Filter bank breaker closing resistor state evaluation method and system

Abstract

The application belongs to the technical field of high-voltage electrical equipment state monitoring and fault diagnosis, and particularly relates to a filter set circuit breaker closing resistor state evaluation method and system. The steps are as follows: based on ±800kV engineering, a refined PSCAD / EMTP simulation model containing an alternating current filter, a circuit breaker and a closing resistor is established, and after being verified by measured data, a multi-working condition database is simulated and constructed; the closing resistor on-off transient process is analyzed, the waveform is segmented by using a mutation point detection algorithm, and time domain and frequency domain characteristics such as inrush current peak value, input time and harmonic component are extracted; the KNN algorithm is adopted, the Euclidean distance is used to measure the similarity between the to-be-measured sample and the database characteristics, the optimal K value 5 is selected, and the closing resistor resistance value estimation is matched and output. The application does not need to be disassembled, realizes accurate online evaluation of the normal, fault and degradation states of the closing resistor, and improves the evaluation reliability and efficiency.

Description

Technical Field

[0001] This invention belongs to the field of high-voltage electrical equipment condition monitoring and fault diagnosis technology, specifically relating to a method and system for assessing the closing resistance of a filter bank circuit breaker. Background Technology

[0002] Converter stations, as core hubs, ensure the efficient and stable flow of electrical energy between different power systems. To guarantee the normal operation of converter stations and minimize the impact on the power grid, AC filters need to be switched on and off frequently to effectively filter out harmonics generated by the converters and provide reactive power compensation, thereby improving the power quality of the system and ensuring the stability and reliability of power transmission. As capacitive devices, AC filters are prone to significant inrush currents and overvoltages during switching operations, and their frequent switching operations can easily lead to performance degradation and frequent failures of the circuit breakers equipped with them. Therefore, filter bank circuit breakers are often equipped with closing resistors to suppress inrush currents and protect the main contacts from arcing. However, during long-term operation, closing resistors are prone to resistance drift, open circuits, or insulation degradation due to thermal stress, mechanical fatigue, and other factors. Investigations have revealed multiple instances of 800kV circuit breaker failures in filter fields caused by defective closing resistors at converter stations in Northwest China, adversely affecting the safety and stability of DC transmission.

[0003] Existing technologies have included research on methods for detecting closing resistance. In the invention patent with publication number CN118731603B, this method, while capable of estimating the closing resistance and determining flashover through wavelet decomposition and TLS-ESPRIT analysis, has inherent limitations. Its underlying mechanism relies on a fixed mapping relationship between the single frequency characteristic of "amplitude-attenuation factor" and the resistance value. In practical applications, TLS-ESPRIT is extremely sensitive to the accuracy of the signal-to-noise ratio and model order. This algorithm can only provide a few modal parameters such as amplitude and attenuation factor. This limited feature dimension essentially restricts the robustness of the established mapping relationship, making it difficult to capture state changes under complex operating conditions. Ultimately, this results in a fragile foundation and insufficient generalization ability of the entire diagnostic model.

[0004] A patent application (CN119205606B) proposes a method for analyzing the closing resistance waveform of industrial substations based on a YOLOv8 neural network. This method parses waveform data files in COMTRADE format to obtain the actual value of the closing current, smooths it, and fits it with a Fourier transform to identify the error type of the current waveform. However, its core problem lies in directly applying the YOLOv8 model, which is specifically designed for image recognition, to one-dimensional current waveform data. This mismatch between the architecture and the characteristics of the problem makes the model a "black box" that relies on a large amount of labeled data. Its decision-making process lacks clear electrical and physical basis, resulting in poor interpretability and questionable generalization ability. Furthermore, this solution can only perform post-event analysis on recorded faults, failing to achieve real-time status monitoring and early warning, thus limiting its engineering value. Currently, research on the relationship between closing resistance and inrush current relies excessively on artificial intelligence models, weakening the explanation of its underlying physical mechanisms and the effective extraction of key features. This leads to insufficient interpretability of the analysis results, making it difficult to support accurate risk assessment. Therefore, there is an urgent need to develop a new method based on real waveform data that can deeply reflect the physical relationship between resistance state and current characteristics, and to achieve accurate identification and risk assessment of closing resistor faults by building a high-quality database. Summary of the Invention

[0005] This invention provides a method and system for evaluating the closing resistance status of a filter bank circuit breaker, in order to solve the technical problems existing in the prior art, such as limited feature dimensions, insufficient model robustness and generalization ability, over-reliance on artificial intelligence and weakening of physical mechanism explanation, and lack of high-quality databases and accurate fault identification and evaluation methods based on real waveform data that reflect the physical relationship between resistance status and current characteristics.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for assessing the closing resistance status of a filter bank circuit breaker, such as... Figure 5 As shown, it includes the following steps:

[0008] Establish a refined PSCAD / EMTP simulation model of the filter field and construct a multi-condition closing resistor simulation database covering normal and fault states;

[0009] Based on circuit theory, the transient process of the closing resistor being engaged and disengaged is analyzed. A sudden change point detection algorithm is used to intelligently segment the waveform, and multidimensional features of the transient process of the closing resistor being engaged and disengaged are extracted from the time domain and frequency domain respectively.

[0010] By using the K-nearest neighbor algorithm, multi-dimensional features are matched with a multi-condition closing resistor simulation database to obtain the closing resistor values ​​of the top n in similarity ranking, as well as the corresponding operating conditions, thereby achieving the assessment of the closing resistor status.

[0011] The aforementioned simulation database for closing resistors under various operating conditions, based on the PSCAD / EMTP simulation model, is constructed. Specifically, based on the ±800kV UHVDC transmission project structure, a refined simulation model including AC filters, circuit breaker bodies, and closing resistors is established in PSCAD / EMTP. The model is then validated using inrush current data collected during actual operation under both normal and fault conditions to ensure its accuracy and robustness. On this basis, a systematic simulation is performed within a sinusoidal cycle at 18° intervals, using the abnormal state, degradation degree, and closing phase angle of the closing resistor as variables, thus constructing a simulation database for closing resistors covering various operating conditions.

[0012] The model is validated using inrush current data collected during actual operation under normal and fault conditions. This is achieved by simulating normal, abnormal, and degraded operating conditions of the closing resistor. Specifically: abnormal operating conditions include breakdown, short circuit, and open circuit; under normal operating conditions, the inrush current waveform of the closing resistor at its rated resistance is simulated; under abnormal operating conditions, the closing process of the closing resistor under breakdown, short circuit, and open circuit fault states is simulated to obtain typical fault characteristic waveforms; under degraded operating conditions, the closing process of the closing resistor under different degrees of resistance drift is simulated to obtain inrush current characteristics reflecting the resistance degradation process; the different degrees of resistance drift include resistance reduction. By comparing and analyzing the consistency of the simulated data and measured inrush current data under the above operating conditions in terms of amplitude and waveform, a comprehensive verification of the model's accuracy and generalization ability is completed.

[0013] The multi-condition closing resistor simulation database includes different operating conditions under normal and fault states, multi-dimensional feature data corresponding to different operating conditions, and closing resistor values ​​corresponding to different operating conditions.

[0014] The transient process of the closing resistor being engaged and disengaged is analyzed based on circuit theory, and a sudden change point detection algorithm is used to intelligently segment the waveform. Specifically, a circuit model is established for the two stages of the closing resistor being engaged and disengaged to obtain the time-domain expression of the corresponding stage current, which theoretically reveals the intrinsic relationship between the closing resistor value, inrush current amplitude, and waveform. Then, the sudden change point detection algorithm is used to intelligently segment the waveforms during the engagement and disengagement stages of the closing resistor.

[0015] The waveforms during the switching resistor's engagement and disengagement phases are intelligently segmented using a mutation point detection algorithm. The specific steps are as follows:

[0016] Determine the closing time, and use the closing time as the starting point, with a window length of 0.1s, to capture the current waveform;

[0017] Based on the captured current waveform, the signal difference is calculated, that is, the difference between adjacent data is calculated to obtain the gradient signal;

[0018] Take 2% of the maximum value in the entire gradient signal as the threshold;

[0019] Traverse the gradient signal and find the index position where the gradient value exceeds the threshold;

[0020] The signals corresponding to the index positions that exceed the threshold are used as candidate index points, and the candidate points corresponding to the index positions are rearranged in chronological order.

[0021] Based on the sorted candidate points, the first candidate point is selected as the initial mutation point. The minimum time interval between mutation points is set to 0.007 seconds. Subsequent candidate points are checked. If the time difference between a candidate point and the previous selected mutation point is ≥0.007 seconds, then the candidate point is selected as the second mutation point. The two mutation points correspond to the time when the closing resistor is put into operation and the time when it is taken out, respectively, so as to realize the intelligent segmentation of the waveform during the operation and exit stages.

[0022] Multidimensional features of the transient process during the two stages of closing resistor connection and disconnection are extracted from the time domain. Specifically, the transient current exhibits inrush current peak value and closing resistor connection time characteristics in the time domain. Specifically, inrush current peak value: when the closing resistor value is normal, the inrush current peak value is effectively suppressed; when the resistor breaks down or its resistance value degrades, the inrush current peak value increases significantly. Closing resistor connection time: under normal operating conditions, the connection time is stable; when the resistor is abnormal, the connection time may be prolonged or shortened, reflecting changes in mechanical or electrical performance.

[0023] Multidimensional features of the transient process during the two stages of closing the resistor being engaged and disengaged are extracted from the frequency domain. Specifically, the characteristics of the transient current in the frequency domain include the distribution of harmonic components and the characteristic frequency amplitude ratio. Specifically, the harmonic component distribution is as follows: under normal resistor conditions, the harmonic content is low and dominated by low-order harmonics; when the resistor breaks down or is open-circuited, the mid-to-high frequency harmonic components are significantly enhanced, and the spectral energy distribution shifts; the characteristic frequency amplitude ratio shows a regular change in the amplitude ratio of the fundamental wave to a specific harmonic under different closing resistor conditions, which can be used as a sensitive indicator for state identification.

[0024] The K-Nearest Neighbors algorithm, specifically, involves: collecting a large amount of known closing data under normal and fault conditions through refined simulation and field tests, and extracting their feature vectors; when a new sample to be tested is input, the K-Nearest Neighbors algorithm calculates its Euclidean distance to every known sample in the database. The closer the distance, the more similar the waveform features.

[0025] A filter bank circuit breaker closing resistance status assessment system includes a simulation module, a feature extraction module, and a status assessment module.

[0026] The simulation module is used to establish a refined PSCAD / EMTP simulation model of the filter field and to build a multi-condition closing resistor simulation database covering normal and fault states.

[0027] The feature extraction module is used to analyze the transient process of the closing resistor being engaged and disengaged based on circuit theory, intelligently segment the waveform using a sudden change point detection algorithm, and extract multi-dimensional features of the transient process of the closing resistor being engaged and disengaged from the time domain and frequency domain respectively.

[0028] The state assessment module is used to match multi-dimensional features with a multi-condition closing resistor simulation database using the K-nearest neighbor algorithm to obtain the closing resistor values ​​of the top n in similarity ranking, as well as the corresponding operating conditions, thereby realizing the state assessment of the closing resistor.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] This invention establishes a refined PSCAD / EMTP simulation model of the filter field and fully verifies the model by combining inrush current data collected during actual operation under normal and fault conditions, effectively ensuring the model's accuracy and robustness. Based on this, a multi-condition closing resistance simulation database is constructed, covering various operating conditions under both normal and fault states. It includes multi-dimensional feature data and closing resistance values ​​corresponding to different operating conditions. The data is comprehensive and deeply reflects the physical relationship between resistance state and current characteristics, completely changing the current situation where feature dimensions are limited and diagnostic models lack robustness and generalization ability. This lays a solid data foundation for subsequent accurate condition assessment. Furthermore, this invention enables online condition assessment of the closing resistance without disassembling the circuit breaker, eliminating the need for complex power outage checks. The operation is simple and does not affect equipment operation, effectively reducing maintenance costs and safety risks. By accurately assessing the resistance value and operating conditions of the closing resistor, abnormal and degraded states of the resistor can be detected in a timely manner, enabling early warning and preventing circuit breaker failures caused by defects in the closing resistor. This significantly improves the safety and stability of the converter station filter field operation and provides strong support for the safe and stable transmission of high-voltage direct current systems.

[0031] This invention, based on circuit theory, deeply analyzes the transient processes of the closing resistor's engagement and disengagement, clearly revealing the intrinsic relationship between the closing resistor's resistance value, inrush current amplitude, and waveform. It overcomes the shortcomings of existing technologies that over-rely on artificial intelligence models and weaken the explanation of physical mechanisms. By using a mutation point detection algorithm to intelligently segment the current waveform, it can accurately define the two transient stages of the closing resistor's engagement and disengagement. Multidimensional features are then extracted from both the time and frequency domains. These features all have clear physical meaning, comprehensively capturing the waveform variation patterns under different resistance states, effectively solving the problems of insufficient feature extraction and poor interpretability of analysis results in existing technologies.

[0032] This invention employs the K-nearest neighbor algorithm for state assessment. It determines the closing resistor value and corresponding operating condition by calculating the similarity between the features of the sample under test and the features of known samples in the simulation database. The algorithm logic is clear and easy to implement. This assessment method does not rely on a large amount of labeled data and avoids the black-box problem caused by mismatch between architecture and problem characteristics. It can achieve accurate judgment of the closing resistor state, overcoming the limitations of existing technologies where some models have questionable generalization ability and can only perform post-fault analysis. This provides a reliable technical means for closing resistor state assessment. Attached Figure Description

[0033] Figure 1 This is a cross-sectional view of the field inlet line of the AC filter in an embodiment of the present invention;

[0034] Figure 2 This is a simulation diagram of the filter bank circuit breaker and closing resistor in an embodiment of the present invention;

[0035] Figure 3 This serves as a validation of the effectiveness of the simulation model under typical operating conditions in the embodiments of the present invention;

[0036] Figure 4 This is a schematic diagram of the equivalent circuit analysis of the AC filter branch in an embodiment of the present invention;

[0037] Figure 5 This is a flowchart of the closing resistor state evaluation method based on transient current characteristics in an embodiment of the present invention;

[0038] Figure 6 This is a schematic diagram illustrating the engagement and disengagement of the closing resistor under normal operating conditions in an embodiment of the present invention.

[0039] Figure 7 This is the distribution of harmonic content of the closing resistor under different operating conditions in the embodiments of the present invention;

[0040] Figure 8 This shows the training effect of the K-nearest neighbor algorithm in this embodiment of the invention. Detailed Implementation

[0041] To further understand the content of this invention, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0042] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0043] This invention proposes a method for evaluating the closing resistance status of a filter bank circuit breaker, comprising the following steps:

[0044] Establish a refined PSCAD / EMTP simulation model of the filter field and construct a multi-condition closing resistor simulation database covering normal and fault states;

[0045] Based on circuit theory, the transient process of the closing resistor being engaged and disengaged is analyzed. A sudden change point detection algorithm is used to intelligently segment the waveform, and multidimensional features of the transient process of the closing resistor being engaged and disengaged are extracted from the time domain and frequency domain respectively.

[0046] By using the K-nearest neighbor algorithm, multi-dimensional features are matched with a multi-condition closing resistor simulation database to obtain the closing resistor values ​​of the top n in similarity ranking, as well as the corresponding operating conditions, thereby achieving the assessment of the closing resistor status.

[0047] Based on the above method steps, this invention will be described in detail with reference to the accompanying drawings. It mainly includes three parts: the construction of the closing resistor simulation database, the division and feature analysis of the closing resistor's input and output process, and the use of the KNN algorithm to evaluate the closing resistor's state. The details are as follows:

[0048] This invention focuses on AC filter fields, and its structural schematic diagram is shown below. Figure 1 As shown, the system includes: a 750kV circuit breaker 1, a 750kV high-voltage AC disconnector 2, a 750kV high-voltage AC grounding switch 3, a BP11 / BP13 filter 4, an SC capacitor 5, a 750kV incoming line surge arrester 6, a 750kV crossover 7, and a 750kV power distribution unit 8. The AC filter field is arranged linearly along the line direction, from left to right: the BP11 / BP13 filter 4, the 750kV high-voltage AC grounding switch 3, the 750kV circuit breaker 1, the 750kV high-voltage AC disconnector 2, the 750kV circuit breaker 1, the 750kV high-voltage AC grounding switch 3, the SC capacitor 5, and the 750kV incoming line surge arrester 6. The 750kV incoming line surge arrester 6 is connected to the 750kV power distribution unit 8 via the 750kV crossover 7. Electrical connections are achieved, and each device is connected sequentially through electrical lines, forming a complete layout and power transmission path for the AC filter field, from filtering equipment and switching equipment to capacitors and surge arresters, and finally connected to the power distribution device via cross-line connections.

[0049] This invention is based on an ±800kV UHVDC transmission project, focusing on the aforementioned AC filter field. The AC filter field comprises parallel single-tuned filter banks BP11 / BP13, dual-tuned filter banks HP24 / 36, a high-pass filter bank HP3, and a parallel capacitor bank SC. The AC filter group is switched using SF6 tank-type circuit breakers with a rated voltage of 800kV. The main break adopts a series double-break structure, with parallel voltage-equalizing capacitors connecting the breaks to balance the voltage. A closing resistor with a resistance of 1500Ω is also installed. Based on the actual equipment configuration of the aforementioned AC filter field, as well as the circuit breaker structural parameters and closing resistor parameters, a refined PSCAD / EMTP simulation model is built in PSCAD / EMTP simulation software. The model covers the AC filters, circuit breaker body, and closing resistor, ensuring accurate reflection of the actual equipment's operating characteristics. The simulation model is as follows: Figure 2 As shown in the diagram, the model starts with a three-phase power supply system. The current in phase B is monitored via a current transformer (CT). The phase B line is connected to a circuit breaker circuit integrating a closing resistor, parallel capacitor, auxiliary break, and main break. This circuit ultimately forms an electrical connection with the BP11 / 13 filter. Simultaneously, three functional branches are connected in parallel: branch 7622 connecting to the HP3 filter, branch 7623 connecting to the HP24 / 36 filter, and branch 7624 connecting to the SC capacitor; thus establishing an electrical interaction between the AC filter bank and the closing resistor.

[0050] Furthermore, based on the refined PSCAD / EMTP simulation model, inrush current data collected during actual operation under normal and fault conditions are used to verify the model, ensuring its good accuracy and robustness. On this basis, using the abnormal state, degradation degree, and closing phase angle of the closing resistor as variables, systematic simulations are performed at 18° intervals within a sinusoidal cycle, constructing a closing resistor simulation database covering various operating conditions. The effectiveness verification of the refined PSCAD / EMTP simulation model is based on typical fault scenarios, specifically simulating normal, abnormal, and degraded operating conditions of the closing resistor for simulation verification. Abnormal operating conditions include breakdown, short circuit, and open circuit. Under normal operating conditions, the closing inrush current waveform of the closing resistor at its rated resistance value is simulated. Under abnormal operating conditions, the closing process of the closing resistor under breakdown, short circuit, and open circuit fault states is simulated, obtaining typical fault characteristic waveforms. Under degraded operating conditions, the closing process of the closing resistor under different degrees of resistance drift is simulated, obtaining inrush current characteristics reflecting the resistance degradation process. The different degrees of resistance drift include resistance reduction. By comparing and analyzing the consistency of the simulation data and the measured inrush flow data under the above working conditions in terms of amplitude and waveform, a comprehensive verification of the model's accuracy and generalization ability is completed.

[0051] The process involves comparing and analyzing the consistency of amplitude and waveform between simulated and measured inrush current data under the aforementioned operating conditions to comprehensively verify the model's accuracy and generalization ability. Specifically, the simulated transient current data is precisely compared with the actual waveform recordings under the corresponding conditions. The comparison results are as follows: Figure 3 As shown, Figure 3 (a) is a comparison diagram of the closing resistor under normal operating conditions; Figure 3 (b) is a comparison diagram under the condition of short circuit of closing resistor; Figure 3 (c) is a comparison diagram of the closing resistor under open circuit conditions. Under normal operating conditions, the simulated curve and the actual recorded curve are highly consistent in terms of waveform trend and amplitude variation. Under short circuit conditions, the two show good consistency in the inrush current mutation and oscillation process. Under open circuit conditions, the curves have a high degree of matching in terms of the mutation time of inrush current characteristics and waveform shape. These highly consistent comparison results fully verify the accuracy and generalization ability of the established PSCAD / EMTP refined simulation model under different operating conditions, and confirm that the established simulation model has high effectiveness and engineering applicability. Based on the verified PSCAD / EMTP refined simulation model, a full-condition scanning simulation of the closing resistor in the resistance range of 0-1500Ω was carried out, with the closing phase covering the complete cycle from 0° to 360°, to construct a closing resistor simulation database that reflects the multi-dimensional characteristics of the closing resistor resistance value and phase combination under different operating conditions.

[0052] To analyze the impact of the closing resistor on inrush current, based on circuit principles, the AC filter branch is considered as an RLC circuit in series. The equivalent capacitance of the AC filter is C, and the equivalent resistance and equivalent inductance of the circuit are R and L, respectively, determined by the AC filter and the AC system it is connected to. Based on this, the equivalent model of the AC filter branch is established as follows: Figure 4 As shown. Based on circuit theory, the transient process of the closing resistor being engaged and disengaged in two stages is analyzed. According to the circuit described above, the KVL loop equations are written and solved to obtain:

[0053]

[0054] In the formula: U m This is the peak power supply voltage. i For loop current, ω Angular frequency, t For time, The closing phase angle, U C The voltage across the capacitor. R The total resistance of the circuit is 1. L The total inductance of the circuit, L For loop capacitance, Z Let σ be the loop impedance and σ be the decay time constant. φ For the impedance angle, α and ω 0 is a custom parameter, defined as follows:

[0055]

[0056] like Figure 4 middle, s 1 and s 2 represents the main break and auxiliary break of the circuit breaker, respectively. Main break s 1. Closing the circuit breaker activates the closing resistor, denoted as "Stage I". At this time, the capacitor... U C =0, closing resistor is applied, circuit impedance is:

[0057]

[0058] At this point, the peak inrush current is negatively correlated with the closing resistor value. If the measured inrush current is much greater than expected, it strongly suggests that the closing resistor may have an abnormally low resistance, such as a partial short circuit in the chip resistor. Therefore, the magnitude of the inrush current in Stage I is a direct electrical quantity for diagnosing whether the resistor has undergone substantial degradation or a short circuit.

[0059] When auxiliary circuit breaker S2 closes, the closing resistor is taken out of service, denoted as "Stage II". At this time, the voltage across the capacitor... U C = u c (0) is the initial voltage stored during the period when the closing resistor is in operation. After the closing resistor is removed from operation, the circuit impedance decreases significantly, which can be equivalent to the instantaneous charging of the capacitor by the power supply. The resistance value of the closing resistor directly affects... U C Resistance level: Normal / High: Slow charging. U C Low resistance. Abnormal / low resistance: rapid charging. U C It is closer to the power supply voltage. If the difference between the power supply voltage and the initial voltage of the capacitor is large, the energy that the two need to exchange will also increase accordingly, resulting in a high inrush current peak at the moment of closing.

[0060] In summary, the magnitude of the inrush current during closing is a "comprehensive output signal" of the closing resistor in its two operating stages. It inherently couples the resistance value and its dynamic functional effect. By capturing this key time-domain characteristic, a reliable bridge can be built between the resistance state and measurable electrical quantities, providing a core basis for achieving accurate non-destructive diagnostics.

[0061] To effectively divide the two stages of the closing resistance, a sudden change point detection algorithm is used to intelligently segment the waveform and automatically identify sudden changes in the current waveform. For example... Figure 6The diagram illustrates the division of the closing resistor's on / off state. First, the closing time is determined, and a window of 0.1 seconds is extracted from this time to capture the current waveform. Second, based on the captured current waveform, the differential gradient of the signal is calculated, and 2% of the maximum gradient value is used as a threshold to filter out all candidate index points exceeding the threshold. Subsequently, the candidate points are arranged in chronological order. Considering the closing resistor's on / off time is 8-11 ms, adjacent candidate points are judged at a minimum interval of 0.007 seconds. Subsequent candidate points are checked; if the time difference between a candidate point and the previous selected abrupt change point is ≥0.007 seconds, that candidate point is selected as the second abrupt change point. The two abrupt change points correspond to the closing resistor's on / off time, respectively. Finally, the key abrupt change points representing the switching state transition of the closing resistor are extracted from the points that meet the interval conditions.

[0062] In a preferred embodiment of the present invention, multi-dimensional feature extraction is performed on the current waveforms during the two transient stages of the closing resistor being engaged and disengaged, using both time and frequency domains. These multi-dimensional features include inrush current peak value, closing resistor engagement time, harmonic component distribution, and characteristic frequency amplitude ratio, to comprehensively characterize its dynamic response characteristics. In the time domain analysis, the following key indicators are examined: inrush current peak value and closing resistor engagement time. In the frequency domain analysis, a Fourier transform is performed on the transient current signal, decomposing the time-domain waveform into harmonic components of different frequencies, and its spectral structure is systematically examined. Specifically: Inrush current peak value: When the closing resistor value is normal, the inrush current peak value is effectively suppressed; when the resistor breaks down or its resistance value degrades, the inrush current peak value increases significantly; Closing resistor engagement time: Under normal operating conditions, the engagement time is stable; when the resistor is abnormal, the engagement time may be prolonged or shortened, reflecting changes in mechanical or electrical performance. Under normal operating conditions, the harmonic content is low and dominated by low-order harmonics. When the resistor breaks down or is open-circuited, the mid-to-high frequency harmonic components are significantly enhanced, and the spectral energy distribution shifts. Characteristic frequency amplitude ratio: Under different closing resistance conditions, the amplitude ratio of the fundamental wave to a specific harmonic shows a regular change and can be used as a sensitive indicator for condition identification. By focusing on the content of each harmonic (e.g., 2nd to 13th), harmonic analysis results under typical operating conditions can be obtained, such as... Figure 7 As shown, the Figure 7 (a) shows the harmonic content distribution of the closing resistor under normal operating conditions; Figure 7 (b) This figure shows the harmonic content distribution under short-circuit conditions of the closing resistor; Figure 7 (c) This figure shows the distribution of harmonic content under the condition of circuit breaking with closing resistor; it clearly shows the significant differences in the spectral characteristics under different closing resistor conditions, providing a distinguishable frequency domain basis for subsequent condition diagnosis.

[0063] In a preferred embodiment of the present invention, the K-nearest neighbor algorithm is used to match the aforementioned multi-dimensional features with a multi-condition closing resistor simulation database to obtain the top n closing resistor values ​​in similarity ranking, as well as the corresponding operating conditions, thereby achieving an assessment of the closing resistor status. Specifically, the core idea of ​​the K-nearest neighbor algorithm is based on the intuitive understanding of "like attracts like," that is, similar samples are close to each other in the feature space. The algorithm is built on two basic assumptions: the local consistency assumption (adjacent samples are likely to have the same category label) and the continuity of the feature space (sample similarity can be measured by distance in the feature space). Each current waveform sample is represented by 16 features, namely:

[0064]

[0065]

[0066] in, H i Indicates the sample waveform of the first i The amplitude of the second harmonic. THD Total harmonic distortion (THD) is used to assess the degree of distortion of the original signal by harmonic components. I m This indicates the peak value of the inrush current upon energization.

[0067] The similarity between waveforms is measured using Euclidean distance:

[0068]

[0069] in, H i The signal to be measured i Second harmonic amplitude THD A The total harmonic distortion of the signal under test. I mA The value represents the peak value of the waveform to be measured. The smaller the Euclidean distance, the more similar the two features are. When the distance is 0, the two features are completely identical.

[0070] After obtaining the Euclidean distances between the signal to be tested and the characteristics of each operating condition in the database, the distances are sorted, and the signal with the smallest distance is selected. K Given a sample, obtain the... K The average resistance value of each sample is taken as the resistance value of the signal under test corresponding to the operating condition. If K Values ​​that are too small: The model is too complex, prone to overfitting, and sensitive to noise; K Values ​​that are too large indicate that the model is too simple and may overlook important local patterns. This invention uses cross-validation to determine the appropriate value. K In this example, the optimal value is 5, meaning the 5 samples with the smallest distance are selected as the matching waveforms. For application to other working conditions, the value needs to be recalculated. The feature matching results are as follows: Figure 8 As shown, the feature matching results are presented from multiple dimensions. The harmonic feature comparison subplot presents the amplitude variation trend under different harmonic orders; the THD comparison subplot uses a bar chart to show the total harmonic distortion percentage of the test waveform and the five matching samples; the current peak value comparison subplot uses a bar chart to show the current peak value of the test waveform and the five matching samples; and the Euclidean distance subplot of the top 20 closest samples uses a bar chart to reflect the change in distance as the samples are sorted. These subplots comprehensively present the feature matching results of the K-nearest neighbor algorithm from dimensions such as harmonics, total harmonic distortion, current peak value, and Euclidean distance, providing multi-dimensional visualization support for the assessment of the closing resistor status.

[0071] In another preferred embodiment of the present invention, based on the above-mentioned method for evaluating the closing resistance status of a filter bank circuit breaker, this embodiment proposes a system for evaluating the closing resistance status of a filter bank circuit breaker, including a simulation module, a feature extraction module, and a status evaluation module, implemented as follows:

[0072] The simulation module is based on PSCAD / EMTP software. A refined PSCAD / EMTP model containing an AC filter and an 800kV SF6 tank-type circuit breaker is built. After verifying the model with measured normal and fault inrush current data, simulation is performed with the abnormal state of the closing resistance, the degree of degradation, and the 18° interval closing phase angle as variables to build a multi-condition simulation database.

[0073] The feature extraction module analyzes the transient process of the closing resistor being switched on and off, and uses a sudden change point detection algorithm to segment the waveform. It extracts the inrush current peak value and the switching time from the time domain, and extracts 16 multi-dimensional features such as harmonic components and THD from the frequency domain.

[0074] The state assessment module uses the K-nearest neighbor algorithm, which measures the similarity between the feature to be tested and the database samples through Euclidean distance. It selects K=5 of the most similar samples, calculates the average resistance value of the corresponding resistors, and outputs state assessment results such as normal, breakdown, or degradation of the closing resistor.

[0075] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for evaluating the closing resistance status of a filter bank circuit breaker, characterized in that, Includes the following steps: Based on the ±800kV UHVDC transmission project structure, a refined simulation model including AC filter, circuit breaker body and closing resistor was established in PSCAD / EMTP. The model was verified by inrush current data under normal and fault conditions collected in actual operation to ensure that it has good accuracy and robustness. On this basis, the abnormal state, degradation degree and closing phase angle of the closing resistor were used as variables, and a systematic simulation was carried out at 18° intervals within a sinusoidal cycle to build a closing resistor simulation database covering various operating conditions. Based on circuit theory, the transient process of the closing resistor during the two stages of connection and disconnection is analyzed. Circuit models are established for each stage to obtain the time-domain expressions of the corresponding stage currents. This theoretically reveals the intrinsic relationship between the closing resistor value, inrush current amplitude, and waveform. A sudden change point detection algorithm is used to intelligently segment the waveforms during the connection and disconnection stages of the closing resistor. The specific steps for intelligently segmenting the waveforms during the connection and disconnection stages using the sudden change point detection algorithm are as follows: Determine the closing time, and use the closing time as the starting point, with a window length of 0.1s, to capture the current waveform; Based on the captured current waveform, the signal difference is calculated, that is, the difference between adjacent data is calculated to obtain the gradient signal; Take 2% of the maximum value in the entire gradient signal as the threshold; Traverse the gradient signal and find the index position where the gradient value exceeds the threshold; The signals corresponding to the index positions that exceed the threshold are used as candidate index points, and the candidate points corresponding to the index positions are rearranged in chronological order. Based on the sorted candidate points, the first candidate point is selected as the initial mutation point. The minimum time interval between mutation points is set to 0.007 seconds. Subsequent candidate points are checked. If the time difference between a candidate point and the previous selected mutation point is ≥0.007 seconds, then the candidate point is selected as the second mutation point. The two mutation points correspond to the time when the closing resistor is put into operation and the time when it is taken out, respectively, so as to realize the intelligent segmentation of the waveform during the operation and exit stages. Multidimensional features of the transient process during the two stages of switching on and off are extracted from the time domain and frequency domain, respectively. By using the K-nearest neighbor algorithm, multi-dimensional features are matched with a multi-condition closing resistor simulation database to obtain the closing resistor values ​​of the top n in similarity ranking, as well as the corresponding operating conditions, thereby achieving the assessment of the closing resistor status.

2. The method for evaluating the closing resistance status of a filter bank circuit breaker according to claim 1, characterized in that, The model is validated using inrush current data collected during actual operation under normal and fault conditions. This is achieved by simulating normal, abnormal, and degraded operating conditions of the closing resistor. Specifically: abnormal operating conditions include breakdown, short circuit, and open circuit; under normal operating conditions, the inrush current waveform of the closing resistor at its rated resistance is simulated; under abnormal operating conditions, the closing process of the closing resistor under breakdown, short circuit, and open circuit fault states is simulated to obtain typical fault characteristic waveforms; under degraded operating conditions, the closing process of the closing resistor under different degrees of resistance drift is simulated to obtain inrush current characteristics reflecting the resistance degradation process; the different degrees of resistance drift include resistance reduction; by comparing and analyzing the consistency of the simulated data and measured inrush current data under the above operating conditions in terms of amplitude and waveform, a comprehensive verification of the model's accuracy and generalization ability is completed.

3. The method for evaluating the closing resistance status of a filter bank circuit breaker according to claim 1, characterized in that, The multi-condition closing resistor simulation database includes different operating conditions under normal and fault states, multi-dimensional feature data corresponding to different operating conditions, and closing resistor values ​​corresponding to different operating conditions.

4. The method for evaluating the closing resistance status of a filter bank circuit breaker according to claim 1, characterized in that, Multidimensional features of the transient process during the two stages of closing resistor connection and disconnection are extracted from the time domain. Specifically, the transient current exhibits inrush current peak value and closing resistor connection time characteristics in the time domain. Specifically, inrush current peak value: when the closing resistor value is normal, the inrush current peak value is effectively suppressed; when the resistor breaks down or its resistance value degrades, the inrush current peak value increases significantly. Closing resistor connection time: the connection time is stable under normal operating conditions; when the resistor is abnormal, the connection time is prolonged or shortened, reflecting changes in mechanical or electrical performance.

5. The method for evaluating the closing resistance status of a filter bank circuit breaker according to claim 1, characterized in that, Multidimensional features of the transient process during the two stages of closing the resistor being engaged and disengaged are extracted from the frequency domain. Specifically, the characteristics of the transient current in the frequency domain include the distribution of harmonic components and the characteristic frequency amplitude ratio. Specifically, the harmonic component distribution is as follows: under normal resistor conditions, the harmonic content is low and dominated by low-order harmonics; when the resistor breaks down or is open-circuited, the mid-to-high frequency harmonic components are significantly enhanced, and the spectral energy distribution shifts; the characteristic frequency amplitude ratio shows a regular change in the amplitude ratio of the fundamental wave to a specific harmonic under different closing resistor conditions, which can be used as a sensitive indicator for state identification.

6. The method for evaluating the closing resistance status of a filter bank circuit breaker according to claim 1, characterized in that, The K-nearest neighbor algorithm is as follows: through refined simulation and field tests, a large amount of known closing data of normal and fault states is collected, and their feature vectors are extracted; when a new sample to be tested is input, the K-nearest neighbor algorithm will calculate its Euclidean distance with each known sample in the database; the closer the distance, the more similar the waveform features.

7. A filter bank circuit breaker closing resistance status assessment system, based on the filter bank circuit breaker closing resistance status assessment method according to any one of claims 1 to 6, characterized in that, It includes a simulation module, a feature extraction module, and a state evaluation module; The simulation module is used to establish a refined PSCAD / EMTP simulation model of the filter field and to build a multi-condition closing resistor simulation database covering normal and fault states. The feature extraction module is used to analyze the transient process of the closing resistor being engaged and disengaged based on circuit theory, intelligently segment the waveform using a sudden change point detection algorithm, and extract multi-dimensional features of the transient process of the closing resistor being engaged and disengaged from the time domain and frequency domain respectively. The state assessment module is used to match multi-dimensional features with a multi-condition closing resistor simulation database using the K-nearest neighbor algorithm to obtain the closing resistor values ​​of the top n in similarity ranking, as well as the corresponding operating conditions, thereby realizing the state assessment of the closing resistor.

Images