Short-circuit current increment prediction analysis method and system in high-proportion new energy power grid

Abstract

The present application belongs to the technical field of power grid short-circuit current analysis and prediction, in order to solve the problem of inaccurate and poor reliability of power grid short-circuit current increment prediction under the background of high proportion of new energy, a short-circuit current increment prediction analysis method and system in high proportion of new energy power grid are proposed, based on the principle of power grid short-circuit calculation, the mechanism of various influencing factors on short-circuit current is analyzed, the key parameter characteristics of each area of the power grid are extracted, and the support vector machine model is trained combined with the corresponding target node short-circuit current; in the training, the physical theory short-circuit current increment sensitivity boundary calculated based on the new energy control strategy amplitude limiting feature is taken as the physical constraint condition, the root mean square of the short-circuit current prediction value and the sample true value is combined to construct the comprehensive fitness function of the sparrow search algorithm, and the penalty factor and the kernel function width of the support vector machine are optimized; the present application realizes accurate and reliable prediction of power grid short-circuit current increment under the background of high proportion of new energy.

Description

Technical Field

[0001] This invention belongs to the technical field of short-circuit current analysis and prediction in power grids, and particularly relates to a method and system for predicting and analyzing short-circuit current increments in high-proportion renewable energy power grids. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] The transmission of new energy sources via ultra-high-voltage AC transmission has exacerbated the complexity of short-circuit current control, and excessive short-circuit current has become a constraint on safe and stable operation. Therefore, with the continuous advancement of the planning and construction of new power systems, it is an urgent need for power grid construction planning to effectively predict the increase in short-circuit current in high-proportion new energy power grids before construction, and to understand the impact of the expansion of different types of power sources in various regions on the short-circuit current of the power grid.

[0004] Short-circuit current is directly related to the power source type and control strategy of the power grid. However, in existing research, the simplification of traditional short-circuit current calculation models for renewable energy units leads to insufficient accuracy and large calculation errors, making it difficult to accurately assess the short-circuit risk under conditions of high renewable energy share. Meanwhile, models applicable to different renewable energy sources for short-circuit current calculation are still imperfect, with insufficient calculation accuracy, making it difficult to effectively assess the impact of various renewable energy construction projects on the power grid's short-circuit current. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, this invention provides a method and system for predicting and analyzing short-circuit current increments in high-proportion renewable energy power grids. This method enables reliable prediction of short-circuit current increments in power grids under the background of high renewable energy, and can quantify the contribution of different power sources in each region to the short-circuit current increments of target nodes.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for predicting and analyzing short-circuit current increments in high-proportion renewable energy power grids, including: The high-proportion renewable energy power grid is partitioned based on the electrical distance between nodes; Collect historical operation data of high-proportion renewable energy power grids, analyze the mechanism of various influencing factors on short-circuit current based on the power grid short-circuit calculation principle, and extract the key parameter characteristics of each region in high-proportion renewable energy power grids that affect short-circuit current. Based on the key parameter characteristics affecting short-circuit current in various regions of the high-proportion renewable energy power grid, and the corresponding short-circuit current of each target node, the support vector machine model is trained. In the training of the support vector machine model, the physical theoretical short-circuit current increment sensitivity boundary calculated based on the limiting characteristics of renewable energy control strategy is used as the physical constraint condition. Combining the root mean square of the predicted short-circuit current value and the sample true value, the comprehensive fitness function of the sparrow search algorithm is constructed. The sparrow search algorithm is used to optimize the penalty factor and kernel function width of the support vector machine model. Using a trained support vector machine model, we predict and analyze the sensitivity of short-circuit current at target nodes of the current high-proportion renewable energy power grid to the incremental power supply of various types in different regions.

[0007] Secondly, the present invention provides a short-circuit current increment prediction and analysis system for high-proportion renewable energy power grids, comprising: The partitioning module is configured to partition the high-proportion renewable energy power grid based on the electrical distance between nodes. The extraction module is configured to: collect historical operation data of high-proportion renewable energy power grids, analyze the mechanism of various influencing factors on short-circuit current based on the power grid short-circuit calculation principle, and extract key parameter characteristics of each region in high-proportion renewable energy power grids that affect short-circuit current; The training module is configured to: train the support vector machine model based on the key parameter characteristics affecting the short-circuit current in each area of ​​the high-proportion renewable energy power grid, and the corresponding short-circuit current of each target node; in the training of the support vector machine model, the physical theoretical short-circuit current increment sensitivity boundary calculated based on the limiting characteristics of renewable energy control strategy is used as the physical constraint condition, and the comprehensive fitness function of the sparrow search algorithm is constructed by combining the root mean square of the predicted short-circuit current value and the sample true value, and the sparrow search algorithm is used to optimize the penalty factor and kernel function width of the support vector machine model; The prediction module is configured to use a trained support vector machine model to predict and analyze the sensitivity of the short-circuit current of the target node in the current high-proportion renewable energy grid to the increment of various power sources in different regions.

[0008] Preferably, the high-proportion renewable energy power grid is partitioned based on the electrical distance between nodes, specifically as follows: Calculate the electrical distance between nodes based on the Thevenin equivalent impedance between each node. Based on the electrical distance matrix between nodes, the K-Means clustering algorithm is used to minimize the sum of the squared errors of the electrical distance feature vectors of nodes in each partition to the cluster center. The node cluster allocation and cluster center update are performed iteratively until the objective function converges, resulting in a preset number of partitions.

[0009] Preferably, based on the superposition principle and Thevenin's equivalence theorem, the mathematical mapping relationship between the type of new energy generating unit, control strategy, access voltage level, and capacity and short-circuit current is analyzed to determine the initial feature library; grey relational analysis is performed on the data in the initial feature library and the short-circuit current data to select the grey relational degree. Initially selected features exceeding a set threshold are used as key parameters affecting short-circuit current. First-order difference and standardization preprocessing are performed on the key parameters and the corresponding short-circuit currents of each target node. The preprocessed data is then used to train the support vector machine model.

[0010] Preferably, the sparrow search algorithm is used to optimize the penalty factor and kernel width of the support vector machine model. Specifically, the position of each sparrow individual is encoded with a penalty factor and kernel width. The sparrow population position is iteratively updated by simulating the behavior of the discoverer, joiner, and watcher. In each iteration, the updated position of each sparrow individual is decoded to obtain a new penalty factor and kernel width. The comprehensive fitness is calculated based on the new penalty factor and kernel width. The global optimal position is updated according to the comprehensive fitness value until the iteration ends, and the optimal penalty factor and kernel width are obtained.

[0011] Preferably, the overall fitness is:

[0012]

[0013]

[0014] in, Regarding the penalty factor and kernel width The root mean square error; It is a one-sided upper bound penalty for physical constraints; A very small bias constant; Indicates the number of samples in the training set; The short-circuit current increment predicted by the support vector machine model; This represents the actual short-circuit current increment. This represents the region predicted by the support vector machine model. node The increase in short-circuit current; This is the physical penalty weighting coefficient. The number of various types of new energy power stations after the power grid is divided; This represents the physical theory of short-circuit current increment sensitivity boundary.

[0015] Preferably, the formula for calculating the physical theory short-circuit current increment sensitivity boundary is:

[0016] in, This represents the physical theory of the sensitivity boundary for short-circuit current increment; This represents the maximum short-circuit current output limiting coefficient of the new energy inverter; Indicates the power access node Rated line voltage at the grid connection point of the new energy generating unit; Let be the self-impedance of the target fault node i in the node impedance matrix. is the mutual impedance between the target fault node i and the power supply access node e in the node impedance matrix; This indicates the increase in installed capacity.

[0017] Thirdly, the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0018] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in the first aspect.

[0019] Fifthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the method described in the first aspect.

[0020] The above one or more technical solutions have the following beneficial effects: This invention employs a sparrow search algorithm to optimize the penalty factor and kernel width of the support vector machine, resulting in stronger stability and generalization performance when solving small-sample problems such as short-circuit faults in large power grids. It uses the physical theoretical short-circuit current increment sensitivity boundary calculated based on the limiting characteristics of new energy control strategies as a physical constraint. A comprehensive fitness function integrating physical information is constructed by combining the root mean square of the predicted short-circuit current value and the actual sample value. This allows the physical constraints to directly influence the parameter optimization process of the sparrow search algorithm, effectively improving the robustness and physical reliability of the model in complex evolution scenarios of high-proportion new energy power grids.

[0021] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0023] Figure 1 This is a flowchart of the support vector machine training process in an embodiment of the present invention; Figure 2 This is a schematic diagram of power grid partitioning and short-circuit current target node selection in an embodiment of the present invention; Figure 3 These are the incremental sensitivity heatmap analysis results for different regions in this embodiment of the invention; where (a) is the incremental sensitivity heatmap analysis result for region 1, (b) is the incremental sensitivity heatmap analysis result for region 2, (c) is the incremental sensitivity heatmap analysis result for region 3, and (d) is the incremental sensitivity heatmap analysis result for region 4. Figure 4 This is a graph showing the contribution of different power sources to the short-circuit current increment in an embodiment of the present invention; Figure 5 This is a comparison chart of the predicted and actual short-circuit current values ​​in an embodiment of the present invention. Detailed Implementation

[0024] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0026] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0027] Example 1 SVM (Structural Risk Minimization) is a supervised learning method based on the principle of structural risk minimization. It maps low-dimensional nonlinear problems to a high-dimensional space to find the optimal hyperplane, demonstrating significant advantages in solving regression prediction problems. Currently, power system short-circuit current increment prediction often employs traditional analytical calculation methods or neural network methods. However, in scenarios with a high proportion of renewable energy grid connection, the dimensionality of input features increases significantly due to the nonlinear control of power electronic equipment and multi-source fluctuations. Traditional neural network methods are prone to problems such as the curse of dimensionality, overfitting, and getting trapped in local extrema when dealing with small sample data. Compared to neural networks, SVM can better solve prediction problems under high-dimensional features, especially demonstrating superior stability and generalization performance in handling small sample learning and nonlinear regression. Furthermore, the introduction of the SSA (Structured Safer Algorithm) algorithm simulates a global optimization mechanism of population foraging, adaptively locking the optimal parameter combination of the model, effectively solving the blindness and randomness of SVM parameter selection. This enables effective and reliable prediction and analysis of short-circuit current increments for various power sources in different regions under the background of high renewable energy.

[0028] In this embodiment, the high-proportion renewable energy power grid is first partitioned based on the electrical distance between nodes. Secondly, power parameters and short-circuit current data within each partition are collected as samples. Input features include the installed capacity of various power sources, control strategies, and network topology access patterns of renewable energy sources, while the short-circuit current of each target node is used as the output feature. The data is preprocessed and divided into training and testing sets. An SVM prediction model is constructed, and the penalty factor and kernel width parameters of the SVM algorithm are optimized using the Sparrow Search Algorithm (SSA). A comprehensive fitness function for Physical Information Fusion (PIML) is created, and the optimal SVM prediction model is obtained through iterative updates of the positions of discoverers, joiners, and watchers. Finally, the model is used for time-series reconstruction and power incremental perturbation is applied. The sensitivity of the target node's short-circuit current to various power increments in different regions is extracted. The model is then used to predict and analyze the short-circuit current of the test set to verify the model's accuracy.

[0029] It should be noted that the "high-proportion renewable energy grid" mentioned in this embodiment refers to a power system in which the amplitude, decay time constant, and transient harmonic characteristics of the grid short-circuit current are significantly constrained by the converter control logic due to the large-scale integration of renewable energy generation such as wind power and photovoltaics. "High proportion" is a qualitative description characterizing the system's fault response characteristics, rather than a strict quantitative limit on a specific percentage of installed capacity (for example, this proportion can be 30%, 50%, or higher under different grid structures). The core criterion is whether the existing short-circuit current analysis algorithm based on synchronous generators produces prediction errors in the system that exceed the engineering allowable range.

[0030] like Figure 1 As shown, a method for predicting and analyzing short-circuit current increments in a high-proportion renewable energy power grid includes the following steps: S1. Based on the electrical impedance distance between nodes in a high-proportion renewable energy power grid, the K-Means algorithm is used to partition the power grid.

[0031] First, calculate the electrical distance between each node. With nodes The strength of the electrical connection between two points is determined by the Thevenin equivalent impedance between them. Define the node. With nodes electrical distance between The impedance magnitude between the two nodes:

[0032] In the formula, and They are nodes and Self-impedance, This is the transfer impedance.

[0033] Based on electrical distance characteristics, the K-Means spatial clustering algorithm is used to partition high-proportion renewable energy power grids. The electrical distance matrix is ​​then used to... Using the row vectors in the dataset as input, the K-Means clustering algorithm is applied to divide all nodes in the network into preset groups. Each region is optimized by minimizing the electrical distance feature vector between nodes within each region. To the cluster center The sum of squared errors :

[0034] In the formula, To be classified as the A set of nodes for each major region; This indicates the number of pre-defined regions for all nodes in the network.

[0035] By iteratively performing node cluster allocation and cluster center updates, until the objective function is achieved... Convergence. The result obtained at this point... The partitioning ensures that the electrical connections between nodes within the same region are the closest, while the electrical connections between different regions are relatively sparse, thereby achieving a reduction in the physical spatial dimension of a high-proportion renewable energy power grid.

[0036] Number of partitions H The determination of the elbow rule should take into account the following three methods: (1) Elbow rule: draw different H Value J (1) Select the curve inflection point; (2) Profile coefficient method: Calculate the curve inflection point of different curves. H The mean of the contour coefficients of each node is selected, and the value that maximizes the mean is chosen. H (3) Engineering experience method: Determine based on the actual zoning habits of power grid dispatching and operation (such as zoning by province, city or load center). H This embodiment selects [the appropriate node] based on the above methods and combined with the actual network structure characteristics of the WECC 179-node system. H =4.

[0037] S2. Collect historical operation data of high-proportion new energy power grids, analyze the mechanism of various influencing factors on short-circuit current based on the power grid short-circuit calculation principle, and determine the initial feature library of support vector machine model.

[0038] This step aims to identify key variables that have a fundamental impact on short-circuit current from power grid operation data based on physical mechanisms.

[0039] Historical operational data of the target high-proportion renewable energy power grid area was collected. To determine the initial feature library, this embodiment analyzes the impact of renewable energy unit type, control strategy, access voltage level, and capacity on short-circuit current based on the superposition principle and Thevenin's equivalence theorem. The mathematical mapping relationship is assumed in this embodiment. It is assumed that the adjustment of the local grid structure in a large power grid in the short term will not have a significant impact on the physical mapping law between characteristic factors and short-circuit current. Therefore, changes in the power grid structure are not considered for the time being.

[0040] (1) Analysis of the impact of unit capacity and power supply access voltage level.

[0041] According to the superposition theorem, the short-circuit current amplitude at a certain node in a high-proportion renewable energy power grid is... It can be approximated as the superposition of the system-side contribution and the new energy-side contribution, and its calculation formula is as follows:

[0042] In the formula, Indicates the voltage coefficient; Indicates the rated voltage level of the power grid system; This represents the equivalent impedance modulus of the system side referred to the fault point when no new energy generating units are present. This indicates the total number of new energy power stations connected to the power grid; Indicates the first The rated installed capacity of each new energy power station; Indicates the first The access voltage level of each new energy power station grid connection point; Indicates the first The dimensionless composite transfer coefficient from a new energy power station to a fault point includes equipment limiting characteristics and network impedance transfer characteristics. Among them, the new energy access level is one of the key factors affecting this coefficient.

[0043] Dimensionless composite transport coefficient λ i The specific calculation formula is as follows:

[0044] In the formula, For the first i The maximum short-circuit current output limiting factor for the converter in a new energy power station is typically 1.1 to 1.2 times the rated current; |Z fi | For the fault node f With the i The mutual impedance modulus of each new energy power station access node; |Z ii |for the first i The self-impedance modulus of each new energy power station access node.

[0045] Dimensionless composite transport coefficient λ i This comprehensively reflects the hardware limiting characteristics and network impedance transfer characteristics of power electronic equipment.

[0046] It should be noted that the simplified model based on the superposition principle above applies to the following conditions: (1) During the fault, all new energy power stations operate in the same segment of the LVRT control strategy, that is, the per-unit value of the grid connection voltage of the new energy units. U t (1) The number of new energy power stations located in the same interval; (2) The number of new energy power stations with a single access point. M ≤ 20, when this number is exceeded, the phase-locked loop coupling oscillation between multiple stations will introduce nonlinear errors; (3) the fault type is a symmetrical three-phase short circuit, and the negative sequence component under asymmetrical fault needs to be modeled separately. When the above conditions are not met, the dimensionless composite transfer coefficient λ i The calculation results are only used as a reference for the construction of the initial feature library. The effectiveness of the final features is guaranteed by secondary verification through grey relational analysis (e.g., >0.9).

[0047] As can be seen from the formula, the rated installed capacity With short-circuit current amplitude Positive correlation, voltage level With impedance weight These factors together determine the degree of current collection, and the grid connection voltage level of new energy sources is generally related to their grid connection mode. Therefore, it is essential to establish installed capacity and grid connection mode as key input characteristics.

[0048] (2) Nonlinear coupling analysis of new energy unit type and control strategy.

[0049] Currently, new energy sources mainly consist of wind turbines, photovoltaics, and energy storage. Among them, doubly-fed induction wind turbines, because their stators are directly connected to the grid, exhibit high-amplitude inrush current characteristics in the initial stage of fault transients; while direct-drive wind turbines, photovoltaics, and energy storage are connected to the grid through full-power converters, and due to the thermal stability limitations of power electronic devices, they exhibit controlled current sources with amplitude limiting characteristics.

[0050] Unlike traditional generating units, the short-circuit output of renewable energy generating units is constrained by power electronic control strategies, exhibiting strong nonlinearity. During fault periods, the short-circuit current output of renewable energy generating units is constrained by low-voltage ride-through (LVRT) control strategies.

[0051] Its reactive current The injection logic is shown in the following formula:

[0052] In the formula, This represents the reactive current component injected into the grid by the converter during a fault. The per-unit value representing the voltage at the grid connection point of the new energy generating unit; and These are the low-voltage ride-through threshold and the deep drop threshold, respectively. This represents the reactive current support coefficient, which indicates the sensitivity of the control strategy. Indicates the rated current of the new energy unit; This indicates the maximum reactive current that the converter is allowed to output.

[0053] To protect power electronic devices, the total output current of the converter is limited to the maximum allowable current. limit:

[0054]

[0055] In the formula, This represents the active current component output by the converter during a fault. This indicates the maximum permissible output current (i.e., the limiting value) of the converter, which is typically 1.1 to 1.2 times the rated current. This represents the total effective value of the short-circuit current contributed by a single new energy generating unit; This represents the reactive current component injected into the grid by the converter during a fault.

[0056] Based on the above analysis and the current construction and planning status of new energy sources in the power grid, this embodiment mainly distinguishes three types of new energy sources (wind turbines, photovoltaics, and energy storage) and their corresponding main control strategies: The short-circuit current characteristics of wind turbines are closely related to their generator type, control strategy, and protection measures. They are mainly divided into two categories: doubly-fed asynchronous wind turbines and permanent magnet direct-drive wind turbines, and their short-circuit current characteristics are fundamentally different.

[0057] Photovoltaics are mainly divided into two categories. Constant power control is mainly applicable to the normal state of the power grid, and its short-circuit current characteristics are affected by the limiting link. When a fault occurs in the power grid and causes a voltage drop, the inverter switches to low voltage ride-through control. Its core is to follow the reactive power priority principle, and the amplitude of the short-circuit current is mainly determined by the voltage drop depth.

[0058] Energy storage can be mainly divided into grid-connected energy storage and grid-connected energy storage. Grid-connected energy storage is a current source synchronized with a phase-locked loop, with strictly limited short-circuit current (1.2~1.5 pu), contributing little to the system's short-circuit capacity. Grid-connected energy storage is a voltage source with self-synchronized power, which can provide 2~3 pu of short-circuit current and exhibit inertial response during faults, significantly improving the system's short-circuit ratio and grid strength.

[0059] Based on the above analysis, this embodiment selects a preliminary feature library, which includes the installed capacity of traditional power plants in different power grid areas, the grid connection mode of new energy sources, and the installed capacity of various new energy sources under different control strategies.

[0060] S3. Perform grey relational analysis on the data in the initial feature library and the short-circuit current data, and select the parameters with high correlation as the input features of the support vector machine model.

[0061] First, the selected original feature sequence After averaging, we get To eliminate dimensional differences and preserve fluctuation trend characteristics, the formula is:

[0062] In the formula, Indicates the first The feature in the first The original values ​​of each sample point; Indicates the first The arithmetic mean of a series of features; This represents the dimensionless numerical value after processing. Indicates the number of samples.

[0063] The formula for calculating the correlation coefficient is:

[0064] In the formula, Indicates that two sequences are in the first position. The absolute difference between the points; Represents the dimensionless numerical value of the reference sequence (short-circuit current); This indicates taking the minimum difference between the two levels; This indicates taking the maximum difference between the two levels; Represents the resolution coefficient, which is taken in this embodiment. =0.5.

[0065] Grey relational degree calculation formula:

[0066] In the formula, This represents the final correlation score. The closer the value is to 1, the more consistent the changing trends of the feature and the target are, and the higher the coupling degree. This indicates the number of samples. In this example, [the sample size is selected]. Features that are greater than a set threshold, such as 0.9, are used as input features for the model.

[0067] S4. Perform first-order difference calculation on the original feature data to eliminate the trend term of non-stationary time series data, and divide the processed dataset into training dataset and test dataset, and perform standardization preprocessing on the training dataset.

[0068] To eliminate trend terms in non-stationary time series data and highlight evolutionary increment features, this embodiment performs first-order difference dimensionality reduction on the original power grid feature sequence and short-circuit current sequence to extract the true physical evolutionary increment matrix. and Subsequently, the processed samples were divided into training and testing sets.

[0069] The first-order difference formula is as follows:

[0070] In the formula, Indicates the first The values ​​of each sample; Indicates the first The values ​​of each sample; Indicates the first difference after the finite element. The values ​​of each sample.

[0071] To avoid the failure of kernel function distance calculation due to the difference in capacity between multiple source devices, this embodiment uses standardization to perform dimensionless processing on the incremental features of the training set. The standardization formula is as follows:

[0072] In the formula, Represents the original sample data; This represents the average value of the characteristic. This represents the standard deviation of the feature column; This represents the standardized data.

[0073] S5. Initialize the parameters of the SSA algorithm, including the sparrow population size. The search space size and sparrow population location are used to determine the penalty factor for the SVM model by decoding the population location. and kernel width .

[0074] In the SSA algorithm, it is assumed that the search space size (dimension) is... The number of sparrows is Only, then the location of individual sparrows in the population. As shown below, the position of each individual sparrow encodes the penalty factor and kernel width:

[0075] In the formula, Location of an individual sparrow in a population. For the first Only sparrows The position of the dimension =1,2, , ; =1,2, , . The population size is set to 50 in this embodiment. In this embodiment, the dimension is the problem dimension. =2, including the penalty factor and kernel parameters Two dimensions.

[0076] S6. Establish an SSA-SVM model based on Physical Information Fusion (PIML), incorporate the control strategy limiting characteristics of new energy sources as physical constraints, and combine the root mean square error of the short-circuit current prediction value generated by SVM with the inverse of the sample true value as the comprehensive fitness function of the SSA optimization algorithm.

[0077] Physical constraints are input into the fitness calculation module using the formula The calculated physical boundary values ​​are used as the comprehensive fitness function. The input parameters of the item; this interaction is performed in each round of SSA iteration, thereby enabling physical information to guide the parameter optimization process in real time.

[0078] In the physical information fusion model constructed in this embodiment, the comprehensive fitness function of the SSA-SVM algorithm... as follows:

[0079] In the formula, For SVM about and The root mean square error; It is a one-sided upper bound penalty for physical constraints; A very small bias constant is used to prevent the denominator from being zero.

[0080]

[0081] In the formula, Indicates the number of samples in the training set; This represents the predicted short-circuit current increment from the SVM model. The actual short-circuit current increment is obtained by performing a first-order difference on the actual short-circuit current sample value.

[0082]

[0083] In the formula, This represents the region predicted by the SVM model. For every 30MW increase in installed capacity of new energy generating units, the node The increase in short-circuit current; This is the physical penalty weighting coefficient, used to balance the magnitude of data error and physical constraints. Its value is determined through preliminary experiments, ensuring that the physical constraint term... With root mean square error term Keep the magnitudes consistent to avoid any one factor dominating the optimization direction. This represents the physical theoretical short-circuit current increment sensitivity boundary obtained by combining the limiting of the new energy inverter control strategy with electrical distance. For every 30MW increase in installed capacity of new energy, at the target node The maximum limit of short-circuit current increment when a short-circuit fault occurs. Its determination process comprehensively considers the hardware tolerance limits and control strategies of power electronic equipment (new energy units, energy storage inverters, etc.), as well as the electrical distance between the fault point and the power source. This represents the number of various types of new energy power stations after the power grid is divided into zones.

[0084] Optionally, in the WECC 179-node system of this embodiment, preliminary experiments tested five sets of values ​​for γ: 1, 2, 3, 4, and 5. The corresponding MAPE values ​​for the test sets were 0.412%, 0.348%, 0.367%, 0.395%, and 0.438%, respectively. Therefore, γ=2 was selected as the optimal value. For power grids with other grid sizes and renewable energy ratios, the recommended value for γ is 1 to 5, and the specific value should be determined through the same preliminary experimental procedure.

[0085] The expression is:

[0086] In the formula, This represents an increase of 30MW in installed capacity. This represents the maximum short-circuit current output limiting coefficient of the new energy inverter; Indicates the power access node Rated line voltage at the grid connection point of the new energy generating unit; Represents the target fault node in the node impedance matrix and power access node Mutual impedance at the point; Represents the target fault node in the node impedance matrix Self-impedance.

[0087] about and Update Notes: The "power expansion" scenario considered in this embodiment refers to the gradual expansion of installed capacity of new energy power plants at planned access points, under the premise that the existing grid structure remains basically unchanged. In this scenario, the node impedance matrix is ​​mainly determined by the network topology and line parameters, and can be considered relatively stable in the short term, without needing to be updated with capacity disturbances; when there are structural adjustments to the grid (such as adding transmission lines or building new substations), the node impedance matrix needs to be recalculated. Z Then, substitute the data into the support vector machine model and retrain.

[0088] In this embodiment, the grid structure is fixed during sensitivity analysis, and incremental perturbations are applied only to the installed capacity. Specifically, the quantitative criterion for triggering model retraining by grid structure adjustments is: when the node impedance matrix... Z The row / column elements corresponding to any target fault node in the training process are compared to the impedance matrix of the baseline node. Z The relative rate of change Δ Z / Z When the error is greater than 5%, the node impedance matrix should be recalculated and substituted into the model for retraining. This 5% threshold is determined based on the engineering requirements for short-circuit current prediction accuracy in the IEEE 1547 standard. When ΔZ / Z≤5%, the prediction results of the original model are used within the allowable error range in engineering, and the corresponding short-circuit current prediction error does not exceed ±3%.

[0089] sparrow population individual location Corresponding fitness As shown below:

[0090] In the formula, Location of individual sparrows in a population Corresponding fitness; This represents the fitness value of an individual (the first sparrow). In this example, d=2, corresponding to the penalty factor. and kernel parameters Two dimensions, namely the fitness value of sparrow 1. in, The corresponding position of Sparrow 1 in the two dimensions, i.e., the corresponding penalty factor. and kernel parameters The value of .

[0091] S7. Iterate through the sparrow population positions using discoverers, joiners, and watchers, and obtain new penalty factors by decoding the new population positions. and kernel width The corresponding SVM model is used to calculate the error value of the new model and the fitness of the SSA, and to update the globally optimal sparrow population position and fitness.

[0092] The formula for updating the discoverer is as follows:

[0093] in, This represents the current iteration number; =1,2,3…, Represents dimensions; For the first The sparrow in the first During the nth iteration, at the... The value that a dimension can take; It is a random number; This represents the maximum number of iterations. This indicates the safety threshold for the current location. This indicates the warning value for the current location. ; It is a random number that follows a normal distribution; It is 1× A dimensional matrix where all elements are 1. When This indicates that a small number of sparrows have discovered a dangerous situation, prompting the flock to move to a safe location; The formula for updating new members is as follows:

[0094] in, For the first The position of the worst joiner in the next iteration; For the first The sparrow in the first During the nth iteration, at the... The value that a dimension can take; For the first The globally optimal joiner position during the next iteration; Randomly assign each element a value of 1 or 1 of 1× A dimensional matrix, and satisfying The superscript T indicates transpose; when When, it indicates the first If a new member fails to find food, its survival rate is low, and it needs to move to other areas to search for food in order to improve its fitness. Here, n represents the total number of individuals in the sparrow population. It is a random number that follows a normal distribution; It is 1× A 3D matrix where all elements are 1.

[0095] The formula for updating the vigilant is as follows:

[0096] in, It is the optimal location for the previous generation of the entire flock; The step size adjustment parameter is a normally distributed random number with a mean of 0 and a variance of 1. for[ Random numbers within the range of [1,1]; These are the fitness of the worst global scout, the best global scout, and the current scout, respectively. It is a minimal constant that avoids a denominator of 0.

[0097] when This indicates that the sparrow (scout) is operating on the edge of the population, making it vulnerable to predators and attacks; when This indicates that the sparrow (scout) is located in the center of the population and has sensed the danger of being attacked, and needs to quickly move to other sparrows in the area.

[0098] After iterating through the sparrow positions in SSA, a new penalty factor is obtained by mapping the current sparrow positions. With kernel parameters A new SVM model is constructed based on the training set. The comprehensive fitness of the new sparrow position is calculated using the SVM model, and the global optimal position and optimal fitness are updated.

[0099] S8. Set the iteration stopping condition. Stop the iteration when the convergence condition is met or the maximum number of iterations is reached, and obtain the optimal parameter combination and the SVM prediction model optimized based on the SSA algorithm.

[0100] This embodiment sets up a dual optimization termination criterion, with the following termination conditions: One is the maximum number of iterations. The maximum number of iterations is set to 200, which serves as the upper limit for the number of iterations the algorithm can run, ensuring the timely response of the model under complex power grid topology data. Secondly, regarding convergence accuracy, the algorithm monitors the changes in the optimal fitness function value in real time during the optimization process. If the preset comprehensive fitness target is reached, the algorithm is considered to have converged, and it exits the loop early and outputs the current optimal penalty factor. With kernel parameters And finally, the SVM model optimized by SSA.

[0101] S9. Apply the optimized SVM model to perform time series reconstruction and apply power supply increment perturbation to the current power grid to extract the sensitivity of the target node short-circuit current to various power supply increments in different regions.

[0102] After training and optimizing the PIML-SSA-SVM model, the penalty factor of the model's internal parameters is fixed. With kernel parameters .

[0103] Regarding the first The first power grid zone Power source characteristics, respectively applying 30MW single-variable installed capacity perturbations. Construct the first p Univariate independent perturbation eigenvectors of power sources :

[0104] In the formula, For feature dimensions; It is a 1*R vector with no power supply expansion, meaning all its internal values ​​are 0.

[0105] The first p Univariate independent perturbation eigenvectors of power sources Input the trained support vector machine model to calculate the true value of the short-circuit current before the disturbance. Short-circuit current response value after disturbance The difference is used to obtain the sensitivity matrix of the target node short-circuit current to the increment of various power sources in different regions; based on the sensitivity matrix, a heat map and contribution curve of each power source to the increment of the target node short-circuit current are plotted to complete the predictive analysis.

[0106] In the data preprocessing stage before training the SVM model, to improve the prediction accuracy of the sensitivity to the physical theory of short-circuit current increments, a first-order difference transformation is performed on both the input feature (power supply capacity) and the output target (short-circuit current) to achieve sequence stabilization. The SVM model trained in this way outputs the increment of the short-circuit current. During actual prediction, based on the recursive relationship of the time series, the current increment output by the model is superimposed with the original short-circuit current value of the predicted sample at the previous moment to reconstruct the absolute value of the short-circuit current at the current moment. This first-order difference processing flow is encapsulated within the SSA-SVM model framework, thus externally representing an end-to-end mapping relationship between the input power supply capacity and the output short-circuit current. Simultaneously, incremental data can still be extracted from intermediate stages of the model for sensitivity analysis and dynamic characteristic studies.

[0107] The first-order differential processing procedure is set as the process in the SSA-SVM model. Therefore, the input and output of the external model are still the power supply capacity and short-circuit current, but the incremental data obtained from the intermediate process can be extracted for research.

[0108] Based on the model prediction results, a heat map of the target node short-circuit current increment under the single-variable expansion of each power source was plotted. Based on the analysis results of the increment sensitivity heat map, four power sources with the highest increment sensitivity were selected, and their short-circuit current contribution curves were plotted.

[0109] S10. Apply the trained SVM model to predict the short-circuit current of the test set, and analyze the accuracy of the model by comparing the error between the predicted value and the true value.

[0110] Based on the trained PIML-SSA-SVM model, the short-circuit current values ​​of the test set samples are predicted, and a comparison chart of the predicted and actual short-circuit current values ​​is plotted. Error analysis is performed on the predicted short-circuit current values ​​to verify the accuracy of the model's prediction results for the short-circuit current increment.

[0111] This case study focuses on a high-proportion renewable energy power grid upgrade based on the WECC 179-node system. The power grid is first divided into zones and clusters. Based on practical engineering needs, this case study divides the grid into four regions, and the node with the largest short-circuit current in each region is selected as the target node for short-circuit current prediction. Figure 2 As shown.

[0112] Taking region 4 after partitioning as an example, a grey relational analysis was performed on the features initially selected for each region and the short-circuit current. The results are shown in Table 1, where NA indicates that this type of power source does not exist in the region. Regarding the engineering processing of NA in the table: When a certain type of power source does not exist in a certain region (i.e., the feature column is always 0 in the historical data), the grey relational degree of this feature to the current region cannot be effectively calculated. The engineering processing method is as follows: (1) The feature column is directly removed in the feature screening of the region and does not participate in the construction of the input features of the region; (2) However, when training the global PIML-SSA-SVM model across regions, the feature column is still retained and filled with 0 values ​​to maintain the uniformity of the model input dimension; (3) In the sensitivity analysis stage, the power source type marked with NA is directly output with the "not applicable to this region" label and no 30MW perturbation operation is performed. This processing method not only ensures the consistency of the model input dimension, but also avoids the interference of invalid features on the training accuracy.

[0113] Table 1: Grey Relational Analysis Results of the Initially Selected Features in Region 4 and the Short-Circuit Current

[0114] (1) Remove influencing factors with a correlation degree of less than 0.9 and retain feature values ​​with a high correlation degree. The final input feature set includes permanent magnet direct drive wind turbine installed capacity, grid-connected energy storage, centralized new energy access capacity, etc. (2) Table 2 shows the original sample data of the model in region 4 of this case, of which 500 sets are used as the training set and 50 sets are used as the test set. When training the model, the input features and short-circuit current data of all regions are input into the PIML-SSA-SVM model, and the short-circuit current increment prediction model can be trained. Figure 3Figures (a)-(d) show the heat map of the sensitivity of the short-circuit current of the target node in different regions of the high-proportion new energy power grid to the incremental power of different types of power sources in each region. The figure shows the increment of the short-circuit current of the target node in each region when each power source increases its installed capacity by 30MW. Figure 4 Based on the incremental sensitivity heatmap analysis results of Region 4, four power sources with the highest incremental sensitivity in this region were selected, and their short-circuit current contribution curves were plotted. In this embodiment, the dynamic extrapolation range of the expansion increment is set to 0~100MW. The basis for selecting this range is that in the planning and engineering construction of this power grid case, the installed capacity of a single expansion of a centralized photovoltaic power station or wind power project is usually within 100MW.

[0115] To verify the prediction accuracy of the PIML-SSA-SVM model, short-circuit currents on the test set were predicted based on the trained model. Taking the short-circuit current prediction results of the target node in region 4 as an example, the mean absolute percentage error (MAPE) of the model predictions was 0.348%. The prediction results are shown in the figure below. Figure 5 Error analysis of the model's prediction results demonstrates that the model can effectively predict the increase in short-circuit current in power grids with a high proportion of new energy sources, and grasp the impact of the expansion of different types of power sources in various regions on the short-circuit current of the power grid.

[0116] Table 2: Sample data for Region 4

[0117] This embodiment uses an improved Sparrow Search Algorithm (SSA) to optimize the parameters of the Support Vector Machine (SVM), which has stronger stability and generalization performance when solving small-sample problems such as short-circuit faults in large power grids.

[0118] This embodiment applies a partitioning clustering algorithm, which allows for the selection of partitioning detail based on project needs. This enables model dimensionality reduction while meeting project requirements, significantly improving model computation efficiency.

[0119] In this embodiment, the SSA algorithm has excellent global search capability and fast convergence speed, and can find the optimal SVM parameter configuration in a short time, thereby improving model training efficiency.

[0120] Compared to traditional methods that use simulation software such as PSASP to calculate short-circuit currents in power grids with a high proportion of renewable energy sources, the method proposed in this invention avoids the complex and cumbersome modeling and parameter settings for the power system topology and various renewable energy sources, and can more easily and efficiently predict short-circuit currents.

[0121] In this embodiment, a comprehensive fitness function based on physical information fusion was created during model training. The control strategy limiting features of new energy sources were added as physical constraints, which effectively improved the robustness and physical reliability of the model in complex evolution scenarios of high-proportion new energy power grids.

[0122] Example 2 The purpose of this embodiment is to provide a short-circuit current increment prediction and analysis system for high-proportion renewable energy power grids, including: The extraction module is configured to: collect historical operation data of high-proportion renewable energy power grids, analyze the mechanism of various influencing factors on short-circuit current based on the power grid short-circuit calculation principle, and extract key parameter characteristics of each region in high-proportion renewable energy power grids that affect short-circuit current; The training module is configured to: train the support vector machine model based on the key parameter characteristics affecting the short-circuit current in each area of ​​the high-proportion renewable energy power grid, and the corresponding short-circuit current of each target node; in the training of the support vector machine model, the physical theoretical short-circuit current increment sensitivity boundary calculated based on the limiting characteristics of renewable energy control strategy is used as the physical constraint condition, and the comprehensive fitness function of the sparrow search algorithm is constructed by combining the root mean square of the predicted short-circuit current value and the sample true value, and the sparrow search algorithm is used to optimize the penalty factor and kernel function width of the support vector machine model; The prediction module is configured to use a trained support vector machine model to predict and analyze the sensitivity of the short-circuit current of the target node in the current high-proportion renewable energy grid to the increment of various power sources in different regions.

[0123] The specific implementation methods of each module in this embodiment correspond to those in Embodiment 1, and will not be repeated here.

[0124] In further embodiments, the following is also provided: An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When executed by the processor, the computer instructions perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0125] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0126] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0127] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0128] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0129] The computer storage medium of this embodiment can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be—but is not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0130] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in connection with an instruction execution system, apparatus, or device.

[0131] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0132] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0133] A computer program product includes a computer program that, when executed by a processor, implements the method described in Embodiment 1.

[0134] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions included in program modules, which execute in a device on a target real or virtual processor to perform the processes / methods described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided among program modules as needed. The machine-executable instructions for the program modules can execute within a local or distributed device. In a distributed device, the program modules can reside in both local and remote storage media.

[0135] The computer program code used to implement the methods of the present invention may be written in one or more programming languages. This computer program code may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the computer or other programmable data processing device, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a computer, partially on a computer, as a stand-alone software package, partially on a computer and partially on a remote computer, or entirely on a remote computer or server.

[0136] In the context of this invention, computer program code or related data may be carried by any suitable carrier to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, and the like. Examples of signals may include electrical, optical, radio, sound, or other forms of propagation signals, such as carrier waves, infrared signals, etc.

[0137] Those skilled in the art will recognize that the units and algorithm steps described in conjunction with the embodiments herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0138] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for predicting and analyzing short-circuit current increments in high-proportion renewable energy power grids, characterized in that, include: Collect historical operation data of high-proportion renewable energy power grids, analyze the mechanism of various influencing factors on short-circuit current based on the power grid short-circuit calculation principle, and extract the key parameter characteristics of each region in high-proportion renewable energy power grids that affect short-circuit current. Based on the key parameter characteristics affecting short-circuit current in each area of ​​the high-proportion new energy power grid, and the corresponding short-circuit current of each target node, the support vector machine model is trained. In the training of the support vector machine (SVM) model, the physical theoretical short-circuit current increment sensitivity boundary calculated based on the limiting characteristics of the new energy control strategy is used as the physical constraint condition. Combining the root mean square of the predicted short-circuit current value and the actual sample value, a comprehensive fitness function for the sparrow search algorithm is constructed. The sparrow search algorithm is used to optimize the penalty factor and kernel width of the SVM model. Specifically, the position of each sparrow individual encodes the penalty factor and kernel width. By simulating the behavior of discoverers, joiners, and watchers, the sparrow population position is iteratively updated. In each iteration, the updated position of each sparrow individual is decoded to obtain a new penalty factor and kernel width. The comprehensive fitness is calculated based on the new penalty factor and kernel width. The global optimal position is updated according to the comprehensive fitness value until the iteration ends, obtaining the optimal penalty factor and kernel width. The comprehensive fitness is: in, Regarding the penalty factor and kernel width The root mean square error; It is a one-sided upper bound penalty for physical constraints; A very small bias constant; Indicates the number of samples in the training set; The short-circuit current increment predicted by the support vector machine model; This represents the actual short-circuit current increment. This represents the region predicted by the support vector machine model. node The increase in short-circuit current; This is the physical penalty weighting coefficient. The number of various types of new energy power stations after the power grid is divided; This represents the physical theory of the sensitivity boundary for short-circuit current increment; Using a trained support vector machine model, we predict and analyze the sensitivity of short-circuit current at target nodes of the current high-proportion renewable energy power grid to the incremental power supply of various types in different regions.

2. The method for predicting and analyzing short-circuit current increments in a high-proportion renewable energy power grid as described in claim 1, characterized in that, The high-proportion renewable energy power grid is partitioned based on the electrical distance between nodes, specifically as follows: Calculate the electrical distance between nodes based on the Thevenin equivalent impedance between each node. Based on the electrical distance matrix between nodes, the K-Means clustering algorithm is used to minimize the sum of the squared errors of the electrical distance feature vectors of nodes in each partition to the cluster center. The node cluster allocation and cluster center update are performed iteratively until the objective function converges, resulting in a preset number of partitions.

3. The short-circuit current increment prediction and analysis method in a high-proportion renewable energy power grid as described in claim 1, characterized in that, Based on the superposition principle and Thevenin's equivalence theorem, this paper analyzes the mathematical mapping relationship between the type of new energy generating unit, control strategy, access voltage level, and capacity and short-circuit current, and determines the initial feature library. Grey relational analysis is then performed on the data in the initial feature library and the short-circuit current data to select the appropriate grey relational degree. Initially selected features exceeding a set threshold are used as key parameters affecting short-circuit current. First-order difference and standardization preprocessing are performed on the key parameters and the corresponding short-circuit currents of each target node. The preprocessed data is then used to train the support vector machine model.

4. The method for predicting and analyzing short-circuit current increments in a high-proportion renewable energy power grid as described in claim 1, characterized in that, The formula for calculating the sensitivity boundary of the short-circuit current increment in the physical theory is as follows: in, This represents the physical theory of the sensitivity boundary for short-circuit current increment; This represents the maximum short-circuit current output limiting coefficient of the new energy inverter; Indicates the power access node The rated line voltage at the grid connection point of the new energy generating unit; Let be the self-impedance of the target fault node i in the node impedance matrix. is the mutual impedance between the target fault node i and the power supply access node e in the node impedance matrix; This indicates the increase in installed capacity.

5. A short-circuit current increment prediction and analysis system for high-proportion renewable energy power grids, characterized in that, include: The partitioning module is configured to partition the high-proportion renewable energy power grid based on the electrical distance between nodes. The extraction module is configured to: collect historical operation data of high-proportion renewable energy power grids, analyze the mechanism of various influencing factors on short-circuit current based on the power grid short-circuit calculation principle, and extract key parameter characteristics of each region in high-proportion renewable energy power grids that affect short-circuit current; The training module is configured to train the support vector machine model based on the key parameter characteristics affecting the short-circuit current in each area of ​​the high-proportion new energy power grid, as well as the corresponding short-circuit current of each target node. In the training of the support vector machine (SVM) model, the physical theoretical short-circuit current increment sensitivity boundary calculated based on the limiting characteristics of the new energy control strategy is used as the physical constraint condition. Combining the root mean square of the predicted short-circuit current value and the actual sample value, a comprehensive fitness function for the sparrow search algorithm is constructed. The sparrow search algorithm is used to optimize the penalty factor and kernel width of the SVM model. Specifically, the position of each sparrow individual encodes the penalty factor and kernel width. By simulating the behavior of discoverers, joiners, and watchers, the sparrow population position is iteratively updated. In each iteration, the updated position of each sparrow individual is decoded to obtain a new penalty factor and kernel width. The comprehensive fitness is calculated based on the new penalty factor and kernel width. The global optimal position is updated according to the comprehensive fitness value until the iteration ends, obtaining the optimal penalty factor and kernel width. The comprehensive fitness is: in, Regarding the penalty factor and kernel width The root mean square error; It is a one-sided upper bound penalty for physical constraints; A very small bias constant; Indicates the number of samples in the training set; The short-circuit current increment predicted by the support vector machine model; This represents the actual short-circuit current increment. This represents the region predicted by the support vector machine model. node The increase in short-circuit current; This is the physical penalty weighting coefficient. The number of various types of new energy power stations after the power grid is divided; This represents the physical theory of the sensitivity boundary for short-circuit current increment; The prediction module is configured to use a trained support vector machine model to predict and analyze the sensitivity of the short-circuit current of the target node in the current high-proportion renewable energy grid to the increment of various power sources in different regions.

6. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, perform the method described in any one of claims 1-4.

8. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the method described in any one of claims 1-4.

Images