New energy-containing power distribution system fault identification method and device

Abstract

The invention discloses a fault identification method and device for a power distribution system containing new energy, which are used for solving the technical problem that the existing fault identification method for the power distribution system containing new energy cannot ensure the accuracy of fault identification when a new energy unit is accessed. The method comprises the steps of collecting input characteristic quantities of a plurality of known fault categories, and performing preprocessing through a K-means clustering algorithm to generate target characteristic quantities; and determining a feature selection quantity by adopting a dynamic optimization strategy. When a fault occurs, a to-be-measured characteristic quantity is extracted according to the fault response electrical data of the power distribution node. And in combination with the pre-processed target feature quantity and the dynamically optimized feature selection quantity, matching the to-be-detected feature quantity with a known fault feature library by using a KNN algorithm, and finally outputting a fault recognition result.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a fault identification method and device for a power distribution system containing new energy sources. Background Technology

[0002] The new energy distribution system refers to a new power system architecture that deeply integrates distributed renewable energy (such as photovoltaic, wind power, and energy storage) with the traditional distribution network. The new energy distribution system undertakes the tasks of distributing power, stepping down and transforming voltage, and improving the reliability of power supply, playing an indispensable role in the operation of the power system.

[0003] Since the new energy units in the distribution system containing new energy generally connect to the grid through power electronic equipment, the fault current exhibits "weak feedback characteristics," which leads to the risk of traditional overcurrent protection failing to operate. Therefore, the complexity and severity of fault types in the distribution system containing new energy are significantly higher than those in the traditional distribution network. Timely and reliable determination of the fault type in the distribution system containing new energy is of great significance for reducing the losses caused by short-circuit faults and improving the reliability of the power grid.

[0004] Existing fault identification methods for power distribution systems incorporating new energy sources primarily rely on measurement-based protection algorithms and signal processing methods. These methods collect fault characteristic quantities through current / voltage transformers, then extract fault frequency domain features using Fourier transform, and finally determine the fault type by combining threshold comparison methods or impedance calculation methods. However, the fault current output by new energy units is limited and controlled by power electronic equipment, and its amplitude may be as low as 20% of that of traditional faults. This results in insensitivity of overcurrent-based protection systems, making it impossible to guarantee the accuracy of fault identification when new energy units are connected. Summary of the Invention

[0005] This invention provides a fault identification method and apparatus for power distribution systems containing new energy sources, which solves the technical problem that existing fault identification methods for power distribution systems containing new energy sources cannot guarantee the accuracy of fault identification when new energy units are connected.

[0006] The first aspect of this invention provides a fault identification method for a new energy power distribution system, comprising:

[0007] Multiple known fault categories containing new energy power distribution system fault input features are obtained, and K-means clustering algorithm is used to preprocess the multiple known fault categories containing new energy power distribution system fault input features to generate multiple target known fault categories containing new energy power distribution system fault input features.

[0008] A preset dynamic optimization strategy is adopted to determine the number of target features to be selected based on the known fault categories of multiple targets, including fault input features of new energy power distribution systems.

[0009] When receiving fault response electrical data from multiple distribution nodes in a power distribution system containing new energy, the fault input characteristic quantity of the power distribution system containing new energy to be tested is determined based on the fault response electrical data of each distribution node.

[0010] The K-nearest neighbor algorithm is used to identify faults based on multiple known fault categories including new energy power distribution system fault input features, the fault input features of the new energy power distribution system to be tested, and the number of target features selected, and outputs the fault identification result corresponding to the fault input features of the new energy power distribution system to be tested.

[0011] Optionally, the preset dynamic optimization strategy includes a preset hyperparameter search space and an accuracy function; the step of determining the number of target features to be selected based on multiple known fault categories, including new energy power distribution system fault input features, using the preset dynamic optimization strategy includes:

[0012] Multiple selection quantities of candidate features are generated in the preset hyperparameter search space;

[0013] The accuracy function is used to determine the accuracy corresponding to the number of selected candidate features based on the number of selected candidate features and the input feature quantity of multiple target known fault categories including new energy power distribution system faults.

[0014] The number of candidate features selected corresponding to the highest accuracy is taken as the number of target features selected.

[0015] Optionally, the fault response electrical data includes current distribution coefficient, short-circuit current, fault clearing time, fault initiation time, node voltage after fault occurrence, and node voltage value during normal operation; determining the fault input characteristic quantity of the distribution system containing new energy under test based on the fault response electrical data of each distribution node includes:

[0016] Based on the current distribution coefficient, short-circuit current, fault clearing time, and fault initiation time of each distribution node, calculate the single-unit energy value in the form of short-circuit current of each distribution node.

[0017] The node voltage drop coefficient of each distribution node is calculated using the node voltage after a fault occurs and the node voltage during normal operation.

[0018] The voltage drop coefficient of each distribution node is compared with a preset coefficient threshold.

[0019] Any distribution node whose voltage drop coefficient is less than the preset coefficient threshold is considered a fault-related node.

[0020] Based on the node voltage drop coefficient and the single-unit energy value in the form of short-circuit current of the fault-related nodes, the fault input characteristic quantities of the power distribution system containing new energy to be tested are generated.

[0021] Optionally, the step of using the K-nearest neighbor algorithm to perform fault identification based on multiple known fault categories including new energy power distribution system fault input features, the fault input features of the new energy power distribution system to be tested, and the number of target features selected, and outputting the fault identification result corresponding to the fault input features of the new energy power distribution system to be tested, includes:

[0022] Calculate the Euclidean distance between the fault input feature quantity of the new energy power distribution system to be tested and the fault input feature quantity of each known fault category of the new energy power distribution system;

[0023] Sort the Euclidean distances in ascending order, and select the number of known fault categories, including new energy power distribution system fault input features, as the final known fault category features.

[0024] The frequency of occurrence of each fault category corresponding to the known fault category feature quantity is statistically analyzed, and the fault category with the highest frequency of occurrence is taken as the fault identification result corresponding to the fault input feature quantity of the new energy power distribution system under test.

[0025] Optionally, the formula for calculating the single-unit energy value in the form of short-circuit current is as follows:

[0026] ;

[0027] in, Let t be the single-unit energy value in the form of the short-circuit current of the i-th distribution network node; ci and t 0i These are the fault clearing time and fault initiation time of the i-th distribution network node, respectively; M i E represents the moment of inertia of the i-th distribution network node. i C represents the internal potential amplitude of the i-th distribution network node; i I is the current distribution coefficient of the i-th distribution network node; fa t represents the amplitude of the short-circuit current at the fault point; t represents any moment during the fault period. Let be the internal potential vector of the i-th distribution network node and the angle of the fault-added current; k is the rate of change of the power angle of different generator sets at the initial moment of the fault. Let be the initial power angle of the i-th distribution network node.

[0028] Optionally, the formula for calculating the node voltage sag factor is as follows:

[0029] ;

[0030] in, Let be the node voltage drop factor of the i-th distribution network node; Let be the node voltage after the fault occurs at the i-th distribution network node; Let be the node voltage value of the i-th distribution network node during normal operation.

[0031] The second aspect of the present invention provides a fault identification device for a new energy power distribution system, comprising:

[0032] The acquisition module is used to acquire multiple known fault categories containing new energy power distribution system fault input features, and to preprocess the multiple known fault categories containing new energy power distribution system fault input features using the K-means clustering algorithm to generate multiple target known fault categories containing new energy power distribution system fault input features.

[0033] The module is used to determine the number of target features to be selected based on the input feature quantities of multiple known fault categories of the target, including faults in new energy power distribution systems, using a preset dynamic optimization strategy;

[0034] The determination module is used to determine the fault input characteristic quantity of the new energy power distribution system under test based on the fault response electrical data of each of the multiple power distribution nodes when receiving fault response electrical data of multiple power distribution nodes in the power distribution system containing new energy.

[0035] The identification module is used to identify faults using the K-nearest neighbor algorithm based on multiple known fault categories including new energy power distribution system fault input features, the fault input features of the new energy power distribution system to be tested, and the number of target features selected, and outputs the fault identification result corresponding to the fault input features of the new energy power distribution system to be tested.

[0036] A computer device provided in a third aspect of the present invention includes a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the fault identification method for new energy power distribution systems as described in any of the preceding claims.

[0037] The fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the steps of the fault identification method for a new energy power distribution system as described in any of the preceding claims.

[0038] The fifth aspect of the present invention provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein, when the program instructions are executed by a computer, the computer performs the steps of the fault identification method for new energy power distribution systems as described in any of the preceding claims.

[0039] As can be seen from the above technical solutions, the present invention has the following advantages:

[0040] The present invention provides a fault identification method for a power distribution system containing new energy sources. This method acquires multiple known fault categories of fault input features for the power distribution system containing new energy sources, and preprocesses these features using a K-means clustering algorithm to generate multiple target known fault categories of fault input features. A preset dynamic optimization strategy is used to determine the number of target features to be selected based on these target known fault categories. When fault response electrical data from multiple distribution nodes in the power distribution system containing new energy sources are received, the method determines the fault input features of the power distribution system containing new energy sources to be tested based on the fault response electrical data of each distribution node. The system uses a set of fault input features; it employs the K-nearest neighbor algorithm to identify faults based on multiple known fault categories including new energy power distribution system fault input features, the fault input features of the new energy power distribution system to be tested, and the number of target features selected, and outputs the fault identification result corresponding to the fault input features of the new energy power distribution system to be tested; based on the above scheme, this invention combines a preset dynamic optimization strategy, the K-nearest neighbor algorithm, and the K-means clustering algorithm to perform fault identification processing on multiple known fault categories including new energy power distribution system fault input features and the fault input features of the new energy power distribution system to be tested, which can increase the focus on the fault area, thereby improving the accuracy of fault identification when new energy units are connected. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating the steps of a fault identification method for a new energy power distribution system according to Embodiment 1 of the present invention.

[0043] Figure 2 This is a schematic diagram of a single-machine infinite bus system provided in Embodiment 1 of the present invention;

[0044] Figure 3 This is a schematic diagram illustrating the changes in fault energy after different types of short-circuit faults occur in a line, as provided in Embodiment 1 of the present invention.

[0045] Figure 4 This is a flowchart illustrating a fault identification method for a new energy power distribution system according to Embodiment 1 of the present invention.

[0046] Figure 5This is a structural block diagram of a fault identification device for a new energy power distribution system provided in Embodiment 2 of the present invention. Detailed Implementation

[0047] This invention provides a method and apparatus for fault identification in power distribution systems containing new energy sources, which addresses the technical problem that existing fault identification methods for power distribution systems containing new energy sources cannot guarantee the accuracy of fault identification when new energy units are connected.

[0048] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0049] Terminology Explanation:

[0050] The KNN algorithm (K-Nearest Neighbors Algorithm) is a non-parametric, supervised learning classification artificial intelligence algorithm where all computation occurs during the classification or prediction process. The core idea of ​​the KNN algorithm is to identify input items and assign them the most frequently occurring labels for classification.

[0051] Fault identification: Fault identification refers to the process of detecting abnormalities and determining the type, location, and cause of faults by monitoring, analyzing, and judging the operating status of power systems or their equipment.

[0052] Please see Figure 1 , Figure 1 The flowchart illustrates the steps of a fault identification method for a new energy power distribution system provided in Embodiment 1 of the present invention.

[0053] This invention provides a fault identification method for a power distribution system containing new energy sources, comprising:

[0054] Step 101: Obtain multiple known fault categories containing new energy power distribution system fault input features, and use K-means clustering algorithm to preprocess the multiple known fault categories containing new energy power distribution system fault input features to generate multiple target known fault categories containing new energy power distribution system fault input features.

[0055] It should be noted that in the operating environment of modern power grids, the volatility of the power transmitted by renewable energy generators makes the changes in electrical quantities after a short-circuit fault more complex; at the same time, the large-scale integration of distributed power sources into the grid further complicates the grid topology. In this context, the traditional KNN algorithm (K-Nearest Neighbors Algorithm), as a relatively simple artificial intelligence algorithm, may experience a significant decrease in efficiency during classification and prediction due to excessively large datasets or external interference. Furthermore, the value of parameter k (i.e., the determination of the number of target features selected) has a significant impact on the final algorithm result, but the traditional KNN algorithm generally chooses a fixed k value for its operation. Therefore, this invention improves the traditional KNN algorithm in two aspects: data optimization processing and dynamic optimization of the k value, to enhance its accuracy and efficiency in fault identification in distribution systems containing renewable energy sources.

[0056] Furthermore, during the process of collecting the required fault identification input features, external factors such as environmental noise interference and monitoring equipment failure may lead to the generation of erroneous data; at the same time, collecting a large number of sample features will cause the KNN algorithm to take too long to calculate, thus reducing the efficiency of fault identification.

[0057] To address the aforementioned issues, this invention introduces the K-Means Algorithm (KMA) to preprocess the sample dataset (containing multiple known fault categories and fault input features of new energy power distribution systems). This process removes erroneous data from the sample set and compresses the sample set size, resulting in multiple target known fault categories containing fault input features of new energy power distribution systems. This ensures the accuracy and efficiency of power grid fault area identification.

[0058] Furthermore, the K-Means algorithm is a classic clustering analysis technique. Its core lies in choosing an appropriate distance formula to measure the similarity of data objects and clustering similar data points together. The algorithm's execution flow is as follows: First, data points are randomly selected as initial cluster centers, and the distance between each cluster center and other data points is calculated. The distance calculation formula is:

[0059] (1)

[0060] Where D and C are the data object and cluster center, respectively; D i and C i These are the specific values ​​for the i-th data object and the i-th cluster center, respectively.

[0061] Furthermore, based on the distance between each data point and its nearest cluster center, the data points are grouped into clusters. Next, the average value of the data in each cluster is calculated to re-establish new cluster centers. Then, the distances between the cluster centers and the remaining data points, as well as the cluster centers themselves, are iteratively calculated until a predetermined stopping condition is met.

[0062] Step 102: Using a preset dynamic optimization strategy, determine the number of target features to be selected based on the known fault categories of multiple targets, including the fault input features of new energy power distribution systems.

[0063] The preset dynamic optimization strategy includes a preset hyperparameter search space and an accuracy function.

[0064] Specifically, step 102 may include the following sub-steps S21-S23:

[0065] Step S21: Generate multiple selection quantities of candidate features in the preset hyperparameter search space;

[0066] Step S22: Using an accuracy function, determine the accuracy corresponding to the number of selected features for each candidate feature based on the number of selected features and the input feature quantity of multiple known fault categories including new energy power distribution system faults.

[0067] Step S23: Select the number of candidate features selected corresponding to the highest accuracy as the number of target features selected.

[0068] It should be noted that before the KNN algorithm is executed, the value of k (i.e., the determination of the number of target features selected) directly determines the accuracy of the algorithm's recognition. When faced with different training samples formed by the number of features collected in different application scenarios, this invention uses the idea of ​​hyperparameter search to dynamically select the optimal value of k. The specific process is as follows: 1) Determine the hyperparameter space and grid: Set the hyperparameter search space as:

[0069] (2)

[0070] Where k represents the number of neighbors in the KNN algorithm; Let k represent the set of possible values ​​for k; i Let k represent the number of times the i-th optional feature is selected; It is a set of integers.

[0071] 2) Initialization and Validation Scheme: The dataset (consisting of multiple known fault categories, including fault input features of new energy power distribution systems) is divided into a training set and a validation set. An accuracy function is initialized to measure the degree of optimization of the k-value. The accuracy function is defined. for:

[0072] (3)

[0073] in, This represents the number of samples misidentified in the validation set for the given value of k; T is the total number of samples in the validation set.

[0074] 3) Evaluation and selection: through traversal For all k values, perform KNN classification and calculate the accuracy function. Select the k value that maximizes the accuracy:

[0075] (4)

[0076] in, This represents the optimal value of k for the current application scenario, i.e., the number of target features selected.

[0077] It is worth mentioning that when new fault samples are collected or the operating environment changes significantly, the above process is repeated to update the k value again, ensuring that the classification performance is dynamically adjusted according to the operating conditions.

[0078] Step 103: When receiving fault response electrical data from multiple distribution nodes in a power distribution system containing new energy, determine the fault input characteristic quantity of the power distribution system containing new energy to be tested based on the fault response electrical data of each distribution node.

[0079] Fault response electrical data includes current distribution coefficient, short-circuit current, fault clearing time, fault initiation time, node voltage after fault occurrence, and node voltage during normal operation.

[0080] Specifically, step 103 may include the following sub-steps S31-S35:

[0081] Step S31: Calculate the single-unit energy value in the form of short-circuit current for each distribution node based on the current distribution coefficient, short-circuit current, fault clearing time, and fault initiation time of each distribution node.

[0082] Step S32: Calculate the voltage drop factor of each distribution node by using the node voltage after the fault occurs and the node voltage during normal operation.

[0083] Step S33: Compare the node voltage drop coefficient of each power distribution node with the preset coefficient threshold.

[0084] Step S34: Take any distribution node whose voltage drop coefficient is less than the preset coefficient threshold as the fault-related node.

[0085] Step S35: Generate the fault input characteristic quantity of the power distribution system containing new energy sources to be tested based on the node voltage drop coefficient and the single-unit energy value in the form of short-circuit current of the fault-related node.

[0086] It should be noted that a single-machine infinite system, such as Figure 2 As shown, when different types of short-circuit faults occur on the line, the fault energy output by the generator will vary depending on the fault type. When a short-circuit fault occurs, the short-circuit current in the fault response information is the most direct indicator of the fault type. By deriving the fault energy function in the form of short-circuit current, and using the fault energy as one of the feature quantities of the improved KNN fault identification model, accurate identification of the fault type can be achieved.

[0087] For any generator i, the transient energy function V after a system fault occurs Ti for:

[0088] (5)

[0089] Among them, V ki V represents the kinetic energy of generator i; pi M represents the potential energy of generator i; i Let be the moment of inertia of generator i; This represents the relative angular velocity of generator i; Let be the initial power angle of generator i; This indicates the generator power angle when the fault of generator i is cleared; For The electromagnetic power of generator i is the independent variable; Let i be the mechanical power of generator i; Let i be the power angle of the generator.

[0090] When the generator is operating stably, the terminal current is the load current. When a short-circuit fault occurs on the line, the current distribution coefficient is used. The additional current generated by the fault can be calculated. By superimposing the two currents, the generator terminal current during the fault can be obtained. for:

[0091] (6)

[0092] Among them, I i Let ΔI be the load current of the i-th distribution network node. Each distribution network node corresponds to a generation unit, such as a synchronous machine (generator); i Additional fault current for the i-th distribution network node; I fa This is the short-circuit current at the fault point.

[0093] Current distribution coefficient of generator i The calculation formula is:

[0094] (7)

[0095] in, This indicates that after a fault occurs, a new node j is added at the location of the short circuit, and the mutual admittance between generator i and node j; This represents the self-admittance of node j.

[0096] The electromagnetic power of the generator can be obtained from this. for:

[0097] (8)

[0098] Among them, P Ei,0 The electromagnetic power of the generator during steady-state operation; and E i These are the generator's internal electromotive force vector and amplitude, respectively; and I fa These are the short-circuit current vectors and their magnitudes at the fault point, respectively. for and ΔI i The angle.

[0099] The equation of motion for the generator after being disturbed is:

[0100] (9)

[0101] in, Let be the derivative of the power angle of generator i; Let be the derivative of the relative angular velocity of generator i.

[0102] The Taylor series expansion of the electromagnetic power during the fault is as follows:

[0103] (10)

[0104] in, Let i be the electromagnetic power of the generator at the initial power angle; Let be the derivative of the electromagnetic power of generator i at the initial power angle; for Truncation error of Taylor series expansion.

[0105] Since power system faults are generally short-lived, the higher-order terms of the electromagnetic power expansion can be ignored, and the above equation can be simplified to:

[0106] (11)

[0107] Based on the above, the fault period was obtained. and The general solution expression is:

[0108] (12)

[0109] Where A is The constant term corresponding to the general solution expression; B is The coefficient of the first term in the general solution expression represents the rate of change of the power angle for different generator sets; C is... The constant term corresponding to the general solution expression; t represents any time during the fault period.

[0110] At the initial moment of the fault =0, = d / dt=k, where k is the rate of change of the power angle of different generator sets at the initial moment of the fault, obtained by combining the above formula. and The particular solution expression:

[0111] (13)

[0112] Furthermore, equation (14) characterizes the period during the fault. and The changes with short-circuit current and time. By combining equations (5) and (12), the fault energy function expression in the form of short-circuit current can be obtained, that is, the calculation formula for the single-machine energy value in the form of short-circuit current, which is as follows:

[0113] (13)

[0114] in, Let t be the single-unit energy value in the form of the short-circuit current of the i-th distribution network node; ci and t 0i These are the fault clearing time and fault initiation time of the i-th distribution network node, respectively; M i E represents the moment of inertia of the i-th distribution network node. i C represents the internal potential amplitude of the i-th distribution network node; i I is the current distribution coefficient of the i-th distribution network node; fa t represents the amplitude of the short-circuit current at the fault point; t represents any moment during the fault period. Let be the internal potential vector of the i-th distribution network node and the angle of the fault-added current; k is the rate of change of the power angle of different generator sets at the initial moment of the fault. Let be the initial power angle of the i-th distribution network node.

[0115] Furthermore, from equation (13), it can be seen that the generator fault energy of this form varies only with the current distribution coefficient C. i Short-circuit current I fa These three variables are functions that change with time t. They can all be directly read from a wide-area measurement system or simply calculated from the read data. Choosing generator fault energy as the input characteristic has the advantage of being easily obtainable.

[0116] When different types of short-circuit faults occur in the line, the fault energy values ​​(single-unit energy values ​​in the form of short-circuit current) calculated by equation (13) during the fault period show significant differences. By constructing a fault energy dataset corresponding to different fault types occurring in the entire line, accurate identification of different fault types can be achieved. The changes in fault energy values ​​after three-phase ground faults, single-phase ground faults, and two-phase-to-phase faults occur in the line are shown in the figure. Figure 3 As shown.

[0117] Furthermore, by Figure 3 The trend of fault energy changes reveals that the fault categories identified in this invention include three types of short-circuit faults: three-phase-to-ground short-circuit faults, two-phase-to-phase short-circuit faults, and single-phase-to-ground short-circuit faults. Among these three types, the fault energy change range caused by a three-phase-to-ground short-circuit fault is the largest, followed by a two-phase-to-phase short-circuit fault, while the fault energy change range after a single-phase-to-ground short-circuit fault is the smallest. The fault energy change ranges of these three types of short-circuit faults do not overlap and exhibit significant differences. Furthermore, regardless of the type of short-circuit fault occurring on the line, the fault energy change range exhibits a linear characteristic. After a short-circuit fault occurs at any location on the line, the calculated fault energy value will fall within the change range of its corresponding short-circuit fault. Therefore, selecting fault energy as a feature quantity for the improved KNN fault identification model can help the model accurately distinguish between various fault types.

[0118] Furthermore, with the increasing penetration rate of new energy units in the power grid, the capacity and complexity of the power system are also constantly increasing. In such large-scale power systems, directly applying fault identification models for fault detection may lead to false alarms, thereby affecting the safe and stable operation of the system and the normal transmission of electricity. Therefore, adding fault area screening indicators as input features can help fault type identification models more accurately locate key areas of concern, thereby improving the accuracy of fault type identification.

[0119] This invention utilizes the node voltage sag coefficient as an input feature to help a fault type identification model locate fault regions. The voltage sag coefficient of a node is obtained by calculating the ratio of the post-fault voltage to the pre-fault voltage. The voltage sag coefficient of distribution network node i is also described. As shown in equation (14):

[0120] (14)

[0121] in, Let be the node voltage drop factor of the i-th distribution network node; Let be the node voltage after the fault occurs at the i-th distribution network node; Let be the node voltage value of the i-th distribution network node during normal operation.

[0122] Furthermore, a boundary voltage value (preset coefficient threshold) is set for a zone. ∈(0,1), when the voltage drop factor obtained after a fault in distribution network node i is less than the set boundary voltage value, that is:

[0123] (15)

[0124] This node is considered a fault-related node, indicating a fault in the power grid. In power grids containing renewable energy units, the transmission characteristics of renewable energy change when the grid connection point voltage drops below 0.9 pu; simultaneously, given the regulatory requirements for supply voltage deviation, a zone-specific critical voltage can be set. It is 0.9 pu. Using the fault identification model as the starting criterion, when the criterion is met, the generator fault energy in the form of voltage drop coefficient and short-circuit current is comprehensively used as input features to improve the KNN model. That is, based on the K nearest neighbor algorithm, fault identification is performed according to the fault input feature quantity of the power distribution system containing new energy under test, which can achieve the goal of fault type identification.

[0125] Step 104: Using the K-nearest neighbor algorithm, fault identification is performed based on multiple known fault categories including new energy power distribution system fault input features, the fault input features of the new energy power distribution system to be tested, and the number of target features selected. The fault identification result corresponding to the fault input features of the new energy power distribution system to be tested is then output.

[0126] Specifically, step 104 may include the following sub-steps S41-S43:

[0127] Step S41: Calculate the Euclidean distance between the fault input characteristic quantity of the new energy power distribution system to be tested and the fault input characteristic quantity of each known fault category of the new energy power distribution system.

[0128] Step S42: Sort the Euclidean distances in ascending order and select the number of known fault categories, including the fault input features of the new energy power distribution system, as the final known fault category features.

[0129] Step S43: Calculate the frequency of occurrence of each fault category corresponding to the feature quantity of each known fault category, and take the fault category with the highest frequency of occurrence as the fault identification result corresponding to the fault input feature quantity of the power distribution system containing new energy under test.

[0130] It should be noted that the KNN algorithm, as a non-parametric, supervised learning classification artificial intelligence algorithm, involves all computations occurring during the classification or prediction process. The core idea of ​​the KNN algorithm is to identify input items and assign the most frequently occurring label for classification. The specific principle of the KNN algorithm is as follows: the operating state of each component in the system, represented by a 1-dimensional vector group, is shown in equation (1):

[0131] (16)

[0132] Among them, S1, S2…S N N represents the operating status of components in the system; N is the total number of components detected during fault identification.

[0133] After establishing a point coordinate system in two-dimensional space, the distance d(s1,s2) between two sample data points S1(x1,y1) and S2(x2,y2) is calculated as follows:

[0134] (17)

[0135] Where x1 is the x-coordinate of the first operating state of the component in the system; y1 is the y-coordinate of the first operating state of the component in the system; x2 is the x-coordinate of the second operating state of the component in the system; and y2 is the y-coordinate of the second operating state of the component in the system.

[0136] Furthermore, in the process of data classification using the KNN algorithm, calculating the distance between two sample data points is crucial, as this calculation directly determines the proximity between the sample points. Common distance calculation methods used in the algorithm execution include Euclidean distance, Manhattan distance, Chebyshev distance, Hamming distance, cosine similarity, and Mahalanobis distance. This invention uses Euclidean distance as the metric to classify sample data. The specific process of the KNN algorithm in handling sample data classification is as follows: First, input the new sample features to be classified (features of new energy distribution system faults to be tested), and calculate the distance between the new feature and each previously input feature (features of known fault categories including new energy distribution system faults) during training. Second, determine the frequency of occurrence of each category among the k nearest features, i.e., select a number of known fault categories including new energy distribution system faults as the final known fault category features, and statistically analyze the frequency of occurrence of each fault category corresponding to each final known fault category feature. Finally, the category with the highest frequency among the first k features is taken as the classification result of the data to be classified. That is, the fault category corresponding to the highest frequency is taken as the fault identification result corresponding to the input feature of the new energy power distribution system to be tested. Generally, in order to balance the accuracy and efficiency of the algorithm decision, the value of k is usually chosen as an odd number not greater than 20.

[0137] For example, suppose there are four known fault category features: the first known fault category is three-phase ground fault, the second is three-phase ground fault, the third is two-phase short circuit, and the fourth is single-phase ground fault. Based on the frequency of occurrence of each category, the three-phase ground fault has the highest frequency. Therefore, the fault identification result corresponding to the fault input feature of the new energy power distribution system under test is determined to be a three-phase ground fault.

[0138] For comparison of technical effectiveness, existing technologies can be referenced. In the research of power system fault identification, wide-area measurement technology plays a crucial role. By preserving the positive and negative characteristics of the fault charge acquired over a wide area using the local integration method and comprehensively considering the basic probability allocation of all positively adjacent lines, an improved evidence theory is employed to achieve reliable identification of faulty lines. A novel wide-area fault identification method based on current polarity and relative energy entropy is proposed by defining the relative entropy of current sampling values ​​and solving the polarity relationship of the boundary current in the associated domain. Based on the measured wide-area measurement information architecture, a fault line detection algorithm based on spectral analysis and wide-area backup protection is proposed to achieve online fault type identification. A fault dataset is formed by extracting bus voltage and current after a fault occurs using a wide-area measurement system, and an artificial neural network fault identification model is built and trained to achieve fault detection and identification.

[0139] However, power system fault identification methods often face problems such as high computational load, high communication requirements, high dependence on data accuracy, and significant susceptibility to noise interference. This makes it difficult to guarantee the accuracy of fault identification when new energy units are connected, and they cannot adequately meet the requirements of modern power system fault identification. Furthermore, the implementation of this method is highly dependent on the transmission speed of real-time measurement data between the measurement system and the control center, requiring very high transmission speed and data accuracy.

[0140] To address the aforementioned problems, this invention provides a fault identification method for power distribution systems containing renewable energy sources. It utilizes an improved KNN algorithm to identify faults in such systems, meeting the power system's need for timely and reliable fault type determination. This method achieves rapid and accurate fault identification, enabling timely detection and resolution of potential faults, reducing power outages, and improving grid reliability. Furthermore, it proposes optimization strategies for the algorithm in two aspects: data optimization processing and dynamic optimization of the k-value. Through these optimizations, rapid identification of power system fault types is achieved. For details, please refer to [link to specific documentation]. Figure 4In the offline training phase, generator fault energy and node voltage sag coefficient data under different fault types of the target power grid are acquired offline as input features. A sample dataset is formed based on the input features of different fault types. A fault type identification model based on the improved KNN algorithm is trained, thus obtaining the number of target features selected. In the online application phase, fault response electrical data measured by WAMS are extracted. The voltage sag coefficient of each node and generator fault energy in the form of short-circuit current are calculated. It is calculated whether the voltage sag coefficient of each node meets the requirements. If it does, the fault identification model is started; otherwise, it is terminated. Under the premise of meeting the requirements, generator fault energy and node voltage sag coefficient are used as input features. The feature values ​​are input to the fault type identification model based on the improved KNN algorithm (i.e., input to the K nearest neighbor algorithm), thereby outputting the fault type identification result.

[0141] Compared with existing technologies, this invention proposes optimization strategies for the traditional KNN algorithm in two aspects: data optimization processing and dynamic optimization of k-values. The improved KNN algorithm is used as the fault identification model. The fault energy in the form of generator short-circuit current, which has significant differences under different fault types and is easy to obtain, is selected as the input feature. At the same time, in order to increase the attention of the fault identification model to the fault area and improve the accuracy of the identification model, the node voltage drop coefficient is added as an input feature and the starting criterion of the fault identification model. After determining that the node voltage drop coefficient is less than the set critical voltage, the fault identification model based on the improved KNN is started, and the generator fault energy and the voltage drop coefficient of each node are calculated as input features and input into the model to achieve accurate identification of the fault type.

[0142] In this embodiment of the invention, a fault identification method for a power distribution system containing new energy sources is provided. The method acquires multiple known fault categories of fault input features for the power distribution system containing new energy sources, and preprocesses these features using a K-means clustering algorithm to generate multiple target known fault categories of fault input features for the power distribution system containing new energy sources. A preset dynamic optimization strategy is used to determine the number of target features to be selected based on these target known fault categories of fault input features. When fault response electrical data from multiple distribution nodes in the power distribution system containing new energy sources is received, the method determines the fault input features to be identified based on the fault response electrical data of each distribution node. The invention identifies fault input features of the power distribution system. It employs the K-nearest neighbor algorithm to perform fault identification based on multiple known fault categories including new energy power distribution system fault input features, the fault input features of the new energy power distribution system to be tested, and the number of target features selected. The result is then output as the fault identification result corresponding to the fault input features of the new energy power distribution system to be tested. Based on this approach, the invention combines a preset dynamic optimization strategy, the K-nearest neighbor algorithm, and the K-means clustering algorithm to perform fault identification processing on multiple known fault categories including new energy power distribution system fault input features and the fault input features of the new energy power distribution system to be tested. This increases the focus on fault areas, thereby improving the accuracy of fault identification when new energy units are connected.

[0143] Please see Figure 5 , Figure 5 This is a structural block diagram of a fault identification device for a new energy power distribution system provided in Embodiment 2 of the present invention.

[0144] The present invention provides a fault identification device for a new energy power distribution system, comprising:

[0145] The acquisition module 501 is used to acquire multiple known fault categories containing new energy power distribution system fault input features, and to preprocess the multiple known fault categories containing new energy power distribution system fault input features using the K-means clustering algorithm to generate multiple target known fault categories containing new energy power distribution system fault input features.

[0146] Module 502 is used to determine the number of target features to be selected based on the known fault categories of multiple targets, including the fault input features of new energy power distribution systems, using a preset dynamic optimization strategy.

[0147] The determination module 503 is used to determine the fault input characteristic quantity of the new energy power distribution system under test based on the fault response electrical data of each power distribution node when receiving fault response electrical data of multiple power distribution nodes in the power distribution system containing new energy.

[0148] The identification module 504 is used to identify faults based on multiple known fault categories including new energy power distribution system fault input features, the fault input features of the new energy power distribution system to be tested, and the number of target features selected, and outputs the fault identification result corresponding to the fault input features of the new energy power distribution system to be tested.

[0149] Furthermore, the preset dynamic optimization strategy includes a preset hyperparameter search space and an accuracy function, using module 502, specifically for:

[0150] Generate multiple candidate feature selection numbers in the preset hyperparameter search space;

[0151] An accuracy function is used to determine the accuracy corresponding to the number of selected features for each candidate feature based on the number of selected features and the input feature quantity of multiple known fault categories including faults in new energy power distribution systems.

[0152] The number of candidate features selected corresponding to the highest accuracy is taken as the number of target features selected.

[0153] Furthermore, the fault response electrical data includes the current distribution coefficient, short-circuit current, fault clearing time, fault initiation time, node voltage after the fault occurs, and node voltage during normal operation; the determination module 503 is specifically used for:

[0154] Based on the current distribution coefficient, short-circuit current, fault clearing time, and fault initiation time of each distribution node, calculate the single-unit energy value in the form of short-circuit current of each distribution node.

[0155] The node voltage drop factor of each distribution node is calculated by using the node voltage after a fault occurs and the node voltage during normal operation.

[0156] The voltage drop coefficient of each distribution node is compared with the preset coefficient threshold.

[0157] Any distribution node whose voltage drop coefficient is less than a preset threshold is considered a fault-related node.

[0158] Based on the node voltage drop coefficient and the single-unit energy value in the form of short-circuit current of the fault-related nodes, the fault input characteristic quantities of the power distribution system containing new energy to be tested are generated.

[0159] Furthermore, the identification module 504 is specifically used for:

[0160] Calculate the Euclidean distance between the fault input characteristics of the power distribution system containing new energy sources to be tested and the fault input characteristics of each known fault category containing new energy sources.

[0161] Sort the Euclidean distances in ascending order and select the number of known fault categories, including new energy power distribution system fault input features, as the final known fault category features.

[0162] The frequency of occurrence of each fault category corresponding to the feature quantity of each known fault category is statistically analyzed, and the fault category with the highest frequency of occurrence is taken as the fault identification result corresponding to the fault input feature quantity of the power distribution system containing new energy under test.

[0163] Furthermore, the formula for calculating the single-unit energy value in the form of short-circuit current is as follows:

[0164] ;

[0165] in, Let t be the single-unit energy value in the form of the short-circuit current of the i-th distribution network node; ci and t 0i These are the fault clearing time and fault initiation time of the i-th distribution network node, respectively; M i E represents the moment of inertia of the i-th distribution network node. i C represents the internal potential amplitude of the i-th distribution network node; i I is the current distribution coefficient of the i-th distribution network node; fa t represents the amplitude of the short-circuit current at the fault point; t represents any moment during the fault period. Let be the internal potential vector of the i-th distribution network node and the angle of the fault-added current; k is the rate of change of the power angle of different generator sets at the initial moment of the fault. Let be the initial power angle of the i-th distribution network node.

[0166] Furthermore, the formula for calculating the node voltage sag factor is as follows:

[0167] ;

[0168] in, Let be the node voltage drop factor of the i-th distribution network node; Let be the node voltage after the fault occurs at the i-th distribution network node; Let be the node voltage value of the i-th distribution network node during normal operation.

[0169] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described device and module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0170] This invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program; when the computer program is executed by the processor, the processor performs the steps of the fault identification method for new energy power distribution systems as described in any of the above embodiments.

[0171] This invention also provides a computer-readable storage medium storing a computer program / instructions thereon, which, when executed by a processor, implements the steps of the fault identification method for new energy power distribution systems as described in any of the above embodiments.

[0172] This invention also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the fault identification method for new energy power distribution systems as described in any of the above embodiments.

[0173] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0174] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0175] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault identification method for a power distribution system incorporating new energy sources, characterized in that, include: Multiple known fault categories containing new energy power distribution system fault input features are obtained, and K-means clustering algorithm is used to preprocess the multiple known fault categories containing new energy power distribution system fault input features to generate multiple target known fault categories containing new energy power distribution system fault input features. A preset dynamic optimization strategy is adopted to determine the number of target features to be selected based on the known fault categories of multiple targets, including fault input features of new energy power distribution systems. When receiving fault response electrical data from multiple distribution nodes in a power distribution system containing new energy, the fault input characteristic quantity of the power distribution system containing new energy to be tested is determined based on the fault response electrical data of each distribution node. The K-nearest neighbor algorithm is used to identify faults based on multiple known fault categories including new energy power distribution system fault input features, the fault input features of the new energy power distribution system to be tested, and the number of target features selected, and outputs the fault identification result corresponding to the fault input features of the new energy power distribution system to be tested.

2. The fault identification method for a power distribution system containing new energy sources according to claim 1, characterized in that, The preset dynamic optimization strategy includes a preset hyperparameter search space and an accuracy function; the step of using the preset dynamic optimization strategy to determine the number of target features to select based on multiple known fault categories, including fault input features of new energy power distribution systems, includes: Multiple selection quantities of candidate features are generated in the preset hyperparameter search space; The accuracy function is used to determine the accuracy corresponding to the number of selected candidate features based on the number of selected candidate features and the input feature quantity of multiple target known fault categories including new energy power distribution system faults. The number of candidate features selected corresponding to the highest accuracy is taken as the number of target features selected.

3. The fault identification method for a power distribution system containing new energy sources according to claim 1, characterized in that, The fault response electrical data includes current distribution coefficient, short-circuit current, fault clearing time, fault initiation time, node voltage after fault occurrence, and node voltage value during normal operation. The step of determining the fault input characteristic quantity of the power distribution system containing new energy sources under test based on the fault response electrical data of each of the power distribution nodes includes: Based on the current distribution coefficient, short-circuit current, fault clearing time, and fault initiation time of each distribution node, calculate the single-unit energy value in the form of short-circuit current of each distribution node. The node voltage drop coefficient of each distribution node is calculated using the node voltage after a fault occurs and the node voltage during normal operation. The voltage drop coefficient of each distribution node is compared with a preset coefficient threshold. Any distribution node whose voltage drop coefficient is less than the preset coefficient threshold is considered a fault-related node. Based on the node voltage drop coefficient and the single-unit energy value in the form of short-circuit current of the fault-related nodes, the fault input characteristic quantities of the power distribution system containing new energy to be tested are generated.

4. The fault identification method for a power distribution system containing new energy sources according to claim 1, characterized in that, The K-nearest neighbor algorithm is used to identify faults based on multiple known fault categories including new energy power distribution system fault input features, the fault input features of the new energy power distribution system to be tested, and the number of target features selected. The fault identification result corresponding to the fault input features of the new energy power distribution system to be tested is output, including: Calculate the Euclidean distance between the fault input feature quantity of the new energy power distribution system to be tested and the fault input feature quantity of each known fault category of the new energy power distribution system; Sort the Euclidean distances in ascending order, and select the number of known fault categories, including new energy power distribution system fault input features, as the final known fault category features. The frequency of occurrence of each fault category corresponding to the known fault category feature quantity is statistically analyzed, and the fault category with the highest frequency of occurrence is taken as the fault identification result corresponding to the fault input feature quantity of the new energy power distribution system under test.

5. The fault identification method for a power distribution system containing new energy sources according to claim 3, characterized in that, The formula for calculating the single-unit energy value in the form of short-circuit current is as follows: ； in, Let t be the single-unit energy value in the form of the short-circuit current of the i-th distribution network node; ci and t 0i These are the fault clearing time and fault initiation time of the i-th distribution network node, respectively; M i E represents the moment of inertia of the i-th distribution network node. i C represents the internal potential amplitude of the i-th distribution network node; i I is the current distribution coefficient of the i-th distribution network node; fa t represents the amplitude of the short-circuit current at the fault point; t represents any moment during the fault period. Let be the internal potential vector of the i-th distribution network node and the angle of the fault-added current; k is the rate of change of the power angle of different generator sets at the initial moment of the fault. Let be the initial power angle of the i-th distribution network node.

6. The fault identification method for a power distribution system containing new energy sources according to claim 3, characterized in that, The formula for calculating the node voltage sag factor is as follows: ； in, Let be the node voltage drop factor of the i-th distribution network node; Let be the node voltage after the fault occurs at the i-th distribution network node; Let be the node voltage value of the i-th distribution network node during normal operation.

7. A fault identification device for a new energy power distribution system, characterized in that, include: The acquisition module is used to acquire multiple known fault categories containing new energy power distribution system fault input features, and to preprocess the multiple known fault categories containing new energy power distribution system fault input features using the K-means clustering algorithm to generate multiple target known fault categories containing new energy power distribution system fault input features. The module is used to determine the number of target features to be selected based on the input feature quantities of multiple known fault categories of the target, including faults in new energy power distribution systems, using a preset dynamic optimization strategy; The determination module is used to determine the fault input characteristic quantity of the new energy power distribution system under test based on the fault response electrical data of each of the multiple power distribution nodes when receiving fault response electrical data of multiple power distribution nodes in the power distribution system containing new energy. The identification module is used to identify faults using the K-nearest neighbor algorithm based on multiple known fault categories including new energy power distribution system fault input features, the fault input features of the new energy power distribution system to be tested, and the number of target features selected, and outputs the fault identification result corresponding to the fault input features of the new energy power distribution system to be tested.

8. A computer device, characterized in that, The system includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of the fault identification method for a new energy power distribution system as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the fault identification method for power distribution systems containing new energy sources as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, wherein when the program instructions are executed by a computer, the computer performs the fault identification method for a power distribution system containing new energy as described in any one of claims 1-6.

Images