating the updated equivalent model.
[0013] The method for verifying the updated equivalent model includes: for any one of the preset multiple verification locations, determining the verification input vector based on the updated equivalent model and the verification information obtained when a short circuit occurs at that verification location; and determining the current prediction error based on the verification input vector, the verification current obtained when a short circuit occurs at that verification location, and the updated equivalent model.
[0014] A load current prediction device for short-circuit simulation includes: a model analysis module for constructing an equivalent model based on a load mechanism model, and for analyzing the equivalent model to obtain equipment characteristic parameters and dynamic operating parameters; a surrogate model construction module for constructing a surrogate model based on the equipment characteristic parameters, dynamic operating parameters, and output parameters of the equivalent model; a surrogate model training module for training the surrogate model based on preset samples and preset labels; an equivalent model update module for fitting the equipment characteristic parameters of the equivalent model using the surrogate model and updating the equivalent model; an applicable range determination module for determining the applicable range of the updated equivalent model; and a current prediction module for predicting short-circuit current using the updated equivalent model when the short-circuit point is within the applicable range, and for predicting short-circuit current using the load mechanism model when the short-circuit point is outside the applicable range.
[0015] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the prediction method when executing the computer program.
[0016] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the prediction method.
[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes a load current prediction method for short-circuit simulation. A surrogate model is constructed based on the model parameters of the equivalent model. The update of the equivalent model is determined through the surrogate model and the training method. Compared to methods that directly optimize parameters based on the equivalent model, this method effectively improves the efficiency of parameter optimization. Based on the current prediction errors corresponding to multiple verification locations, the applicable range of the updated equivalent model is determined. By considering the applicable range and the target location of the short circuit, the choice between using the load mechanism model or the updated equivalent model is made. When the current prediction accuracy of the updated equivalent model meets the requirements, the updated equivalent model can be selected to improve the efficiency of short-circuit current prediction. Attached Figure Description
[0018] The present invention will now be described in further detail with reference to the accompanying drawings.
[0019] Figure 1 : Flowchart of the prediction method of this invention; Figure 2 A schematic diagram illustrating the characteristics of the internal electromotive force of an induction motor during a short circuit. Figure 3 : Schematic diagram of the prediction device of the present invention; Figure 4 : Schematic diagram of the computer device of the present invention. Detailed Implementation
[0020] To better understand the present invention, the content of the invention is further clearly illustrated below with reference to embodiments and accompanying drawings. However, the scope of protection of the present invention is not limited to the embodiments described below. Numerous specific details are set forth in the following description to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the present invention can be practiced without one or more of these details.
[0021] In one embodiment, such as Figure 1 As shown, a load current prediction method for short-circuit simulation is provided, including the following steps: S101. Construct an equivalent model based on the load mechanism model, and analyze the equivalent model to obtain equipment characteristic parameters and dynamic operating parameters.
[0022] The load mechanism model refers to a load model established based on a deep understanding of the physical principles, electrical characteristics, and operating mechanisms of load equipment. The load mechanism model is a conventional model and will not be elaborated upon here. In final application, the equivalent model has the same output parameters (short-circuit current) as the load mechanism model.
[0023] The load short-circuit current is composed of the superposition of the constant impedance load current, the constant current load current, the constant power load current, and the dynamic induction motor load current. Therefore, the short-circuit current and its equivalent model can all be expressed as: Where U(t) is the system voltage at time t, and Z is the impedance value of the constant impedance load. P represents the current value of the constant current load, and P represents the power value of the constant power load. Let E(t) be the internal impedance of the induction motor, and E(t) be the internal electromotive force of the induction motor at time t.
[0024] like Figure 2 The figure shows the characteristics of the internal electromotive force of an induction motor when it is short-circuited. This represents the peak value of the internal potential. The short circuit time point This is the point in time for recovery. When hour, ;when hour, ;when hour, .in and These are the first attenuation coefficient and the second attenuation coefficient, respectively.
[0025] Based on the above analysis, the equipment characteristic parameters are determined to include the impedance value Z of the constant impedance load, the power value P of the constant power load, and the internal impedance of the induction motor. Current value of constant current load First attenuation coefficient Second attenuation coefficient Dynamic operating parameters include system voltage U(t) and peak internal potential. Current time point t, short-circuit time point and recovery time point Among them, the equipment characteristic parameters are constants that describe the inherent physical properties of the load equipment, while the dynamic operating parameters are dynamic quantities.
[0026] Compared to the load mechanism model, the equivalent model has simpler computational logic, lower computational complexity, and higher computational efficiency, resulting in more efficient prediction of short-circuit current. When using the equivalent model, it is necessary to determine the specific values of the equipment characteristic parameters to ensure the accuracy of short-circuit current prediction.
[0027] S102. Construct a proxy model based on the equipment characteristic parameters, dynamic operating parameters, and output parameters of the equivalent model.
[0028] The surrogate model can use existing regression models, such as multilayer perceptrons and fully connected neural network models. The device characteristic parameters and dynamic operating parameters are concatenated as the input vector of the surrogate model, and the output is the output value (current value) of the output parameters of the equivalent model.
[0029] S103. Train the surrogate model based on preset samples and preset labels.
[0030] The preset samples can be generated randomly. The range of random generation and physical constraints of the preset samples are limited according to prior information. The surrogate model is trained based on the preset samples and their corresponding preset labels so that the surrogate model learns the mapping relationship between input parameters and output parameters.
[0031] Specifically, this step includes: Initial values are randomly generated for the input parameters (equipment characteristic parameters and dynamic operating parameters) based on physical constraints and preset boundaries. The initial values corresponding to the input parameters are concatenated into a sample vector, which is then used as a preset sample. Based on the initial values and equivalent models corresponding to the input parameters, the sample output values of the equivalent model are calculated, and the sample output values are used as preset labels. The preset sample is input into the surrogate model for prediction to obtain the sample current; The first training loss is calculated based on the sample current, preset label, and mean square error loss function. The surrogate model is then trained based on the first training loss until the first training loss converges, resulting in a well-trained surrogate model.
[0032] Since the training of the surrogate model is only to enable the surrogate model to learn the mapping relationship between the model input parameters and the model output parameters, any preset samples can be used for training. However, it is advisable to provide a large number of preset samples and their corresponding preset labels to ensure that the trained surrogate model can effectively represent the mapping relationship between the M model input parameters and the model output parameters.
[0033] S104. Use the surrogate model to fit the equipment characteristic parameters of the equivalent model and update the equivalent model.
[0034] Based on the reference information, reference current, and trained surrogate model obtained when a short circuit occurs at the reference location, the reference values corresponding to the device characteristic parameters are determined.
[0035] The reference location can be represented by the distance between the load connection location and the short circuit location, and the reference value refers to the optimization result of the corresponding model parameters.
[0036] The reference information corresponds to the dynamic operating parameters, namely: reference voltage, reference internal potential peak value, reference short-circuit time point, reference recovery time point, and reference current time point.
[0037] Specifically, methods for fitting the device characteristic parameters of the equivalent model using a surrogate model include: Obtain reference information and reference current when a short circuit occurs at the reference location; The dynamic operating parameters are assigned values based on the reference information to obtain the reference values corresponding to each dynamic operating parameter. Generate corresponding values for each device characteristic parameter, and form a reference input vector by combining the reference values corresponding to the reference information and the generated values corresponding to the device characteristic parameters. The reference input vector is input into the trained surrogate model to obtain the reference predicted current; Based on the reference predicted current, the reference current, and the preset loss function, the second training loss is calculated. Based on the second training loss, the generated values of the device characteristic parameters in the reference input vector are updated until the second training loss converges, and the reference values corresponding to the device characteristic parameters are obtained.
[0038] Among them, dynamic operating parameters are the model input parameters that need to be given, and equipment characteristic parameters are the model parameters that need to be optimized.
[0039] The reference information may include reference values for each dynamic operating parameter when a short circuit occurs at the reference location.
[0040] If a short circuit occurs at the reference location, which is a real scenario, the reference current can be obtained directly. If a short circuit occurs at the reference location, which is a simulated scenario, the output of the load mechanism model can be used as the reference current based on the reference information.
[0041] Specifically, the second training loss can include a first sub-loss and a second sub-loss. The first sub-loss can be calculated based on the reference predicted current, the reference current, and the mean square error loss. The second sub-loss is calculated as follows: a mask vector of size 1*M is constructed (M is the total number of dynamic operating parameters and device characteristic parameters). In the mask vector, the first J elements are 1 (J is the number of dynamic operating parameters), and the other elements are 0. After iterating to obtain the updated reference input vector, the reference input vector before the update is multiplied by the mask vector to obtain the first multiplication result. The updated reference input vector is then multiplied by the mask vector to obtain the second multiplication result. The second sub-loss is calculated based on the first multiplication result, the second multiplication result, and the mean square error loss function. The second sub-loss aims to ensure that the reference values corresponding to the dynamic operating parameters remain unchanged and only the device characteristic parameters are updated.
[0042] It should be noted that, based on the second training loss, when updating the generated values of the device characteristic parameters in the reference input vector, the model parameters of the trained proxy model are fixed, and the device characteristic parameters can be updated normally using methods such as stochastic gradient descent.
[0043] Compared to directly using an optimization algorithm based on an equivalent model for parameter optimization, this embodiment can greatly save the time spent on parameter optimization and improve the efficiency of parameter optimization.
[0044] Finally, the reference values of each device characteristic parameter are assigned to the corresponding device characteristic parameter to obtain the updated equivalent model. The updated equivalent model only needs to input the actual values corresponding to the dynamic operating parameters to obtain the short-circuit predicted current.
[0045] S105. Verify the updated equivalent model.
[0046] For any of the preset multiple (N) verification locations, the verification input vector is determined based on the updated equivalent model and the verification information obtained when a short circuit occurs at that verification location. The current prediction error is determined based on the verification input vector, the verification current obtained when a short circuit occurs at that verification location, and the updated equivalent model.
[0047] The verification location can also be represented by the distance between the load connection location and the short-circuit location. If the short circuit at the verification location is a real scenario, the verification current can be directly obtained. If the short circuit at the verification location is a simulated scenario, the output result of the load mechanism model can be used as the verification current based on the verification information input.
[0048] The verification information includes the verification voltage, the peak value of the internal potential, the verification short-circuit time point, the verification recovery time point, and the current verification time point.
[0049] Specifically, methods for determining the verification input vector include: Based on the verification information, values are assigned to the dynamic operating parameters to obtain the verification values corresponding to each dynamic operating parameter; Based on the updated equivalent model, the equipment characteristic parameters are assigned values to obtain the verification values corresponding to the equipment characteristic parameters respectively; The verification input vector is formed by the verification values corresponding to the dynamic operating parameters and the verification values corresponding to the equipment characteristic parameters.
[0050] The verification information may include the verification values of each dynamic operating parameter when a short circuit occurs at the verification location.
[0051] Specifically, methods for determining current prediction error include: The validation input vector is then input into the updated equivalent model to obtain the validation predicted current. Based on the verification predicted current and the verification current, the current prediction error corresponding to the verification location is calculated.
[0052] Among them, the current prediction error can represent the difference between the short-circuit current predicted by the load mechanism model and the equivalent model.
[0053] Specifically, the current prediction error can be obtained by comparing the absolute value of the difference between the predicted current and the verification current with the verification current to achieve a normalization effect.
[0054] S106. Determine the applicable range of the updated equivalent model.
[0055] Based on the current prediction errors corresponding to the N verification locations, the applicable range of the equivalent model is determined.
[0056] Where N is a positive integer, the applicable range can be used to determine whether the short-circuit current prediction at the target location using the updated equivalent model is reliable.
[0057] Specifically, this step includes: Obtain the current prediction error corresponding to multiple (N) verification locations; The N current prediction errors are clustered using a preset clustering method to obtain R error cluster sets, where R is a positive integer; Based on the mean error of each error cluster set, the error cluster set with the smallest mean error is determined as the first target set; The verification locations corresponding to each current prediction error in the first target set are clustered using a preset clustering method to obtain S location cluster sets, where S is a positive integer; Based on the number of verification locations contained in each location cluster set, the location cluster set with the largest number of verification locations is determined as the second target set; The applicable range of the equivalent model is determined based on the maximum and minimum values at each verification location in the second target set.
[0058] The preset clustering method can use the DBSCAN clustering algorithm, which can perform clustering without setting the number of cluster sets.
[0059] Specifically, in this embodiment, the smaller the mean error, the more reliable the short-circuit current prediction using the updated equivalent model when a short circuit occurs at each verification location in the corresponding error cluster set. Therefore, the error cluster set with the smallest mean error is determined as the first target set. According to the actual request, the implementer can select the top W error cluster sets with the smallest mean error as the first target set, where W is a positive integer.
[0060] The purpose of clustering the verification locations corresponding to each current prediction error in the first target set using a preset clustering method is to eliminate some outliers in order to ensure the reliability of the applicable range.
[0061] S107. When the short circuit point is within the applicable range, the updated equivalent model is used to predict the short circuit current; otherwise, the load mechanism model is used to predict the short circuit current.
[0062] When the target location of the short circuit is within the applicable range, the updated equivalent model is selected for short-circuit current prediction.
[0063] When the target location of the short circuit is within the applicable range, the accuracy of short circuit current prediction using the updated equivalent model is similar to that using the load mechanism model. Therefore, the updated equivalent model is used to predict the short circuit current, which improves the efficiency of short circuit current prediction.
[0064] When the short-circuit point is outside the reference location, a load mechanism model is selected for short-circuit current prediction.
[0065] When the target location of the short circuit is outside the applicable range, the accuracy of short circuit current prediction using the updated equivalent model differs significantly from that using the load mechanism model. Therefore, the load mechanism model is used for short circuit current prediction to ensure the accuracy of short circuit current prediction.
[0066] In this embodiment, the update of the equivalent model is determined by using a proxy model and training method. Compared with the method of directly optimizing parameters based on the equivalent model, this method can effectively improve the efficiency of parameter optimization. Based on the current prediction error corresponding to multiple verification locations, the applicable range is determined. By using the applicable range and the target location where a short circuit occurs, the load mechanism model or the updated equivalent model is selected. When the current prediction accuracy of the updated equivalent model meets the requirements, the updated equivalent model can be selected to improve the efficiency of short circuit current prediction.
[0067] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0068] In one embodiment, a load current prediction device for short-circuit simulation is provided, which corresponds one-to-one with the prediction methods described in the above embodiments. For example... Figure 3 As shown, the prediction module includes a model acquisition module, a model parsing module, a model construction module, a model training module, a reference value determination module, a model update module, an error calculation module, an interval determination module, a first model selection module, and a second model selection module. Detailed descriptions of each functional module are as follows: The model parsing module 301 is used to construct an equivalent model based on the load mechanism model, and to parse the equivalent model to obtain equipment characteristic parameters and dynamic operating parameters. The proxy model construction module 302 is used to construct a proxy model based on the device characteristic parameters, dynamic operating parameters, and the output parameters of the equivalent model. The proxy model training module 303 is used to train the proxy model based on preset samples and preset labels; The equivalent model update module 304 is used to fit the equipment characteristic parameters of the equivalent model using the surrogate model and update the equivalent model. The equivalent model verification module 305 is used to verify the updated equivalent model; The applicable range determination module 306 is used to determine the applicable range of the updated equivalent model; The current prediction module 307 is used to predict the short-circuit current using the updated equivalent model when the short-circuit point is within the applicable range, and to predict the short-circuit current using the load mechanism model when the short-circuit point is outside the applicable range.
[0069] For specific limitations regarding the prediction device, please refer to the limitations of the prediction method above, which will not be repeated here. Each module in the aforementioned prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in hardware or independently of the processor in a computer device, or stored in software in the memory of a computer device, so that the processor can call and execute the operations corresponding to each module.
[0070] In one embodiment, a computer device server is provided, the internal structure of which can be shown in the following diagram. Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores preset samples, preset tags, reference information, reference currents, verification information, and verification currents. The network interface communicates with external terminals via a network connection. The computer program is executed by the processor to implement the prediction method.
[0071] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the prediction method described in the above embodiments, for example... Figure 1 S101-S107, as shown, will not be described again here to avoid repetition. Alternatively, the processor, when executing the computer program, implements the functions of each module / unit in this embodiment of the prediction device, for example... Figure 3 The functions of 301-307 shown will not be repeated here to avoid duplication.
[0072] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When executed by a processor, the computer program implements the prediction method described in the above embodiment, for example... Figure 1 S101-S107, as shown, will not be described again here to avoid repetition. Alternatively, when the computer program is executed by the processor, it implements the functions of each module / unit in this embodiment of the prediction device, for example... Figure 3 The functions of 301-307 shown will not be repeated here to avoid duplication.
[0073] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.
[0074] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0075] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A load current prediction method for short-circuit simulation, characterized in that, include: An equivalent model is constructed based on the load mechanism model, and the equipment characteristic parameters and dynamic operating parameters are obtained by analyzing the equivalent model. Construct a proxy model, which takes device characteristic parameters and dynamic operating parameters as input and output parameters of an equivalent model as output; The proxy model is trained based on preset samples and preset labels; The equipment characteristic parameters of the equivalent model are fitted using a surrogate model, and the equivalent model is then updated. Determine the applicable range of the updated equivalent model; When the short circuit point is within the applicable range, the updated equivalent model is used to predict the short circuit current; otherwise, the load mechanism model is used to predict the short circuit current.
2. The load current prediction method, apparatus, equipment, and medium for short-circuit simulation according to claim 1, characterized in that, Methods for training surrogate models include: Initial values are randomly generated for the input parameters based on physical constraints and preset boundaries. The initial values corresponding to the input parameters are concatenated into a sample vector, which is then used as a preset sample. Based on the initial values and equivalent models corresponding to the input parameters, the sample output values of the equivalent model are calculated, and the sample output values are used as preset labels. The preset sample is input into the surrogate model for prediction to obtain the sample current; The first training loss is calculated based on the sample current, preset label, and mean square error loss function. The surrogate model is then trained based on the first training loss until the first training loss converges, resulting in a well-trained surrogate model.
3. The load current prediction method for short-circuit simulation according to claim 1, characterized in that, The equipment characteristic parameters include the impedance value of the constant impedance load, the power value of the constant power load, the internal impedance of the induction motor, the current value of the constant current load, the first attenuation coefficient, and the second attenuation coefficient; the dynamic operating parameters include the system voltage, the peak value of the internal potential, the current time point, the short circuit time point, and the recovery time point.
4. The load current prediction method for short-circuit simulation according to claim 1, characterized in that, Methods for fitting equipment characteristic parameters of equivalent models using surrogate models include: Obtain reference information and reference current when a short circuit occurs at the reference location; The dynamic operating parameters are assigned values based on the reference information to obtain the reference values corresponding to each dynamic operating parameter. Generate corresponding values for each device characteristic parameter, and form a reference input vector by combining the reference values corresponding to the reference information and the generated values corresponding to the device characteristic parameters. The reference input vector is input into the trained surrogate model to obtain the reference predicted current; Based on the reference predicted current, the reference current, and the preset loss function, the second training loss is calculated. Based on the second training loss, the generated values of the device characteristic parameters in the reference input vector are updated until the second training loss converges, and the reference values corresponding to the device characteristic parameters are obtained.
5. The load current prediction method for short-circuit simulation according to claim 1, characterized in that, Methods for determining the applicable range of the updated equivalent model include: Obtain the current prediction error corresponding to multiple verification locations; The current prediction error is clustered using a preset clustering method to obtain an error cluster set; Based on the mean error of each error cluster set, the error cluster set with the smallest mean error is determined as the first target set; The verification locations corresponding to each current prediction error in the first target set are clustered using a preset clustering method to obtain a location cluster set; Based on the number of verification locations contained in each location cluster set, the location cluster set with the largest number of verification locations is determined as the second target set; The applicable range of the equivalent model is determined based on the maximum and minimum values at each verification location in the second target set.
6. The load current prediction method for short-circuit simulation according to claim 1, characterized in that, Also includes: The updated equivalent model is validated.
7. The load current prediction method for short-circuit simulation according to claim 6, characterized in that, The method for verifying the updated equivalent model includes: for any one of the preset multiple verification locations, determining the verification input vector based on the updated equivalent model and the verification information obtained when a short circuit occurs at that verification location; and determining the current prediction error based on the verification input vector, the verification current obtained when a short circuit occurs at that verification location, and the updated equivalent model.
8. A load current prediction device for short-circuit simulation, characterized in that, include: The model parsing module is used to construct an equivalent model based on the load mechanism model, and to parse the equivalent model to obtain equipment characteristic parameters and dynamic operating parameters; The proxy model building module is used to build a proxy model based on device characteristic parameters, dynamic operating parameters, and the output parameters of the equivalent model. The proxy model training module is used to train the proxy model based on preset samples and preset labels. The equivalent model update module is used to fit the equipment characteristic parameters of the equivalent model using the surrogate model and update the equivalent model. The applicable range determination module is used to determine the applicable range of the updated equivalent model; The current prediction module is used to predict short-circuit current using the updated equivalent model when the short-circuit point is within the applicable range, and to predict short-circuit current using the load mechanism model when the short-circuit point is outside the applicable range.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the prediction method according to any one of claims 1-7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the prediction method according to any one of claims 1-7.