Power grid transient equivalent modeling method and system based on grnn stepwise regression, electronic device and medium

Abstract

The present application relates to the technical field of power system regulation, and more particularly to a power transmission network transient equivalent modeling method and system based on GRNN step-by-step regression, an electronic device and a medium; the method comprises: constructing a power transmission network transient equivalent mathematical model; establishing a detailed simulation model of the power transmission network containing a high proportion of new energy units, and dividing the sample set into a training set and a test set; using the training set to perform step-by-step training on the generalized regression neural network; after the training is completed, calling the regression model, calculating the short-circuit current component of the second branch based on the number of low-penetration units and the voltage correction parameter, and calculating the short-circuit current component of the first branch based on the equivalent impedance, and synthesizing the total short-circuit current. In this way, the technical problem of insufficient short-circuit current calculation accuracy of the existing power transmission network transient equivalent modeling technology in the high proportion of new energy access scene is solved, and the accuracy, adaptability and reliability of the calculation results of the equivalent model are improved.

Description

Technical Field

[0001] This invention relates to the field of power system control technology, and in particular to a method, system, electronic device and medium for transient equivalent modeling of transmission networks based on GRNN stepwise regression. Background Technology

[0002] With the rapid development and large-scale application of renewable energy technologies, the penetration rate of new energy units such as wind power and photovoltaics in power transmission networks has significantly increased, placing higher demands on the accuracy of power system short-circuit current calculations. In traditional power transmission networks, due to the significant capacity difference between the transmission and distribution networks, the transmission network is often considered an equivalent "infinite system" with extremely high inertia and constant voltage amplitude and frequency. Its effect on the distribution network can be simplified to an equivalent model consisting of an ideal voltage source connected in series with an equivalent impedance. The distribution network is located at the end of the power system, and short-circuit faults occurring in it can be regarded as remote short circuits from the perspective of the transmission network. The aforementioned traditional equivalent model has been widely used in distribution network short-circuit current calculations due to its simple structure and clear parameters.

[0003] However, with the grid connection of large-scale renewable energy units in the transmission network, when a short-circuit fault occurs in the distribution network near the transmission-distribution boundary, the voltage at the grid connection point of the adjacent renewable energy units in the transmission network may drop, and the units enter a low-voltage ride-through state. During the low-voltage ride-through, renewable energy units need to maintain grid-connected operation and inject dynamic reactive current into the system to support voltage recovery. At this time, their transient characteristics exhibit voltage-controlled current source behavior, which differs significantly from the characteristics of the ideal voltage source series impedance in the traditional equivalent model of the transmission network. In the transient equivalent modeling of transmission networks with a high proportion of renewable energy, especially for short-circuit faults near the transmission-distribution boundary, the traditional equivalent model fails to account for the nonlinear transient response characteristics of renewable energy units, leading to deviations in the short-circuit current calculation results.

[0004] In the field of equivalent modeling for new energy sources, existing technologies have established a certain foundation. Reference document 1 (application publication number CN118886328B) discloses an equivalent modeling method suitable for the safe and stable operation of new energy power plants. This method transforms the original topology into a radial structure by establishing the road matrix and impedance matrix of the power plant's collector lines, and performs clustering based on the terminal voltage vector at fault times, ultimately achieving equivalent modeling of the new energy power plant. This method focuses on equivalent optimization within the new energy power plant, improving accuracy through clustering algorithms. However, its application scenarios are concentrated at the power plant level, failing to fully consider the coordination issues between new energy units and synchronous units in the transient equivalent model at the transmission network level, and particularly lacking a unified representation of the transient characteristics of new energy units under short-circuit faults at the transmission and distribution boundary. Specifically, existing technologies suffer from two prominent problems: First, traditional transmission network equivalent models fail to integrate the voltage-controlled current source characteristics of renewable energy units, resulting in insufficient applicability of short-circuit current calculation models in scenarios with high renewable energy penetration. Second, existing renewable energy equivalent methods primarily focus on internal substation optimization, making it difficult to directly extend to transient equivalent modeling of transmission networks. This fails to address the coupling problem between the transient response of renewable energy units and system equivalent parameters during transmission-distribution boundary faults. These intertwined issues make it difficult for existing equivalent technologies to achieve accurate short-circuit current calculations in transmission networks with a high proportion of renewable energy, affecting not only the reliability of fault analysis but also potentially limiting the effectiveness of protection settings and safety control strategies. Therefore, existing transmission network transient equivalent modeling technologies suffer from insufficient short-circuit current calculation accuracy in scenarios with high renewable energy integration. Summary of the Invention

[0005] To address the aforementioned shortcomings or deficiencies, this invention provides a method, system, electronic device, and medium for transient equivalent modeling of power transmission networks based on stepwise regression using GRNN (General Regression Neural Network). This method can solve the technical problem of insufficient short-circuit current calculation accuracy in existing transient equivalent modeling technologies for power transmission networks under scenarios with a high proportion of renewable energy access.

[0006] This invention provides a method for transient equivalent modeling of power transmission networks based on GRNN stepwise regression, comprising: A transient equivalent mathematical model of the power transmission network is constructed. The mathematical model includes a first branch representing the short-circuit current response characteristics of synchronous generator units and a second branch representing the short-circuit current response characteristics of new energy generator units.

[0007] A detailed simulation model of the power transmission network with a high proportion of new energy generating units was established. Transmission and distribution boundary nodes were selected as observation points. Multiple short-circuit fault scenarios were generated by adjusting the impedance parameters of the distribution network lines. Voltage data of transmission and distribution boundary nodes were collected based on the short-circuit fault scenarios to form a sample set, which was then divided into a training set and a test set.

[0008] The generalized regression neural network is trained step-by-step using the training set to generate a first regression model, a second regression model, and a third regression model with different equivalent parameters for the regression mathematical model.

[0009] In response to the completion of training of the generalized regression neural network, the voltage data of the test set is input into the regression model. The first regression model is called to regress the number of low-voltage units, the second regression model is called to regress the voltage correction parameters, and the third regression model is called to regress the equivalent impedance. Based on the number of low-voltage units and the voltage correction parameters, the short-circuit current component of the second branch is calculated, and based on the equivalent impedance, the short-circuit current component of the first branch is calculated, and the total short-circuit current is synthesized.

[0010] According to a second aspect, this invention provides a power grid transient equivalent modeling system based on GRNN stepwise regression, comprising: The mathematical model construction module is used to construct a transient equivalent mathematical model of the power transmission network. The mathematical model includes a first branch representing the short-circuit current response characteristics of synchronous generator units and a second branch representing the short-circuit current response characteristics of new energy generator units.

[0011] The simulation model building module is used to establish a detailed simulation model of the power transmission network containing a high proportion of new energy units. It selects the transmission and distribution boundary nodes as observation points, generates multiple short-circuit fault scenarios by adjusting the impedance parameters of the distribution network lines, collects voltage data of the transmission and distribution boundary nodes based on the short-circuit fault scenarios to form a sample set, and divides the sample set into a training set and a test set.

[0012] The GRNN step-by-step training module is used to train the generalized regression neural network step by step using the training set, generating a first regression model, a second regression model, and a third regression model with different equivalent parameters for the regression mathematical model.

[0013] The transient equivalent modeling result output module is used to input the voltage data of the test set into the regression model when the generalized regression neural network has been trained. It sequentially calls the first regression model to regress the number of low-voltage units, the second regression model to regress the voltage correction parameters, and the third regression model to regress the equivalent impedance. Based on the number of low-voltage units and the voltage correction parameters, it calculates the short-circuit current component of the second branch and the short-circuit current component of the first branch based on the equivalent impedance, and synthesizes the total short-circuit current.

[0014] According to a third aspect, the present invention provides an electronic device comprising: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to execute any of the GRNN-based stepwise regression-based power transmission network transient equivalent modeling methods in the embodiments of the present invention.

[0015] According to another aspect of the present invention, a non-transient computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute any of the GRNN-based stepwise regression-based transient equivalent modeling methods for power transmission networks in the embodiments of the present invention.

[0016] The present invention provides a method for transient equivalent modeling of power transmission networks based on stepwise regression using a generalized regression neural network (GRNN). This method is achieved through four core steps: constructing a mathematical model, generating a sample set, training the regression model stepwise, and synthesizing the short-circuit current. Specifically, a transient equivalent mathematical model containing a first branch and a second branch is constructed to characterize the differentiated response characteristics of synchronous generating units and renewable energy generating units under short-circuit faults, overcoming the limitations of traditional single ideal voltage source models in scenarios with high proportions of renewable energy integration. A detailed simulation model is established, and a multi-scenario sample set is generated by adjusting line impedance parameters to simulate changes in the location of the short-circuit fault point, ensuring the coverage and representativeness of the training data. The training set is used to train the generalized regression neural network stepwise to generate three independent regression models for accurately regressing different equivalent parameters in the mathematical model, avoiding the coupling effects between parameters. After the model training is completed, the regression model is called stepwise to obtain parameters and calculate the current components of the two branches, then the total short-circuit current is synthesized, realizing a closed-loop process from boundary voltage observation to accurate short-circuit current calculation.

[0017] In this technical solution, the present invention addresses the problem that traditional equivalent models, as described in the background art, cannot accurately reflect the transient characteristics of new energy generating units. By constructing a transient equivalent mathematical model that includes a second branch of a voltage-controlled current source, it achieves an accurate description of the current response characteristics of new energy generating units under low-voltage ride-through conditions, thus solving the defect of insufficient applicability of traditional models for short-circuit current calculation due to neglecting this characteristic. Furthermore, addressing the issues of variable operating modes and complex fault scenarios brought about by new energy access, the invention establishes a detailed simulation model and adjusts line impedance to generate diverse sample sets, constructing a training data foundation that fully reflects changes in the electrical distance to the fault location. This solves the problem of traditional methods being unable to accurately reflect the transient characteristics of new energy generating units. The invention addresses the problem of weak model generalization ability due to the limited scope of the scenario; it also addresses the difficulty in accurately obtaining equivalent parameters due to the intertwined characteristics of new energy units and synchronous units. By employing a generalized regression neural network for step-by-step training and generating three independent regression models, it achieves high-precision, decoupled regression of the number of low-voltage units, voltage correction parameters, and equivalent impedance, overcoming the shortcomings of traditional methods such as low parameter fitting accuracy and mutual error interference. Furthermore, it addresses the complexity and difficulty in guaranteeing accuracy in short-circuit current calculation by step-by-step regression of parameters and sequential calculation of the current components of the two branches before synthesizing the total current. This establishes a clear and traceable calculation process, ensuring the reliability of the final short-circuit current result. Therefore, the technical solution of this invention solves the technical problem of insufficient short-circuit current calculation accuracy in existing power grid transient equivalent modeling techniques under high-proportion new energy access scenarios, improving the accuracy, adaptability, and reliability of the equivalent model and calculation results. Attached Figure Description

[0018] Figure 1 This is a flowchart of a method for transient equivalent modeling of power transmission networks based on stepwise regression of GRNN according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the principle of the equivalent parameter stepwise regression and short-circuit current calculation process according to another embodiment of the present invention; Figure 3 This is a schematic diagram of the complete process of equivalent parameter stepwise regression and short-circuit current calculation according to another embodiment of the present invention; Figure 4 This is a schematic diagram of the execution flow for calculating the short-circuit current using an equivalent model according to another embodiment of the present invention; Figure 5 This is a schematic diagram of a typical topology of a transmission network containing a high proportion of new energy generating units, according to another embodiment of the present invention. Figure 6 This is a schematic diagram of the structure of a power grid transient equivalent modeling system based on GRNN stepwise regression according to an embodiment of the present invention; Figure 7 This is a block diagram of an electronic device used to implement embodiments of the present invention. Detailed Implementation

[0019] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0020] During the development of this invention, the inventors, through extensive experiments and data analysis, revealed the intrinsic relationship between the voltage characteristics of transmission and distribution boundary nodes and the transient response characteristics of renewable energy units: the electrical distance between the short-circuit fault point and the transmission and distribution boundary node directly determines the degree of boundary voltage drop, thereby affecting the low-voltage ride-through state switching and short-circuit current output characteristics of renewable energy units. Based on this relationship, the inventors innovatively proposed this technical solution, utilizing the nonlinear fitting capability of a generalized regression neural network (GRNN), by training three independent regression models in stages, and combining an equivalent model structure of a voltage-controlled current source branch and an ideal voltage source branch in parallel, thereby achieving accurate calculation of short-circuit current in transmission networks with a high proportion of renewable energy, embodying the core concept of "multi-parameter decoupled regression and step-by-step collaborative calculation".

[0021] Specifically, through comparative experiments, the invention team discovered that traditional equivalent modeling methods for power transmission networks cannot reflect the low-voltage ride-through characteristics of renewable energy units. Traditional ideal voltage source series impedance models are only applicable to synchronous units, while renewable energy units exhibit voltage-controlled current source characteristics during voltage dips, with their short-circuit current output showing a non-linear relationship with the grid connection voltage. These technical deficiencies lead to a significant increase in short-circuit current calculation errors in scenarios with a high proportion of renewable energy integration, affecting the accuracy of protection setting. This invention proposes a transient equivalent model that includes a voltage-controlled current source branch, and performs stepwise regression of the number of low-voltage ride-through units, voltage correction parameters, and equivalent impedance based on boundary voltage amplitude and phase characteristics. This improves the model's accuracy in representing the transient characteristics of renewable energy units, achieving a technical effect where the decoupled calculation errors for the active and reactive components of the short-circuit current are both below 0.06%.

[0022] Therefore, this invention provides a transient equivalent modeling method for power transmission networks based on GRNN stepwise regression, which can be applied to a power system short-circuit current analysis and calculation platform (hereinafter referred to as the "system"). This system can run in a power system simulation and analysis environment via a local server cluster or cloud computing to complete transient equivalent modeling and accurate short-circuit current calculation for power transmission networks with a high proportion of renewable energy.

[0023] Specifically, this system can be deployed in various hardware environments, including but not limited to: dedicated servers in power dispatch centers, edge computing devices in new energy power plant control centers, and high-performance computing platforms in research institutes. This flexible deployment architecture allows the system to meet both the high real-time requirements of online analysis applications and the needs of offline simulation research. In terms of operation, the system achieves functional decoupling of data acquisition, model training, parameter regression, and current calculation through modular design, thereby improving the maintainability and scalability of the calculation process.

[0024] like Figure 1 As shown, the method may include: Step S110: Construct a transient equivalent mathematical model of the power transmission network.

[0025] The mathematical model includes a first branch representing the short-circuit current response characteristics of synchronous generator units and a second branch representing the short-circuit current response characteristics of renewable energy generator units. The first branch is configured as an ideal voltage source connected in series with an equivalent impedance to characterize the short-circuit current response characteristics of synchronous generator units in the transmission network during short-circuit faults. The second branch is configured as a voltage-controlled current source to characterize the short-circuit current response characteristics of all renewable energy generator units in the transmission network during short-circuit faults. Renewable energy generator units include units that have entered the low-voltage ride-through state and units that have not entered the low-voltage ride-through state.

[0026] Specifically, the system can be obtained through the voltage phasor and equivalent impedance (including the resistive component) of an ideal voltage source. With reactance component A mathematical model is constructed by combining the control function of the voltage-controlled current source (with the voltage of the transmission and distribution boundary node as input).

[0027] For example, in the IEEE 39-node test system, node 16 is used as the transmission and distribution boundary node, and the voltage phasor of the ideal voltage source in the first branch is set as follows: The equivalent impedance is (per unit value), the voltage-controlled current source control function of the second branch is implemented based on formula (3) described below.

[0028] Step S120: Establish a detailed simulation model of the power transmission network with a high proportion of new energy units, select the transmission and distribution boundary nodes as observation points, generate multiple short-circuit fault scenarios by adjusting the impedance parameters of the distribution network lines, collect voltage data of the transmission and distribution boundary nodes based on the short-circuit fault scenarios to form a sample set, and divide the sample set into a training set and a test set.

[0029] Among them, high proportion of new energy units refers to scenarios where the installed capacity of new energy units accounts for no less than 30% of the total system capacity; transmission and distribution boundary nodes are the physical connection points where the distribution network connects to the transmission network, usually located in areas with concentrated loads; short-circuit impedance is composed of the impedance of the step-down transformer and the impedance of the distribution network line in series, and the change of electrical distance between the short-circuit fault point and the transmission and distribution boundary node is simulated by changing the line impedance value; the sample set includes a training set and a test set, and their ratio is usually set to 4:1.

[0030] Specifically, the system can use a simulation platform (such as electromagnetic transient simulation software) to set distribution network lines of different lengths (corresponding impedance values ​​from...) to ), generating 1601 short-circuit fault scenarios, and collecting the voltage amplitude at the transmission and distribution boundary nodes in each scenario. (unit: ) and phase angle (Unit: degrees), forming a voltage dataset.

[0031] For example, the system sets the line impedance step value in the simulation to be... A total of 1601 samples were generated, of which 1280 were used as the training set and 321 were used as the test set for subsequent model training and validation.

[0032] In some embodiments, the system can calculate the relationship between the voltage phasor of the transmission and distribution boundary node and the total short-circuit current when a short-circuit fault occurs by the following formula (1); (1) Formula (1) describes the voltage-current relationship on the short-circuit impedance formed by the series connection of the step-down transformer impedance and the distribution network line impedance, and is used to establish the electrical equation of the fault observation point. This represents the voltage phasor at the transmission and distribution boundary node during a short-circuit fault. Voltage amplitude (unit: ), Voltage phase angle (unit: degrees); This represents the total short-circuit current phasor flowing through this short-circuit impedance. Active component of current (unit: ), The reactive component of the current (unit: ); Indicates short-circuit impedance. Resistance component (unit: ), Reactance component (unit: ).

[0033] Next, in this embodiment, the system can also decompose the total short-circuit current into components contributed by different power supply types using formula (2); (2) Formula (2) embodies the core idea of ​​the dual-branch equivalent model constructed in this invention, namely, the total short-circuit current is provided by the new energy unit branch and the synchronous unit branch. This represents the short-circuit current phasor output from the voltage-controlled current source branch, and represents the contribution of all new energy units. This represents the short-circuit current phasor output from the ideal voltage source branch, and represents the contribution of the synchronous generator unit.

[0034] Furthermore, in this embodiment, the system can also calculate the short-circuit current output of the new energy unit branch through the voltage-controlled current source control function defined by formula (3); (3) Formula (3) is a custom voltage-controlled current source control function used to characterize the nonlinear current response characteristics of new energy units under short-circuit faults, especially when entering the low-voltage ride-through state. The control function is represented, and its specific form is determined by the grid connection standards and dynamic characteristics of the new energy unit. This indicates the number of new energy generating units that enter low-voltage ride-through mode during a fault (unit: number). This parameter reflects the scale of generating units affected by voltage dips.

[0035] Furthermore, in this embodiment, the system can also calculate the short-circuit current output of the synchronous generator branch using formula (4); (4) Formula (4) describes the relationship between voltage drop and current in the synchronous generator branch, following Ohm's law. The voltage phasor representing an ideal voltage source is typically set as follows: (unit: ); This represents the equivalent impedance of the synchronous generator branch. Resistance component (unit: ), Reactance component (unit: ).

[0036] Therefore, by combining the above formulas (1) to (4), the system can fully construct and solve the equivalent model of the transmission network that takes into account the nonlinear transient characteristics of new energy units, thereby accurately calculating the total short-circuit current when a short-circuit fault occurs near the transmission and distribution boundary in the scenario of high proportion of new energy access.

[0037] Step S130: Use the training set to train the generalized regression neural network step by step to generate a first regression model, a second regression model, and a third regression model with different equivalent parameters for the regression mathematical model.

[0038] Step-by-step training refers to independently training three GRNN models to avoid coupling between parameters; the first regression model is used to regress the number of low-altitude units. (Unit: number), its input is the voltage amplitude of the transmission and distribution boundary node. The second regression model is used to regress the voltage correction parameters. and (Dimensionless), its input is The third regression model is used to regress equivalent impedance. and (unit: Its input is the sinusoidal component of the voltage phasor. Sum and cosine components .

[0039] Specifically, the system can first use training set data (voltage amplitude U and the corresponding number of low-voltage units) Train the first regression model, then use the same... and corresponding , Train the second regression model and finally use and corresponding , The third regression model was trained; the radial basis function was used as the activation function during training, and the mean absolute percentage error was used as the training termination condition (the threshold was set to 0.01%).

[0040] For example, the training results show that the first regression model is effective. The regression mean error of the first model is 0%, and the second regression model is... , The average errors were 0.0006% and 0.0002%, respectively. The third regression model... , The average errors were 0.0055% and 0.0009%, respectively (see Table 2 below).

[0041] Step S140: In response to the completion of training of the generalized regression neural network, the voltage data of the test set is input into the regression model. The first regression model is called to regress the number of low-voltage units, the second regression model is called to regress the voltage correction parameters, and the third regression model is called to regress the equivalent impedance. The short-circuit current component of the second branch is calculated based on the number of low-voltage units and the voltage correction parameters, and the short-circuit current component of the first branch is calculated based on the equivalent impedance. The total short-circuit current is then synthesized.

[0042] Among them, the number of low-profile units The number of new energy generating units that enter low-voltage ride-through mode during a short-circuit fault; voltage correction parameters. , Used to correct the error between the voltage at the transmission and distribution boundary node and the voltage at the grid connection point of new energy units; the short-circuit current component includes the active component. (unit: and reactive components (unit: ).

[0043] Specifically, the system can calculate through the following process: First, the test set voltage... Input the first regression model to get Then Input the second regression model to get , And calculate the current in the second branch. Then Inputting the third regression model yields , Calculate the current in the first branch. Finally, the currents of the two branches are added together to obtain the total current. .

[0044] For example, in the test set The sample was calculated to obtain , The relative errors between the actual and simulated values ​​were 0.053% and 0.002% respectively (see Table 3 below), verifying the accuracy of the method.

[0045] In some embodiments, the system can determine the control strategy of the equivalent voltage-controlled current source characterizing the transient characteristics of all new energy units using the following formula (5): (5) Formula (5) defines the control function framework of the equivalent voltage-controlled current source, which is used to aggregate the transient current response of all new energy units. This represents the total output current function of the equivalent voltage-controlled current source; This indicates the number of new energy generating units that enter low-voltage ride-through mode when a short-circuit fault occurs (unit: units). This represents the voltage magnitude at the observation point (usually a transmission and distribution boundary node) (unit: ); This indicates the total number of new energy generating units in the power transmission network (unit: units). , They represent the first Taiwan and the Rated capacity of new energy units (unit: ); , These represent the first and second faults during a short circuit fault. Taiwan and the Voltage at grid connection point of new energy units (unit: ); For a single renewable energy unit in low-voltage ride-through state; This is the control function for a single new energy unit when it has not entered the low-voltage ride-through state (i.e., normal operation state).

[0046] Next, in this embodiment, the system can also calculate the total active current and total reactive current injected into the grid by the new energy unit group using formulas (6) and (7), respectively: Formula (6) is used to calculate the total active current injected into the system by all new energy units. (unit: ). Indicates the first Active current components of the unit (not yet in low-voltage ride-through state); Indicates the first The active current component of the unit (entering low-voltage ride-through state); , The first Taiwan and the Rated current of the unit (unit: ); For the first The current limiting proportional coefficient (dimensionless) for the unit is specified by the unit's grid connection standard. For the first Reactive current components of the generator set (unit: This formula shows that units that have not entered the low-voltage ride-through process output active power at their rated current, while the active current of units that have entered the low-voltage ride-through process is limited by the geometric relationship between the maximum current and the reactive current. Formula (7) is used to calculate the total reactive current injected into the system by all new energy units. (unit: ). Indicates the first The reactive current component of the generator set; Indicates the first The reactive current component of the generator set; For the first The dynamic reactive current proportional coefficient (dimensionless) of the generator unit is specified by the grid connection standard; the constant 0.9 is the voltage threshold for low-voltage ride-through starting (unit: This formula indicates that units that have not entered low-voltage ride-through do not provide reactive power support. Units entering low-voltage ride-through need to dynamically output reactive current based on the degree of voltage drop, and the total output is limited by its maximum current capacity. .

[0047] Furthermore, in this embodiment, the system can also calculate the rated current of a single new energy unit using formula (8), which serves as the basic parameter for formulas (6) and (7): (8) Formula (8) is the definition of the rated current of a new energy generator unit. , The first Taiwan and the Rated current of the new energy unit (unit: ); , Their rated capacities (unit: ); , Their rated voltages (unit: ); The coefficients used to calculate line current in a three-phase AC system.

[0048] Therefore, by combining the above formulas (5) to (8), the system can accurately model and calculate the active and reactive current output of the equivalent voltage-controlled current source formed by the aggregation of a large number of dispersed new energy units when a short circuit fault occurs in the power grid under the scenario of high proportion of new energy access, starting from the control characteristics of the underlying units. This provides an accurate physical description basis for the subsequent transient equivalent modeling based on GRNN stepwise regression.

[0049] In some embodiments, the system can calculate the active current component of the equivalent voltage-controlled current source using the following formula (9); (9) Formula (9) is used to calculate the total active current component injected into the grid by the equivalent voltage-controlled current source branch (unit: ). This represents the total active current component of the equivalent voltage-controlled current source (unit: ); The equivalent rated current of a new energy power generation group entering low-voltage ride-through state ("low-voltage ride-through group") (unit: ); The equivalent rated current (unit: ) of new energy power plants that have not entered the low-voltage ride-through state ("non-low-voltage ride-through power plants") is given. ); This represents the current limiting proportional coefficient of all units in a low-voltage generator group. The weighted average (dimensionless) reflects the ratio of the maximum allowable current of the entire group to its rated current. Represents the total reactive current component of the equivalent voltage-controlled current source (unit: This formula shows that the total injected active current is composed of the equivalent rated current of the non-low-voltage group. It consists of the sum of the active current contributions of the low-voltage power generation group and the active current contribution of the low-voltage power generation group, and the active current contribution of the low-voltage power generation group is limited by the maximum allowable current. With the already output reactive current The geometric relationship between them.

[0050] Next, in this embodiment, the system can also calculate the reactive current component of the equivalent voltage-controlled current source using formula (10); (10) Formula (10) is used to calculate the total reactive current component injected into the grid by the equivalent voltage-controlled current source branch (unit: ). Represents the total reactive current component of the equivalent voltage-controlled current source (unit: ); This represents the dynamic reactive current proportionality coefficient of all units in a low-voltage generator group. The weighted average value (dimensionless) determines the reactive current injection intensity when the generator group responds to a voltage drop. The voltage amplitude at the observation point (transmission and distribution boundary node) is expressed in units of: The constant 0.9 represents the voltage threshold for low-voltage ride-through startup (unit: This formula indicates that the injected total reactive current takes the smaller of two values: one is the expected injection value determined by the dynamic reactive power coefficient of the generator group and the degree of voltage drop. Secondly, the maximum allowable current limit for this generator group. This reflects the dual constraints of reactive current demand and equipment current limitation in the low-voltage ride-through standard.

[0051] Furthermore, in this embodiment, the system can also calculate the equivalent rated current of the low-voltage generator group and the non-low-voltage generator group using formula (11). (11) Formula (11) is used to calculate the equivalent rated current (unit: ) of two new energy generator groups in different states. ( ), so as to uniformly convert to the voltage level of the observation point. , These represent the equivalent rated currents of low-voltage and non-low-voltage power supply groups, respectively. , They represent the first Taiwan and the Rated capacity of new energy units (unit: ); This indicates the number of new energy generating units that enter low-voltage ride-through mode when a short-circuit fault occurs (unit: units). This indicates the total number of new energy generating units (unit: units). The coefficient for calculating line current in a three-phase AC system; , The voltage correction parameters (dimensionless) for the two generator groups are used to compensate for the voltage at the observation points. The measurement or estimation error between the actual voltage at the grid connection point of each unit; Indicates the rated voltage of the transmission and distribution boundary node (unit: This formula obtains the equivalent current representing the overall power base of the unit group by summing the capacities of all units within each unit group and dividing by the corrected rated voltage.

[0052] Furthermore, in this embodiment, the system can also calculate the weighted average dynamic reactive current ratio coefficient and current limiting ratio coefficient of the low-voltage generator group using formula (12). (12) Formula (12) is used to calculate the weighted average control parameters of the entire low-voltage generator group to reflect the comprehensive characteristics of different capacity units within the group. This represents the dynamic reactive current proportionality coefficient of all units in a low-voltage generator group. The weighted average; This represents the current limiting ratio of all units in a low-voltage generator group. The weighted average; Indicates the first Rated capacity of new energy units (unit: ); , They represent the first The dynamic reactive current proportionality coefficient and current limitation proportionality coefficient of the generating unit (both dimensionless) are parameters typically specified by the unit's manufacturing standards or grid connection specifications. This formula uses a weighted average with the unit's rated capacity as the weight, ensuring that the overall equivalent characteristics are dominated by units with larger capacities, which conforms to physical laws.

[0053] Therefore, by combining the above formulas (9) to (12), the system can accurately model and calculate the active and reactive current output of the equivalent voltage-controlled current source formed by the aggregation of a large number of heterogeneous new energy units when a short circuit fault occurs in the scenario of high proportion of new energy access, starting from the individual control characteristics of the underlying unit. This provides an accurate physical description and calculation basis for the subsequent transient equivalent modeling based on GRNN stepwise regression.

[0054] In other embodiments, Figure 2 This diagram illustrates the principle of the stepwise regression of equivalent parameters and the calculation of short-circuit current. It specifically depicts the core calculation path of transient equivalent modeling based on GRNN. The input quantities are clearly marked on the left side of the diagram, namely the voltage data of the transmission and distribution boundary nodes collected from the detailed simulation model, expressed in complex form as follows: The voltage data was fed in parallel into three regression models trained in stages: the first regression model was used to regress the number of low-voltage units. The second regression model is used to regress the voltage correction parameters; the third regression model is used to regress the equivalent impedance. The diagram clearly shows the output current of the voltage-controlled current source in the second branch of the new energy unit. The calculation process, its functional relationship is as follows: This reflects the output current and the number of low-voltage units. and voltage The mapping relationship between them. At the same time, Figure 2 A model of the synchronous generator branch (first branch) is also depicted, including an ideal voltage source. and equivalent impedance Ultimately, the output currents of the two branches are superimposed through phasor superposition to form the total short-circuit current. This intuitively demonstrates the complete, step-by-step computational logic and data flow of the GRNN-based step-by-step regression-based transient equivalent modeling method for power transmission networks, from voltage input to total current output.

[0055] In other embodiments, Figure 3 This diagram illustrates the complete process of equivalent parameter step-by-step regression and short-circuit current calculation. It details the systematic steps from establishing a detailed simulation model of the transmission network to the final short-circuit current synthesis. The initial step is clearly defined as "establishing a detailed model of the transmission network and setting different short-circuit impedances." Here, "short-circuit impedance" refers to the equivalent impedance value formed by the series connection of the step-down transformer impedance and the variable distribution network line impedance. Changing the line impedance value simulates the change in electrical distance between the short-circuit fault point and the transmission / distribution boundary node. Subsequently, the process enters the data recording stage, namely, "recording short-circuit impedance, renewable energy grid connection point voltage, transmission / distribution boundary node voltage, and short-circuit current." These electrical quantities provide the original data foundation for subsequent modeling. "Renewable energy grid connection point voltage" refers to the node voltage at which the renewable energy unit connects to the transmission network (unit: ...). ).

[0056] The core of the process demonstrates the step-by-step regression and calculation path: First, the number of grid-connected units is calculated based on the voltage at the new energy grid connection point; then, three pre-trained regression models are called sequentially—the input to which is the voltage amplitude at the transmission and distribution boundary node. The output is Model 1 with a low number of turbine units, and the input is... The output is a voltage correction coefficient. , Model 2, and the input is... and The output is an equivalent impedance ( Model 3. Finally, the process synthesizes the calculated current components of the two branches to obtain the total short-circuit current, and terminates at the "End" node. This flowchart intuitively and systematically illustrates the implementation of this embodiment. The technical implementation chain provides clear engineering guidance for the accurate calculation of short-circuit current in scenarios with a high proportion of new energy access.

[0057] In other embodiments, Figure 4 A schematic diagram illustrating the execution flow of short-circuit current calculation using the equivalent model is provided. Figure 4 Presented in the form of process nodes, the complete calculation process of the total short-circuit current is intuitively shown, based on a new set of transmission and distribution boundary node voltage parameters, sequentially calling three trained regression models and finally synthesizing the total short-circuit current. The process begins with the "Start" node, clearly indicating the start of the calculation task. The first step, "Input a new set of voltage parameters into each model," defines the input of the entire calculation process, namely the voltage data to be analyzed obtained from the actual or simulated scenario. Next, the process strictly follows the step-by-step calling logic: "Use Model 1 to perform regression on the number of low-voltage units" to obtain the affected scale of new energy units; "Use Model 2 to perform regression on the voltage correction coefficient" to obtain voltage compensation parameters; subsequently, based on the aforementioned regression results, "Calculate the short-circuit current of the voltage-controlled current source branch" is executed to complete the current contribution calculation of the new energy unit branch. The process continues by calling "Use Model 3 to perform regression on the equivalent impedance" to obtain the impedance parameters of the synchronous unit branch, and accordingly executes "Calculate the short-circuit current of the ideal voltage source branch." Finally, the process summarizes the current components of the two branches through the "Calculate the total short-circuit current" node, and marks the completion of the calculation task with the "End" node. This flowchart concretizes the above step S140 into an executable sequence of engineering operations in a clear and linear manner, providing direct guidance for the field application and system implementation of this technical solution.

[0058] In other embodiments, Figure 5 A schematic diagram of a typical topology of a power transmission network containing a high proportion of renewable energy generating units is shown in a detailed simulation model. Figure 5The diagram vividly depicts the power system network architecture upon which the transient equivalent modeling method for transmission networks is based, showcasing the grid connection of various power generation methods, including traditional synchronous power sources and a high proportion of renewable energy units. The left side of the diagram labels thermal power units representing conventional synchronous generators with the letter "G," while the middle and right sides label wind turbines (blue) and photovoltaic units (green), visually illustrating the penetration and distribution of renewable energy units within the grid. These power generation units are connected by a crisscrossing network of transmission lines, forming a complex power network with multiple nodes and branches. The numerical designations in the diagram identify key system nodes, some of which constitute the "transmission and distribution boundary nodes" described in the technical solution. These nodes are crucial for selecting voltage observation points and performing short-circuit calculations in this transient equivalent modeling method. This topology diagram provides a concrete and realistic physical system reference for constructing detailed simulation models, setting different short-circuit fault scenarios, and collecting boundary node voltage data, validating the applicability of this transient equivalent modeling method for transmission networks in modern power grids with complex power source structures.

[0059] Therefore, according to the above implementation method, the system first achieves its goal through four core steps: constructing a mathematical model, generating a sample set, training the regression model step by step, and synthesizing the short-circuit current. Specifically, a transient equivalent mathematical model containing the first and second branches is constructed to characterize the differentiated response characteristics of synchronous generator units and new energy generator units under short-circuit faults, overcoming the limitations of the traditional single ideal voltage source model in scenarios with a high proportion of new energy access. A detailed simulation model is established, and a multi-scenario sample set is generated by adjusting the line impedance parameters to simulate the changes in the location of the short-circuit fault point, ensuring the coverage and representativeness of the training data. The training set is used to train the generalized regression neural network step by step to generate three independent regression models for accurately regressing different equivalent parameters in the mathematical model, avoiding the coupling effect between parameters. After the model training is completed, the regression model is called step by step to obtain parameters and calculate the current components of the two branches, then the total short-circuit current is synthesized, realizing a closed loop from boundary voltage observation to accurate short-circuit current calculation.

[0060] Specifically, in this implementation, the technical solution addresses the problem that traditional equivalent models, as described in the background technology, cannot accurately reflect the transient characteristics of new energy generating units. By constructing a transient equivalent mathematical model that includes the second branch of the voltage-controlled current source, an accurate description of the current response characteristics of new energy generating units under low-voltage ride-through conditions is achieved, thus solving the deficiency of insufficient applicability of the short-circuit current calculation model due to neglecting this characteristic in traditional models. Furthermore, addressing the issues of variable operating modes and complex fault scenarios brought about by new energy access, a detailed simulation model is established, and diverse sample sets are generated by adjusting the line impedance. This constructs a training data foundation that can fully reflect changes in the electrical distance to the fault location, solving the problem of insufficient applicability of traditional models. This paper addresses the issue of weak model generalization ability due to the limited scope of the current model in a single scenario. It also addresses the difficulty in accurately obtaining equivalent parameters due to the intertwined characteristics of renewable energy units and synchronous generators. By employing a generalized regression neural network for step-by-step training and generating three independent regression models, high-precision, decoupled regression of the number of low-voltage generators, voltage correction parameters, and equivalent impedance is achieved, overcoming the shortcomings of low parameter fitting accuracy and mutual error interference in traditional methods. Furthermore, to address the complexity and difficulty in guaranteeing accuracy in short-circuit current calculation, a clear and traceable calculation process is established by step-by-step regression of parameters and sequential calculation of the current components of the two branches before synthesizing the total current, ensuring the reliability of the final short-circuit current result. Therefore, the technical solution implemented here solves the technical problem of insufficient short-circuit current calculation accuracy in existing power grid transient equivalent modeling techniques under high-proportion renewable energy integration scenarios, improving the accuracy, adaptability, and reliability of the equivalent model and calculation results.

[0061] In some embodiments, the system can calculate the short-circuit impedance parameters used to construct the sample set using the following formulas (13) and (14): (13) (14) Formulas (13) and (14) together define the rules for setting the short-circuit impedance, which are used to simulate fault scenarios with different electrical distances between the short-circuit fault point and the transmission and distribution boundary node through linear variation. Represents the resistive component of short-circuit impedance (unit: ); The reactance component representing the short-circuit impedance (unit: ); It is an integer variable used as the scene index, and its value ranges from 0 to 1600. This variable determines the superposition increment of the line impedance. Specifically, the constant term... and These represent the reference resistance and reference reactance values, respectively, of the step-down transformer impedance; step term and These represent the resistance increment and reactance increment of the distribution network line impedance added to simulate different fault locations, respectively. Each increment of 1 represents an increase of one preset unit length in the line length, corresponding to a specific electrical distance between the fault point and the transmission and distribution boundary node.

[0062] Therefore, by combining the above formulas (13) and (14), the system can automatically generate a sample set containing 1601 different short-circuit impedance parameters, where each set of parameters This corresponds to a unique short-circuit fault scenario. The sample set was further divided into a training set (1280 groups) and a test set (321 groups) in a 4:1 ratio, providing a comprehensive and controllable data foundation for the subsequent step-by-step training of the generalized regression neural network and the performance verification of transient equivalent modeling.

[0063] Next, in the above embodiments, Table 1 below shows the specific grouping and distribution of the sample set used for training and GRNN. This table systematically lists the number of simulation samples corresponding to different low-voltage generator unit numbers (unit: units) and the range of short-circuit impedance values ​​set to simulate changes in fault location (unit: ... Specifically, the sample set is grouped according to the number of new energy generating units entering the low-voltage ride-through state, ranging from 3 to 7 units, into 5 groups. The number of samples in each group is configured according to the probability of occurrence in typical operating scenarios. For example, the number of samples is largest when there are 5 units (736 samples), while the number of samples is smallest when there are 3 units (105 samples). Meanwhile, the short-circuit impedance range corresponding to each group (in complex form) is also specified. (Indicated) Continuous and non-overlapping, its resistive and reactive components both monotonically increase with increasing electrical distance. For example, when the number of units is 7, the impedance range is... When the number of units is 3, the impedance range is This grouping method ensures that the training data can comprehensively cover typical fault scenarios from near-end to far-end short-circuit locations and different scales of grid disconnection of new energy units, providing a key data foundation for the subsequent regression model to achieve high accuracy and strong generalization ability.

[0064] In other embodiments, Table 2 below shows the error evaluation results of the first, second, and third regression models obtained after stepwise training of the generalized regression neural network based on the training set. The table specifically lists the mean absolute percentage error (MAS) of each model when regressing its target isoparameter. (unit: %) and maximum absolute percentage error ( (Unit: %), quantitatively reflecting the training accuracy and generalization performance of the model. As shown in the table, the first regression model for low-load units ( Regression error and All values ​​are 0%, indicating that the model achieves a completely accurate fit to the discrete number of units. The second regression model adjusts the voltage correction parameters. , regression error They were 0.0006% and 0.0002% respectively. The percentages were 0.0194% and 0.0031% respectively, demonstrating high accuracy and smooth regression capability for continuous parameters. The third regression model applied to equivalent impedance... and regression error They were 0.0055% and 0.0009% respectively. The error rates are 0.1553% and 0.0185%, respectively, indicating that the method also possesses superior regression accuracy for the key impedance parameters characterizing the network structure. This error data table provides direct and quantitative evidence to verify the effectiveness of the proposed step-by-step training method in avoiding parameter coupling and achieving high-precision transient equivalent modeling.

[0065] In other embodiments, Table 3 below shows a comparison and error analysis between the results of short-circuit current calculations using the method described in this invention and the results of detailed electromagnetic transient simulations. This table systematically lists the baseline data (“simulation data”) obtained from the detailed simulation model and the results (“calculated data”) calculated by the transient equivalent model of this embodiment, along with their relative errors, under different scenarios with varying numbers of low-voltage generator units (from 3 to 7). The table specifically includes the active component of the short-circuit current (…). Unit: per unit value, abbreviated as ), reactive power component ( ,unit: ) and total current amplitude ( ,unit: A comparison of the three key indicators is shown in Table 3 below. In all test scenarios, the calculated data and simulation data of the method in this embodiment are extremely close. Taking a scenario with 5 units as an example, the simulation results show... 2.9382 respectively 9.4397 9.8864 The result calculated using the method in this embodiment is 2.9397. 9.4394 9.8866 The corresponding relative errors were 0.053%, 0.002%, and 0.0027%, respectively. The maximum relative error in the other scenarios did not exceed 0.053% (see [reference needed]). (components), all total current amplitudes The relative errors are all below 0.0036%. This error analysis table provides conclusive data to prove that the transient equivalent model based on GRNN stepwise regression constructed in this embodiment can reproduce the short-circuit current characteristics of grids with a high proportion of renewable energy access with extremely high accuracy. It effectively solves the problem of insufficient short-circuit current calculation accuracy caused by the inability of traditional models in the background technology to reflect the transient characteristics of renewable energy units.

[0066] In some embodiments, the step of constructing a transient equivalent mathematical model of a power transmission network includes: The first branch is configured as an ideal voltage source connected in series with an equivalent impedance to characterize the short-circuit current response characteristics of synchronous generators in the transmission network during short-circuit faults.

[0067] Among them, the ideal voltage source is used to provide constant voltage support to simulate the internal potential of the synchronous generator set; the equivalent impedance is used to characterize the internal resistance and network impedance of the synchronous generator set, which together reflect the amplitude and phase characteristics of the short-circuit current.

[0068] Specifically, the system sets the voltage phasor of the ideal voltage source and the resistive component of the equivalent impedance. With reactance component (unit: The branch is constructed using the voltage phasor angle in degrees ( ). The unit is ). For example, in the IEEE 39-node test system, with node 16 as the transmission and distribution boundary node, the voltage phasor of the ideal voltage source of the first branch is set to . The equivalent impedance is ,in It is the imaginary unit (representing the imaginary part of a complex number).

[0069] The second branch is configured as a voltage-controlled current source to characterize the short-circuit current response characteristics of all new energy generating units in the power grid during short-circuit faults. The new energy generating units include those that have entered the low-voltage ride-through state and those that have not.

[0070] Among them, the voltage-controlled current source is used to dynamically adjust the output current according to the voltage at the transmission and distribution boundary node to simulate the nonlinear response of the new energy unit when the voltage drops; the low-voltage ride-through state refers to the situation where the voltage at the grid connection point of the new energy unit is lower than a threshold (e.g., 0.9). When in operation, it is necessary to maintain grid connection and inject reactive current into the system.

[0071] Specifically, the system implements this branch by defining a control function for the voltage-controlled current source, with the control function controlling the transmission and distribution boundary node voltage. (unit: () is the input, and the output is the active current component. and reactive current components (unit: ), and introduce voltage correction parameters. , (Dimensionless) Compensation for voltage measurement error.

[0072] Therefore, according to the above implementation method, the system can accurately construct the transient equivalent model of the power transmission network containing new energy sources, provide a physical consistency basis for short-circuit current calculation, and support the efficient execution of the subsequent GRNN stepwise regression process.

[0073] In some embodiments, the steps of establishing a detailed simulation model of a transmission network containing a high proportion of new energy generating units, selecting transmission and distribution boundary nodes as observation points, generating multiple short-circuit fault scenarios by adjusting the impedance parameters of the distribution network lines, and collecting voltage data of transmission and distribution boundary nodes based on the short-circuit fault scenarios to form a sample set include: The short-circuit impedance is configured as a structure consisting of the step-down transformer impedance and the distribution network line impedance connected in series.

[0074] Among them, the impedance of the step-down transformer refers to the equivalent impedance value of the step-down substation, which is used to characterize the resistance and reactance characteristics of the transformer itself; the impedance of the distribution network line refers to the equivalent impedance value of the line connecting the transmission and distribution boundary node and the short-circuit fault point, which is used to characterize the resistance and reactance characteristics of the line.

[0075] Specifically, the system sets the specific value of the short-circuit impedance using electromagnetic transient simulation software (such as PSCAD or EMTP), where the impedance of the step-down transformer is set to a fixed value (e.g., ,in (The unit is imaginary). The impedance of the distribution network lines is set as a variable parameter. For example, in the IEEE 39-bus test system, with node 16 as the transmission and distribution boundary node, the impedance of the step-down transformer is fixed at [value missing]. The impedance of the distribution network lines is varied by adjusting the line length. PSCAD stands for Power System Computer Aided Design, and EMTP stands for Electromagnetic Transients Program; both are electromagnetic transient simulation software.

[0076] By changing the impedance parameters of the distribution network line sections, the change in electrical distance between the short-circuit fault point and the transmission and distribution boundary node is simulated.

[0077] Among them, electrical distance refers to the strength of electrical connection between the short-circuit fault point and the transmission and distribution boundary node, which is indirectly reflected by the impedance value. The larger the impedance value, the farther the electrical distance.

[0078] Specifically, the system simulates the scenario where the fault point gradually moves away from the transmission and distribution boundary node by linearly increasing the resistive and reactive components of the distribution network line impedance. The impedance variation range covers from the minimum impedance (containing only the step-down transformer impedance) to the maximum impedance (including the long line impedance).

[0079] For example, setting the impedance parameter value to increase in steps: the resistance component from... by Increase the step size to The reactance component from by Increase the step size to A total of 1601 short-circuit fault scenarios were generated, with m ranging from 0 to 1600 (m is an integer).

[0080] Simulation calculations were performed based on short-circuit fault scenarios corresponding to different impedance parameters, and voltage amplitude and phase angle data of transmission and distribution boundary nodes were collected.

[0081] Among them, voltage amplitude refers to the effective value of the voltage at the transmission and distribution boundary node, and the unit is per unit value ( Phase angle data refers to the phase angle of the voltage phasor, in degrees. ).

[0082] Specifically, the system performs three-phase short-circuit fault simulation under each set of short-circuit impedance settings, with the fault occurrence time... Second Lasting 0.1 seconds After disconnection, the voltage amplitude at the transmission and distribution boundary node in steady state is recorded. and phase angle .

[0083] For example, for impedance parameters The scenario (short-circuit impedance is) The voltage amplitude was collected through simulation. Phase angle ;for The scenario (short-circuit impedance is) ), collected , .

[0084] The collected voltage amplitude and phase angle data are combined with the corresponding short-circuit impedance parameters to form a sample set.

[0085] The sample set refers to the data set used for training and testing machine learning models, which includes input features (such as voltage data) and output labels (such as impedance parameters).

[0086] Specifically, the system will assign each set of short-circuit impedance parameters ( and ) and the corresponding voltage data ( and Pairing samples together creates a single sample record. All samples are then randomly divided into a training set (80%) and a test set (20%). For example, a total of 1601 samples are generated, with 1280 used as the training set and 321 as the test set. The data is stored in matrix form, with each row containing... Four fields.

[0087] Therefore, according to the above implementation method, the system can generate a sample set covering different short-circuit fault scenarios, providing a sufficient and diverse data foundation for the subsequent step-by-step training of the generalized regression neural network, and supporting the accuracy of transient equivalent modeling.

[0088] In some embodiments, a generalized regression neural network is trained stepwise using a training set to generate a first regression model, a second regression model, and a third regression model for different equivalent parameters of the regression mathematical model, including: The input to the first regression model is the voltage amplitude at the transmission and distribution boundary node, and the output is the number of low-voltage units.

[0089] Among them, the number of low-voltage ride-through units refers to the number of new energy generating units that enter the low-voltage ride-through state during a short-circuit fault (unit: number), used to characterize the scale of generating units affected by voltage dips in the transmission network; the voltage amplitude at the transmission and distribution boundary node refers to the effective value of the voltage at the transmission and distribution boundary node during a fault (unit: ).

[0090] Specifically, the system receives voltage amplitude data through the input layer of a generalized regression neural network, and the output layer regresses the integer value of the number of low-voltage generator units. During training, a radial basis function is used as the activation function, and the mean absolute percentage error (MASE) is used as the training termination condition. For example, based on the sample set shown in Table 1 above, when the voltage amplitude... At that time, the corresponding number of low-voltage units The training model 1 had an error of 0% on the test set.

[0091] The input to the second regression model is the voltage amplitude at the transmission and distribution boundary node, and the output is the voltage correction parameter.

[0092] Among them, the voltage correction parameters include (Dimensionless) is used to compensate for the measurement error between the voltage at the transmission and distribution boundary node and the voltage at the grid connection point of new energy units; the voltage correction parameter is calculated by weighted average to reflect the voltage correction requirements of different units.

[0093] Specifically, the system inputs the voltage amplitude into the second regression model and outputs two voltage correction parameters. During training, the network output is calculated using the maximum probability principle to ensure the smoothness of parameter regression and generalization ability. For example, regarding the voltage amplitude... The sample was obtained from the second regression model. The actual value is The errors were 0.0009% and 0.0035%, respectively.

[0094] The input to the third regression model is the sine and cosine components of the voltage amplitude at the transmission and distribution boundary nodes, and the output is the equivalent impedance.

[0095] The equivalent impedance includes a resistive component. and reactance component (unit: ), used to characterize the impedance characteristics of synchronous generator branch; sine and cosine components are expressed through voltage amplitude. and phase angle Calculated, i.e. and (unit: This is to standardize the units of measurement and avoid training bias.

[0096] Specifically, the system will and As a dual-channel input, the real and imaginary parts of the output equivalent impedance are processed in steps during training to reduce the influence of parameter coupling. For example, for , The sample, calculate , Model 3 regression yields , The actual value is , The errors are respectively and .

[0097] Based on the training set, the first regression model, the second regression model, and the third regression model were trained independently.

[0098] Independent training refers to training the three models sequentially to avoid mutual interference between equivalent parameters; the ratio of training set to test set is set to 4:1 to ensure the generalization performance of the model.

[0099] Specifically, the system first uses training set data (voltage amplitude) and the corresponding number of low-altitude units Train the first regression model, then use the same... and corresponding , Train the second regression model and finally use and corresponding Training the third regression model; the training process follows... Figure 3 , 4The equivalent parameters are shown in the step-by-step training steps. For example, out of 1601 samples, 1280 samples were used as the training set. After training, the maximum absolute percentage error of each model was less than 0.1553% (see Table 2 above), which meets the engineering accuracy requirements.

[0100] Therefore, according to the above implementation method, the system can accurately regress the equivalent parameters through step-by-step training, providing a reliable basis for short-circuit current calculation and supporting the application of transient equivalent models in high-proportion new energy scenarios.

[0101] In some embodiments, the voltage data of the test set is input into the regression model, and the first regression model is called sequentially to regress the number of low-voltage units, the second regression model is called to regress the voltage correction parameters, and the third regression model is called to regress the equivalent impedance, including: Input the voltage amplitude of the transmission and distribution boundary node in the test set into the first regression model to obtain the regression value of the number of low-voltage units.

[0102] Among them, the number of low-voltage ride-through units refers to the number of new energy generating units that enter the low-voltage ride-through state during a short-circuit fault (unit: number), used to characterize the scale of units affected by voltage dips; the voltage amplitude at the transmission and distribution boundary node refers to the effective value of the voltage at the transmission and distribution boundary node during a fault (unit: ).

[0103] Specifically, the system receives voltage amplitude data through the input layer of a generalized regression neural network, and the output layer regresses the integer value of the number of low-voltage units. The regression process is calculated based on radial basis function activation and the maximum probability principle. For example, when the voltage amplitude in the test set... At that time, the first regression model obtained the number of low-power units. (See the sample group with 5 units in Table 1 above).

[0104] Input the voltage amplitude of the transmission and distribution boundary node in the test set into the second regression model to obtain the regression value of the voltage correction parameter.

[0105] Among them, the voltage correction parameters include , (Dimensionless) is used to compensate for the measurement error between the voltage at the transmission and distribution boundary node and the voltage at the grid connection point of new energy units; the voltage correction parameter is calculated by weighted average to reflect the voltage correction requirements of different units.

[0106] Specifically, the system will measure the voltage amplitude Input the second regression model, and the model outputs two voltage correction parameters. , During regression, the trained GRNN network parameters are used for forward inference. For example, for the test set... The sample was obtained by the second regression model. .

[0107] The sinusoidal and cosine components of the voltage amplitude at the transmission and distribution boundary nodes in the test set are input into the third regression model to obtain the regression value of the equivalent impedance.

[0108] The equivalent impedance includes a resistive component. and reactance component (unit: ), used to characterize the impedance characteristics of synchronous generator branch; sine and cosine components are expressed through voltage amplitude. and phase angle Calculated, i.e. and (unit: This is to standardize the units of measurement and avoid training bias.

[0109] Specifically, the system first calculates the sinusoidal components of the voltage phasor. Sum and cosine components The dual-channel data is input into the third regression model, and the real part of the equivalent impedance is output. and the virtual part For example, for the test set , The sample, calculate , The third regression model yields the following results: , .

[0110] Therefore, according to the above implementation method, the system can accurately obtain the number of low-voltage units, voltage correction parameters and equivalent impedance through stepwise regression, providing accurate input parameters for short-circuit current calculation and supporting engineering applications of transient equivalent modeling.

[0111] In some embodiments, the step of calculating the short-circuit current component of the second branch based on the number of low-voltage units and voltage correction parameters, and calculating the short-circuit current component of the first branch based on the equivalent impedance, and synthesizing the total short-circuit current includes: Based on the number of low-voltage units obtained from the first regression model and the voltage correction parameters obtained from the second regression model, the output current of the voltage-controlled current source of the second branch is calculated.

[0112] The voltage-controlled current source output current includes an active current component. (unit: ) and reactive current components (unit: The voltage is calculated using a voltage-controlled current source control function; the number of low-voltage ride-through units refers to the number of new energy generating units that enter low-voltage ride-through state during a short-circuit fault (unit: units); voltage correction parameters include... , (Dimensionless) is used to compensate for the measurement error between the voltage at the transmission and distribution boundary node and the voltage at the grid connection point of new energy units.

[0113] Specifically, the system calculates through the voltage-controlled current source control function. and Firstly, based on the number of low-voltage units and voltage correction parameters , Calculate the equivalent rated current of the two generator groups. and Combined with weighted average parameters and (Obtained by weighting the capacity of new energy units) and transmission and distribution boundary node voltage Solve using the minimum value function and square root operation. and For example, when the first regression model yields a low number of units... The voltage correction parameters are obtained from the second regression model. And the voltage at the transmission and distribution boundary node At that time, the system calculates the output current of the voltage-controlled current source. , .

[0114] Based on the equivalent impedance and transmission / distribution boundary node voltage obtained from the third regression model, the ideal voltage source output current of the first branch is calculated.

[0115] The output current of an ideal voltage source includes an active current component. (unit: ) and reactive current components (unit: The equivalent impedance is calculated using Ohm's law; it includes the resistive component. and reactance component (unit: The impedance characteristics of the synchronous generator branch are characterized; the transmission and distribution boundary node voltage refers to the voltage phasor of the transmission and distribution boundary node during a fault, including amplitude. and phase angle .

[0116] Specifically, the system calculates using Ohm's law. and First, convert the voltage phasors at the transmission and distribution boundary nodes. With the potential of an ideal voltage source (Default setting) Take the difference to get the voltage drop, then divide by the equivalent impedance. The current component is solved using complex number operations. The calculation process uses the following formula: For example, when the third regression model yields equivalent impedance... , And the voltage at the transmission and distribution boundary node , At that time, the system calculates the output current of the ideal voltage source. , .

[0117] The output current of the second branch is superimposed with the output current of the first branch to form the total short-circuit current.

[0118] Phasor superposition refers to the algebraic addition of the active current component and the reactive current component to obtain the active component of the total short-circuit current. and reactive components (unit: ), total current amplitude pass and It is calculated by taking the square root of the sum of the squares.

[0119] Specifically, the system first converts the second branch... , With the first branch road The total current is obtained by adding the corresponding components. , Then calculate the total current amplitude. This process ensures that the phasor characteristics of the short-circuit current are fully preserved.

[0120] For example, based on the above embodiments , , , The system calculates the total short-circuit current. , , , and the simulation results of the sample ( , The error is less than 0.1%.

[0121] Therefore, according to the above implementation method, the system can accurately synthesize the total short-circuit current through step-by-step calculation and phasor superposition, verifying the effectiveness of the transient equivalent modeling method in short-circuit current calculation.

[0122] In some embodiments, based on the training set, the first regression model, the second regression model, and the third regression model are trained independently, including: The voltage amplitude of the transmission and distribution boundary nodes in the training set is used as input data, and the corresponding number of low-voltage units is used as the output target to train the first regression model.

[0123] Among them, the number of low-voltage ride-through units refers to the number of new energy generating units that enter the low-voltage ride-through state during a short-circuit fault (unit: number), used to characterize the scale of units affected by voltage dips; the voltage amplitude at the transmission and distribution boundary node refers to the effective value of the voltage at the transmission and distribution boundary node during a fault (unit: ).

[0124] Specifically, the system receives voltage amplitude data through the input layer of the GRNN, and the output layer regresses the integer value of the number of low-voltage units. During the training process, the radial basis function is used as the activation function, and the mean absolute percentage error is used as the training error index for monitoring.

[0125] For example, based on the sample set shown in Table 1 above, when the voltage amplitude in the training set... At that time, the corresponding number of low-voltage units After training, the mean absolute percentage error of Model 1 on the test set was 0% (see Table 2 above).

[0126] The voltage amplitudes of the transmission and distribution boundary nodes in the training set are used as input data, and the corresponding voltage correction parameters are used as output targets to train the second regression model.

[0127] Among them, the voltage correction parameters include , (Dimensionless) is used to compensate for the measurement error between the voltage at the transmission and distribution boundary node and the voltage at the grid connection point of new energy units; the voltage correction parameter is calculated by weighted average to reflect the voltage correction requirements of different units.

[0128] Specifically, the system inputs the voltage amplitude U into the second regression model and outputs two voltage correction parameters. , During training, the network output is calculated using the maximum probability principle to ensure the smoothness of parameter regression. For example, for the training set... The sample was obtained from the second regression model. The mean absolute percentage errors after training were 0.0009% and 0.0035%, respectively (see Table 2 above).

[0129] The third regression model is trained by using the sinusoidal and cosine components of the voltage amplitude at the transmission and distribution boundary nodes in the training set as input data and the corresponding equivalent impedance as the output target.

[0130] The equivalent impedance includes a resistive component. and reactance component (unit: ), used to characterize the impedance characteristics of synchronous generator branch; sine and cosine components are expressed through voltage amplitude. and phase angle Calculated, i.e. and (unit: This is to standardize the units of measurement and avoid training bias.

[0131] Specifically, the system will and As a dual-channel input, the real part of the output equivalent impedance. and the virtual part During training, parameter coupling is reduced by processing parameters in stages. For example, for the training set... , The sample, calculate , Model 3 regression yields , The mean absolute percentage errors after training were 0.0010% and 0.0016%, respectively (see Table 2 above).

[0132] The training process for each regression model is conducted independently. When the training error index of the model meets the preset termination condition, its training is terminated. After all regression models have been trained, the training of the generalized regression neural network is considered complete. The preset termination condition refers to the training error threshold set in advance for the regression model.

[0133] Specifically, the training error metric is the mean absolute percentage error (in %), and the preset termination condition is set to terminate training when the mean absolute percentage error is lower than 0.01%. The training process is carried out sequentially: first, the training of model one is completed, then the training of model two is started, and finally the training of model three is carried out to ensure parameter decoupling.

[0134] For example, out of 1601 samples, 1280 were used as the training set. The training error of Model 1 was 0%, the training error of Model 2 was 0.0006%, and the training error of Model 3 was 0.0055%, all of which met the termination condition (see Table 2 above). After training was completed, the system entered the testing phase.

[0135] Therefore, according to the above implementation method, the system can provide a reliable basis for short-circuit current calculation by training the accurate regression equivalent parameters step by step independently, and support the application of transient equivalent models in high-proportion new energy scenarios.

[0136] Figure 6 This is a structural block diagram of a power transmission network transient equivalent modeling system based on GRNN stepwise regression according to an embodiment of the present invention.

[0137] like Figure 6 As shown, the power grid transient equivalent modeling system based on GRNN stepwise regression includes: The mathematical model construction module 210 is used to construct a transient equivalent mathematical model of the power transmission network. The mathematical model includes a first branch representing the short-circuit current response characteristics of synchronous generator units and a second branch representing the short-circuit current response characteristics of new energy generator units.

[0138] The simulation model construction module 220 is used to establish a detailed simulation model of the power transmission network containing a high proportion of new energy units. It selects the transmission and distribution boundary nodes as observation points, generates multiple short-circuit fault scenarios by adjusting the impedance parameters of the distribution network lines, collects voltage data of the transmission and distribution boundary nodes based on the short-circuit fault scenarios to form a sample set, and divides the sample set into a training set and a test set.

[0139] The GRNN step-by-step training module 230 is used to train the generalized regression neural network step-by-step using the training set, generating a first regression model, a second regression model, and a third regression model with different equivalent parameters for the regression mathematical model.

[0140] The transient equivalent modeling result output module 240 is used to input the voltage data of the test set into the regression model when the generalized regression neural network has been trained. It sequentially calls the first regression model to regress the number of low-voltage units, the second regression model to regress the voltage correction parameters, and the third regression model to regress the equivalent impedance. Based on the number of low-voltage units and the voltage correction parameters, it calculates the short-circuit current component of the second branch and the short-circuit current component of the first branch based on the equivalent impedance, and synthesizes the total short-circuit current.

[0141] The specific functions and examples of each module and submodule of the device in this embodiment can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0142] According to embodiments of the present invention, the above-described method of the present invention can be applied to an electronic device and a readable storage medium.

[0143] Figure 7 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0144] like Figure 7As shown, the electronic device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. The RAM 603 may also store various programs and data required for the operation of the electronic device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0145] Multiple components in electronic device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of displays, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows electronic device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0146] The computing unit 601 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as a transmission network transient equivalent modeling method based on GRNN stepwise regression. For example, in some embodiments, a transmission network transient equivalent modeling method based on GRNN stepwise regression can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the transmission network transient equivalent modeling method based on GRNN stepwise regression described above can be performed. Alternatively, in other embodiments, the computing unit 601 may be configured, by any other suitable means (e.g., by means of firmware), to perform a GRNN-based stepwise regression method for power grid transient equivalent modeling.

[0147] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0148] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0149] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0150] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0151] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0152] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0153] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0154] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for transient equivalent modeling of power transmission networks based on GRNN stepwise regression, characterized in that, include: A transient equivalent mathematical model of the power transmission network is constructed. The mathematical model includes a first branch representing the short-circuit current response characteristics of synchronous generator units and a second branch representing the short-circuit current response characteristics of new energy generator units. A detailed simulation model of a power transmission network with a high proportion of new energy generating units is established. Transmission and distribution boundary nodes are selected as observation points. Multiple short-circuit fault scenarios are generated by adjusting the impedance parameters of the distribution network lines. Voltage data of transmission and distribution boundary nodes are collected based on the short-circuit fault scenarios to form a sample set. The sample set is then divided into a training set and a test set. The training set is used to train the generalized regression neural network GRNN step by step to generate a first regression model, a second regression model and a third regression model for regressing different equivalent parameters of the mathematical model. In response to the completion of training of the generalized regression neural network, the voltage data of the test set is input into the regression model, and the first regression model is called to regress the number of low-voltage units, the second regression model is called to regress the voltage correction parameters, and the third regression model is called to regress the equivalent impedance. The short-circuit current component of the second branch is calculated based on the number of low-voltage units and the voltage correction parameters, and the short-circuit current component of the first branch is calculated based on the equivalent impedance, and the total short-circuit current is synthesized.

2. The method according to claim 1, characterized in that, The steps for constructing the transient equivalent mathematical model of the power transmission network include: The first branch is configured as a structure in series between an ideal voltage source and an equivalent impedance to characterize the short-circuit current response characteristics of synchronous generators in the power transmission network during short-circuit faults. The second branch is configured as a voltage-controlled current source to characterize the short-circuit current response characteristics of all new energy units in the power grid during short-circuit faults. The new energy units include units that have entered the low-voltage ride-through state and units that have not entered the low-voltage ride-through state.

3. The method according to claim 1, characterized in that, The steps of establishing a detailed simulation model of the transmission network containing a high proportion of new energy generating units, selecting transmission and distribution boundary nodes as observation points, generating multiple short-circuit fault scenarios by adjusting the impedance parameters of the distribution network lines, and collecting voltage data of the transmission and distribution boundary nodes based on the short-circuit fault scenarios to form a sample set include: The short-circuit impedance is configured as a structure consisting of the step-down transformer impedance and the distribution network line impedance connected in series; the short-circuit impedance refers to the equivalent impedance consisting of the step-down transformer impedance and the distribution network line impedance connected in series. By changing the impedance parameter values ​​of the distribution network line section, the change in electrical distance between the short-circuit fault point and the transmission and distribution boundary node is simulated. Simulation calculations are performed based on the short-circuit fault scenarios corresponding to different impedance parameters, and the voltage amplitude and phase angle data of the transmission and distribution boundary nodes are collected. The collected voltage amplitude and phase angle data are combined with the corresponding short-circuit impedance parameters to form the sample set.

4. The method according to claim 1, characterized in that, The step-by-step training of the generalized regression neural network using the training set to generate a first regression model, a second regression model, and a third regression model for regressing different equivalent parameters of the mathematical model includes: The input to the first regression model is the voltage amplitude at the transmission and distribution boundary node, and the output is the number of low-voltage units. The input to the second regression model is the voltage amplitude of the transmission and distribution boundary node, and the output is the voltage correction parameter; The input to the third regression model is configured to be the sine and cosine components of the voltage amplitude at the transmission and distribution boundary node, and the output is the equivalent impedance. Based on the training set, the first regression model, the second regression model, and the third regression model are trained independently, respectively.

5. The method according to claim 1, characterized in that, The step of inputting the voltage data of the test set into the regression model, sequentially calling the first regression model to regress the number of low-voltage units, calling the second regression model to regress the voltage correction parameters, and calling the third regression model to regress the equivalent impedance includes: Input the voltage amplitude of the transmission and distribution boundary node in the test set into the first regression model to obtain the regression value of the number of low-voltage units; Input the voltage amplitude of the transmission and distribution boundary node in the test set into the second regression model to obtain the regression value of the voltage correction parameter; The sinusoidal and cosine components of the voltage amplitude at the transmission and distribution boundary nodes in the test set are input into the third regression model to obtain the regression value of the equivalent impedance.

6. The method according to claim 1, characterized in that, The step of calculating the short-circuit current component of the second branch based on the number of low-voltage units and the voltage correction parameters, and calculating the short-circuit current component of the first branch based on the equivalent impedance, and synthesizing the total short-circuit current, includes: Based on the number of low-voltage units obtained from the first regression model and the voltage correction parameters obtained from the second regression model, the output current of the voltage-controlled current source of the second branch is calculated. Based on the equivalent impedance obtained from the third regression model and the voltage of the transmission and distribution boundary node, the ideal voltage source output current of the first branch is calculated. The output current of the second branch is phasor-superimposed with the output current of the first branch to synthesize the total short-circuit current.

7. The method according to claim 4, characterized in that, The step of independently training the first regression model, the second regression model, and the third regression model based on the training set includes: The voltage amplitude of the transmission and distribution boundary nodes in the training set is used as input data, and the corresponding number of low-voltage units is used as the output target to train the first regression model. The voltage amplitude of the transmission and distribution boundary nodes in the training set is used as input data, and the corresponding voltage correction parameters are used as output targets to train the second regression model. The third regression model is trained by using the sinusoidal and cosine components of the voltage amplitude of the transmission and distribution boundary nodes in the training set as input data and the corresponding equivalent impedance as the output target. The training process for each regression model is carried out independently. When the training error index of the model meets the preset termination condition, the training is terminated. After all regression models have been trained, the training of the generalized regression neural network is considered complete. The preset termination condition refers to the training error threshold set in advance for the regression model.

8. A transient equivalent modeling system for power transmission networks based on GRNN stepwise regression, characterized in that, include: The mathematical model construction module is used to construct a transient equivalent mathematical model of the power transmission network. The mathematical model includes a first branch representing the short-circuit current response characteristics of synchronous generator units and a second branch representing the short-circuit current response characteristics of new energy generator units. The simulation model construction module is used to establish a detailed simulation model of the power grid containing a high proportion of new energy units. It selects the transmission and distribution boundary nodes as observation points, generates multiple short-circuit fault scenarios by adjusting the impedance parameters of the distribution network lines, collects voltage data of the transmission and distribution boundary nodes based on the short-circuit fault scenarios to form a sample set, and divides the sample set into a training set and a test set. The GRNN step-by-step training module is used to train the generalized regression neural network step-by-step using the training set, and to generate a first regression model, a second regression model and a third regression model for regressing different equivalent parameters of the mathematical model. The transient equivalent modeling result output module is used to input the voltage data of the test set into the regression model when the generalized regression neural network has been trained, and sequentially call the first regression model to regress the number of low-voltage units, call the second regression model to regress the voltage correction parameters, and call the third regression model to regress the equivalent impedance. The short-circuit current component of the second branch is calculated based on the number of low-voltage units and the voltage correction parameters, and the short-circuit current component of the first branch is calculated based on the equivalent impedance, and the total short-circuit current is synthesized.

9. An electronic device, characterized in that, include: At least one processor; and a memory that is communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, in, Computer instructions are used to cause a computer to perform the method according to any one of claims 1-7.

Images