Transformer on-load tap-changer testing method and testing device based on alternating current and direct current characteristics

Abstract

The invention discloses a transformer on-load tap-changer testing method and testing device based on alternating current and direct current characteristics, and relates to the technical field of power equipment detection.The testing method comprises the steps that in response to adjustable direct current voltage applied to an on-load tap-changer, direct current resistance values at all tapping positions are measured, and direct current resistance deviation is calculated; the method comprises the following steps: in response to a sweep frequency alternating current signal injected into an on-load tap-changer in a preset frequency range, acquiring voltage and current transient waveforms at the moment of switch action, and calculating a sweep frequency impedance phase angle and arc energy intensity and contact jitter variance in the transient waveforms; the direct current resistance deviation, the sweep frequency impedance phase angle, the arc energy characteristic and the contact jitter variance form a fusion characteristic vector; and inputting the fusion feature vector into a pre-trained deep belief network diagnosis model, and outputting fault probability distribution. According to the invention, the operation state information of the on-load tap-changer can be comprehensively obtained from multiple dimensions, and the fault of the on-load tap-changer of the transformer can be accurately diagnosed.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment detection, and in particular to a transformer on-load tap changer testing method and testing device based on AC and DC characteristics. Background Art

[0002] In modern power systems, transformers are critical electrical equipment, and the stability of their operating status is directly related to the reliability and safety of power supply. The transformer on-load tap-changer (OLTC) is a key component in transformers for voltage regulation. It can change the transformer's tap position without interrupting the load, achieving flexible voltage adjustment. However, during long-term operation, OLTCs are susceptible to various factors such as mechanical wear, electrical aging, and contact oxidation, which can lead to performance degradation and failures such as contact erosion, mechanical loosening, arc extinguishing failure, and spring fatigue, threatening the safe operation of the entire transformer system. Therefore, effective status monitoring and fault diagnosis of transformer on-load tap-changers are of great significance to ensuring the stable operation of power systems.

[0003] Traditional transformer on-load tap-changer testing methods primarily focus on single DC resistance measurements or AC withstand voltage tests. While these methods can provide some indication of the switch's performance, they have significant limitations. DC resistance measurements only reflect the static contact resistance of the contacts, making it difficult to effectively diagnose dynamic mechanical failures and insulation aging. AC withstand voltage tests, on the other hand, are primarily used to detect severe insulation defects and are unable to accurately identify early-stage faults and potential hazards. Traditional methods struggle to construct a feature vector that comprehensively reflects the switch's status, resulting in insufficient fault diagnosis accuracy and failing to meet the high-precision requirements of modern power systems for equipment status monitoring and fault warning. Summary of the Invention

[0004] To address the problem in existing technologies where it is difficult to construct a feature vector that can fully reflect the switch status, resulting in insufficient fault diagnosis accuracy, the present invention provides a transformer on-load tapchanger testing method and device based on AC / DC characteristics. This method can comprehensively obtain on-load tapchanger operating status information from multiple dimensions, enabling accurate diagnosis of transformer on-load tapchanger faults. The specific technical solution is as follows:

[0005] In a first aspect, the present invention provides a transformer on-load tap changer testing method based on AC / DC characteristics, comprising:

[0006] In response to an adjustable DC voltage applied to the on-load tap changer, the DC resistance value at each tap position is measured and the DC resistance deviation is calculated;

[0007] In response to injecting a swept frequency AC signal in a preset frequency range into the on-load tap changer, the system collects the voltage and current transient waveforms at the moment of switching action, and calculates the swept frequency impedance phase angle, arc energy intensity, and contact jitter variance in the transient waveform;

[0008] The DC resistance deviation, the swept frequency impedance phase angle, the arc energy characteristic and the contact jitter variance are used to form a fusion feature vector;

[0009] The fused feature vector is input into a pre-trained deep belief network diagnostic model to output a fault probability distribution.

[0010] Preferably, the pre-trained deep belief network diagnostic model is generated by training a historical fault sample library; the historical fault sample library includes a DC resistance anomaly data set, an AC impedance phase angle offset data set, a transient arc energy exceeding standard data set, and a contact jitter exceeding standard data set.

[0011] Preferably, the calculation process of the DC resistance deviation includes:

[0012] Perform multiple DC resistance measurements for each tap position and calculate the average resistance;

[0013] Calculate the DC resistance deviation based on the factory reference resistance and the average resistance of switches of the same model.

[0014] Preferably, the calculation process of the swept-frequency impedance phase angle includes:

[0015] Based on the swept frequency AC signal in the preset frequency range injected into the switch, a fast Fourier transform is performed on the steady-state response at different frequencies to calculate the complex impedance;

[0016] The swept frequency impedance phase angle at each frequency point is calculated according to the complex impedance.

[0017] Preferably, the calculation process of the arc energy intensity includes:

[0018] performing wavelet packet decomposition on the current transient waveform to extract frequency band energy of different frequencies;

[0019] The arc energy intensity is obtained by the ratio of the frequency band energy to the total energy.

[0020] Preferably, the calculation process of the contact jitter variance includes:

[0021] Identify the oscillation segment at the moment of contact separation from the voltage waveform;

[0022] The variance of the zero-crossing interval within the oscillation segment is calculated to obtain the contact jitter variance.

[0023] Preferably, the extraction of the contact jitter variance includes:

[0024] Identify contact action interval from voltage transient waveform;

[0025] Detect the time series of the waveform crossing the zero point, calculate the zero crossing interval, and calculate the contact jitter variance.

[0026] Preferably, after injecting a swept frequency AC signal within a preset frequency range into the on-load tap changer, the method further comprises:

[0027] The voltage-current dynamic trajectory is continuously drawn during the switching action, and the drawn trajectory is dynamically time-warped and matched with the standard trajectory library. An alarm is triggered when the trajectory deviation exceeds the preset alarm threshold.

[0028] Preferably, inputting the fused feature vector into a pre-trained deep belief network diagnostic model to output a fault probability distribution comprises:

[0029] Inputting the fused feature vector into the model input layer for normalization;

[0030] The normalized fusion feature vector is input into the three RBM layers of the model and then output through the Softmax layer to obtain the fault probability distribution.

[0031] In a second aspect, the present invention further provides a transformer on-load tap-changer testing device based on AC / DC characteristics, which applies the aforementioned transformer on-load tap-changer testing method based on AC / DC characteristics, comprising:

[0032] a DC steady-state test module for measuring the DC resistance value at each tap position in response to applying an adjustable DC voltage to the on-load tap changer and calculating the DC resistance deviation;

[0033] An AC transient test module is used to respond to a swept-frequency AC signal within a preset frequency range injected into the on-load tap-changer, collect the voltage and current transient waveforms at the moment of switching action, and calculate the swept-frequency impedance phase angle, as well as the arc energy intensity and contact jitter variance in the transient waveform;

[0034] a feature fusion module, configured to form a fusion feature vector from the DC resistance deviation, the swept frequency impedance phase angle, the arc energy feature, and the contact jitter variance;

[0035] The fault diagnosis module is used to input the fused feature vector into a pre-trained deep belief network diagnosis model and output a fault probability distribution.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] The AC / DC characteristics-based transformer on-load tap-changer testing method proposed in this paper comprehensively acquires multi-dimensional operating status information of the on-load tap-changer by combining DC resistance measurement, swept-frequency impedance phase angle analysis, arc energy intensity in transient waveforms, and contact jitter variance calculation. This operating status information can identify fault characteristics in dynamic processes such as mechanical loosening, arc extinguishing failure, and spring fatigue. By fusing these characteristic parameters from different dimensions into a vector and inputting it into a pre-trained deep belief network (DBN) diagnostic model, it accurately outputs fault probability distributions, enabling precise diagnosis and early warning of transformer on-load tap-changer faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly describes the drawings required for the specific embodiments or the description of the prior art. Similar elements or parts are generally identified by similar reference numerals throughout the drawings. Elements or parts in the drawings are not necessarily drawn to scale.

[0039] Figure 1 The figure is a flow chart of the transformer on-load tap changer testing method based on AC and DC characteristics of the present invention.

[0040] Figure 2 This is a schematic diagram of the transformer on-load tap changer testing system based on AC and DC characteristics of the present invention. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0042] It will be understood that when used in this specification and the appended claims, the terms “comprises” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0043] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0044] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0045] Unless otherwise specified, the raw materials used in the examples of this application are all commercially available.

[0046] Please refer to the following examples Figure 1 and Figure 2 .

[0047] The present application provides a method for testing a transformer on-load tap changer based on AC / DC characteristics, including:

[0048] Step S1, in response to an adjustable DC voltage applied to the on-load tap changer, measuring the DC resistance value at each tap position, and calculating the DC resistance deviation;

[0049] By selecting a DC resistance tester with an adjustable DC voltage output, resistance measurement requirements at different tap positions can be met. Existing resistance testers generally meet these requirements. In this embodiment, the TH2516B DC resistance tester, the three-channel WBL-BL3395A DC resistance tester, or the RK2512N / RK2512N+ DC low resistance tester can be used.

[0050] In a specific implementation, the output end of the DC resistance tester is connected to the designated terminal of the transformer on-load tap changer, and the stable transmission of current is measured during the measurement process.

[0051] Specifically, the calculation process of the DC resistance deviation includes:

[0052] Turn on the DC resistance tester and apply an adjustable DC voltage to the on-load tap changer. Starting from the lowest tap position, adjust the switch to each tap position in sequence. Under the DC steady-state test environment, perform three DC resistance measurements for each tap position to eliminate the influence of contact noise. The corresponding data for each measurement is recorded as

[0053] Calculate the resistance average of 3 measurements

[0054]

[0055] At the same time, the factory reference resistance value R corresponding to the same type of switch is retrieved from the historical database. base .

[0056] Calculation of DC resistance deviation ΔR dc :

[0057]

[0058] It should be noted that the DC resistance deviation value ΔR dc It can be used to diagnose whether there is oxidation or ablation of the contacts and serious contact failures, ΔR dc Associated contact erosion / oxidation type fault.

[0059] Step S2: In response to injecting a swept frequency AC signal within a preset frequency range into the on-load tap changer, collecting the voltage and current transient waveforms at the moment of switching operation, and calculating the swept frequency impedance phase angle, arc energy intensity, and contact jitter variance in the transient waveforms;

[0060] Select a signal generator capable of generating a swept-frequency AC signal within a preset frequency range (0.1Hz-20kHz). Existing resistance testers can generally meet these requirements. In this embodiment, the TH1312-20 / TH1312-100 audio swept-frequency signal generator, the CRY5520 audio swept-frequency signal generator, or the GVS7116 automatic swept-frequency signal generator can be used.

[0061] In the specific implementation, the signal source and acquisition equipment are connected to the corresponding test port of the on-load tap changer to inject the sweep frequency signal and capture the voltage and current transient waveforms at the moment of switch action.

[0062] Specifically, the calculation process of the swept-frequency impedance phase angle includes:

[0063] Start the AC sweep frequency signal source and inject a swept frequency AC signal in the frequency range of 0.1Hz-20kHz into the on-load tap changer. Simultaneously, measure the voltage u(t) and current i(t).

[0064] For a specific frequency f k The steady-state response of is subjected to fast Fourier transform (FFT) to obtain:

[0065]

[0066] Where U k and I k Respectively represent the frequency f k At , the Fourier transform results of the voltage signal u(t) and the current signal i(t).

[0067] and through Calculate the complex impedance.

[0068] By scanning the entire frequency band, the swept impedance phase angle at each frequency point is obtained and the output θ(f k )=arg(Z k ) corresponding frequency response curve.

[0069] It should be noted that the swept-frequency impedance phase angle can be used to identify faults such as mechanical looseness and insulation degradation.

[0070] Specifically, the calculation process of the arc energy intensity includes:

[0071] At the moment of switch action, that is, when the contacts are separated or closed, the current i(t) is collected at a sampling frequency of not less than 5 MHz;

[0072] Perform wavelet packet decomposition on i(t) to obtain the energy E of different frequency bands j (j represents the node number).

[0073] Calculate the sum of the energy within the arc extinguishing characteristic frequency band (3kHz-100kHz), that is, E band =∑ j∈Ω E j , where Ω is defined as {j|3kHz≤f j ≤100kHz}.

[0074] Calculate the arc energy intensity E arc

[0075]

[0076] It should be noted that the arc energy intensity is used to determine whether the arc extinguishing chamber has failed and whether there is a risk of arc restrike.

[0077] Specifically, the calculation process of the contact jitter variance includes:

[0078] Identify the contact action interval from the voltage transient waveform, set as t∈[t0,t0+ΔT];

[0079] Detect a series of time points at which the waveform passes through zero to form a time series

[0080] pass (i=1,2,…,N-1) calculate the time interval between adjacent zero crossing points;

[0081] Calculate the contact jitter variance σ 2 :

[0082]

[0083] in,

[0084] It should be noted that the contact jitter variance is used to diagnose whether the spring is fatigued and whether the machine has jamming problems.

[0085] Among them, θ(f k) reflects the frequency domain phase shift characteristics and is related to mechanical looseness / insulation degradation; E arc The problem of associated arc extinguishing failure is obtained by transient waveform analysis; σ 2 It reflects the time dispersion of mechanical action and can be used to diagnose spring fatigue / transmission mechanism abnormalities. It is also obtained based on transient waveform analysis.

[0086] Step S3, forming a fusion feature vector by combining the DC resistance deviation, the swept frequency impedance phase angle, the arc energy characteristic and the contact jitter variance;

[0087] The fused feature vector x is generated by fusing the DC resistance deviation, the swept frequency impedance phase angle, the arc energy feature and the contact jitter variance. raw :

[0088]

[0089] Z-score normalization is used to eliminate dimensional differences:

[0090]

[0091] where μ i and δ i are the mean and standard deviation of the i-th dimension features of the training set, respectively.

[0092] Frequency domain feature expansion The swept impedance phase angle spectrum θ(f) is a frequency domain vector and needs to be expanded as:

[0093] x θ =[θ(f1)θ(f2)…θ(f m )] T ∈R m

[0094] Where m is the number of frequency sweep points, and m = 20 corresponds to the logarithmic average division of 0.1 Hz to 20 kHz.

[0095] Step S4: input the fused feature vector into a pre-trained deep belief network diagnostic model and output a fault probability distribution.

[0096] Specifically, the pre-trained deep belief network diagnostic model is generated by training a historical fault sample library; the historical fault sample library includes a DC resistance anomaly data set, an AC impedance phase angle offset data set, a transient arc energy exceeding standard data set, and a contact jitter exceeding standard data set.

[0097] The constructed fused feature vector is input into a pre-trained DBN diagnostic model, which analyzes the correlation between the feature vector and various fault modes and ultimately outputs a fault probability distribution. For example, the model output includes the probability of contact erosion / oxidation, mechanical loosening, arc extinguishing failure, and spring fatigue. Based on the output of these fault probabilities, operators can determine the risks associated with the on-load tap-changer and conduct timely inspections.

[0098] Specifically, inputting the fused feature vector into a pre-trained deep belief network diagnostic model and outputting a fault probability distribution includes:

[0099] Inputting the fused feature vector into the model input layer for normalization;

[0100] The input layer vector constructs the final input DBN vector as a normalized scalar + frequency domain vector, including 23 neurons, expressed as:

[0101]

[0102] Where, is the standardized DC resistance deviation; θ(f1)~θ(f 20 ) is the impedance phase angle of 20 sweep frequency points; is the normalized arc energy peak; is the normalized contact jitter variance.

[0103] The normalized fusion feature vector is input into the three RBM layers of the model and then output through the Softmax layer to obtain the fault probability distribution.

[0104] The network structure of the deep belief network diagnosis model includes three RBM layers as hidden layers and one output layer.

[0105] The three RBM layers are hidden layers as follows:

[0106] First layer RBM: The number of nodes is set to 120, Gaussian-binary RBM is used to process continuous features, and the activation function is Sigmoid, which is expressed as:

[0107]

[0108] Among them, h represents the activation state of the hidden layer neurons, represents the jth hidden unit in the first layer; v represents the input vector of the input layer, v i represents the i-th element in the input vector; b represents the bias term, represents the bias of the jth hidden unit in the first layer; w represents the weight, represents the weight from input unit i to hidden unit j in the first layer.

[0109] The second layer RBM: The number of nodes is set to 60, using binary-binary RBM, and the activation function is Sigmoid, expressed as:

[0110]

[0111] Second-layer RBM: The number of nodes is set to 30, and binary-binary RBM is used to extract abstract fault patterns. The activation function is Sigmoid, which is expressed as:

[0112]

[0113] It should be noted that the Sigmoid activation function is used to map the input value to the (0, 1) interval and is a commonly used activation function in RBM.

[0114] The output layer consists of four neurons, the activation function is Softmax, and the output vector.

[0115] Among them, the failure probability calculation P k as follows:

[0116]

[0117] Where w s,k and b s,k They represent the weight vector and bias of the k-th output node in the Softmax layer respectively.

[0118] Confidence Assessment:

[0119]

[0120] When Confidence<0.85, a warning indicating uncertain diagnosis is output.

[0121] By calculating the confidence level of the diagnostic results, the reliability of the fault probability distribution output by the model can be evaluated. When the confidence level is lower than the set threshold of 0.85, an uncertainty warning is issued in a timely manner to remind operation and maintenance personnel to make prudent decisions, effectively avoiding incorrect maintenance operations caused by model misjudgment.

[0122] The final output vector is represented as:

[0123]

[0124] Among them, P1 represents the probability of contact ablation / oxidation; P2 represents the probability of mechanical loosening; P3 represents the probability of arc extinguishing failure; and P4 represents the probability of spring fatigue.

[0125] In this embodiment, multiple key characteristics of the transformer on-load tap-changer are comprehensively considered, including static DC resistance deviation, dynamic swept-frequency impedance phase angle, arc energy and contact jitter during transient processes, etc., to comprehensively reflect the operating status of the switch from different angles, avoiding the one-sided diagnosis caused by a single feature.

[0126] By analyzing the fused feature vectors using a deep belief network (DBN), the system accurately classifies fault types and quantifies their probabilities, providing operators with more accurate fault information. The final output includes probability distribution vectors for four specific fault types (contact erosion / oxidation, mechanical looseness, arc extinguishing failure, and spring fatigue). This provides operators with detailed equipment health status information, reducing downtime and repair costs.

[0127] It should be noted that the training of the pre-trained deep belief network diagnostic model is as follows:

[0128] In the first stage, unsupervised pre-training (layer-by-layer greedy training) is used to train the unlabeled test data v in , RBM parameter update (taking the first layer as an example):

[0129] Positive phase:

[0130] Inversion: Reconstruct data v recon

[0131] Weight update: Where ∈ = 0.01 (learning rate)

[0132] The second stage uses supervised fine-tuning, which is trained through backpropagation + Adam optimizer as follows:

[0133] Network construction: stack the pre-trained 3-layer RBM+Softmax output layer, freeze the weights of the first layer RBM1 and the second layer RBM2, and only train the third layer RBM3 and Softmax layer

[0134] The loss function is as follows:

[0135]

[0136] Back propagation, calculate the output layer gradient:

[0137]

[0138] Calculate the gradient of the third layer RBM3:

[0139]

[0140] Among them, z (3) =W (3) h(2) +b( 3) .

[0141] Parameter update (Adam optimizer):

[0142]

[0143] Here, η=0.001 and batch size=325.

[0144] Through the layered training mechanism, unsupervised pre-training solves the problem of sample scarcity, and supervised fine-tuning optimizes the classification boundaries.

[0145] Specifically, in a preferred embodiment, after injecting a swept frequency AC signal within a preset frequency range into the on-load tap changer, the method further includes:

[0146] The voltage-current dynamic trajectory is continuously drawn during the switching action, and the drawn trajectory is dynamically time-warped and matched with the standard trajectory library. An alarm is triggered when the trajectory deviation exceeds the preset alarm threshold.

[0147] During implementation, a swept-frequency AC signal within a preset frequency range is injected into the on-load tap changer. The voltage-current (VI) dynamic trajectory is continuously plotted during the switching operation. Using the Dynamic Time Warping (DTW) algorithm, the real-time VI trajectory is matched with trajectories in a pre-built standard trajectory library. The DTW algorithm handles nonlinear alignment of time series data and evaluates the similarity between trajectories by calculating the shortest distance. When the trajectory deviation exceeds a preset alarm threshold, the system automatically triggers an alarm.

[0148] By monitoring changes in the VI trajectory during switch operation in real time and comparing it with the standard trajectory, anomalies can be detected promptly. The application of the DTW algorithm improves the accuracy and robustness of trajectory matching, enabling the system to adapt to varying movement speeds and time delays. When a trajectory deviates significantly from the standard trajectory, a timely alarm alerts operators to conduct inspections and maintenance, preventing potential failures.

[0149] In a second aspect, the present invention further provides a transformer on-load tap-changer testing device based on AC / DC characteristics, which applies the aforementioned transformer on-load tap-changer testing method based on AC / DC characteristics, comprising:

[0150] a DC steady-state test module for measuring the DC resistance value at each tap position in response to applying an adjustable DC voltage to the on-load tap changer and calculating the DC resistance deviation;

[0151] An AC transient test module is used to respond to a swept-frequency AC signal within a preset frequency range injected into the on-load tap-changer, collect the voltage and current transient waveforms at the moment of switching action, and calculate the swept-frequency impedance phase angle, as well as the arc energy intensity and contact jitter variance in the transient waveform;

[0152] a feature fusion module, configured to form a fusion feature vector from the DC resistance deviation, the swept frequency impedance phase angle, the arc energy feature, and the contact jitter variance;

[0153] The fault diagnosis module is used to input the fused feature vector into a pre-trained deep belief network diagnosis model and output a fault probability distribution.

[0154] The beneficial effects of this embodiment are the same as those of the transformer on-load tap-changer testing method based on AC / DC characteristics, and will not be repeated here.

[0155] Those skilled in the art will appreciate that the modules of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0156] In the embodiments provided by the present invention, it should be understood that the division of modules is merely a logical function division, and there may be other division methods in actual implementation, for example, multiple modules can be combined into one module, one module can be split into multiple modules, or some features can be ignored, etc.

[0157] In addition, the functional modules in various embodiments of the present invention may be integrated into a single processing module, or each module may exist physically separately, or two or more modules may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules.

[0158] If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-0nly Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc., various media that can store program code.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the specification of the present invention.

Claims

1. A transformer on-load tap changer testing method based on AC / DC characteristics, characterized in that: include: In response to an adjustable DC voltage applied to the on-load tap changer, the DC resistance value at each tap position is measured and the DC resistance deviation is calculated; In response to injecting a swept frequency AC signal in a preset frequency range into the on-load tap changer, the system collects the voltage and current transient waveforms at the moment of switching action, and calculates the swept frequency impedance phase angle, arc energy intensity, and contact jitter variance in the transient waveform; The DC resistance deviation, the swept frequency impedance phase angle, the arc energy characteristic and the contact jitter variance are used to form a fusion feature vector; The fused feature vector is input into a pre-trained deep belief network diagnostic model to output a fault probability distribution.

2. The transformer on-load tap changer testing method based on AC / DC characteristics according to claim 1, characterized in that: The pre-trained deep belief network diagnostic model is generated by training a historical fault sample library; the historical fault sample library includes a DC resistance anomaly data set, an AC impedance phase angle offset data set, a transient arc energy exceeding standard data set, and a contact jitter exceeding standard data set.

3. The transformer on-load tap changer testing method based on AC / DC characteristics according to claim 2, characterized in that: The calculation process of the DC resistance deviation includes: Perform multiple DC resistance measurements for each tap position and calculate the average resistance; Calculate the DC resistance deviation based on the factory reference resistance and the average resistance of switches of the same model.

4. The transformer on-load tap changer testing method based on AC / DC characteristics according to claim 2, characterized in that: The calculation process of the swept-frequency impedance phase angle includes: Based on the swept frequency AC signal in the preset frequency range injected into the switch, a fast Fourier transform is performed on the steady-state response at different frequencies to calculate the complex impedance; The swept frequency impedance phase angle at each frequency point is calculated according to the complex impedance.

5. The transformer on-load tap changer testing method based on AC / DC characteristics according to claim 2, characterized in that: The calculation process of the arc energy intensity includes: performing wavelet packet decomposition on the current transient waveform to extract frequency band energy of different frequencies; The arc energy intensity is obtained by the ratio of the frequency band energy to the total energy.

6. The transformer on-load tap changer testing method based on AC / DC characteristics according to claim 2, characterized in that: The calculation process of the contact jitter variance includes: Identify the oscillation segment at the moment of contact separation from the voltage waveform; The variance of the zero-crossing interval within the oscillation segment is calculated to obtain the contact jitter variance.

7. The transformer on-load tap-changer testing method based on AC / DC characteristics according to claim 2, characterized in that: The extraction of the contact jitter variance includes: Identify contact action interval from voltage transient waveform; Detect the time series of the waveform crossing the zero point, calculate the zero crossing interval, and calculate the contact jitter variance.

8. The transformer on-load tap-changer testing method based on AC / DC characteristics according to claim 1, characterized in that: After injecting a swept frequency AC signal within a preset frequency range into the on-load tap changer, the method further comprises: The voltage-current dynamic trajectory is continuously drawn during the switching action, and the drawn trajectory is dynamically time-warped and matched with the standard trajectory library. An alarm is triggered when the trajectory deviation exceeds the preset alarm threshold.

9. The transformer on-load tap changer testing method based on AC / DC characteristics according to claim 1, characterized in that: Inputting the fused feature vector into a pre-trained deep belief network diagnostic model and outputting a fault probability distribution includes: Inputting the fused feature vector into the model input layer for normalization; The normalized fusion feature vector is input into the three RBM layers of the model and then output through the Softmax layer to obtain the fault probability distribution.

10. A transformer on-load tap-changer test device based on AC / DC characteristics, characterized in that: The transformer on-load tap changer testing method based on AC / DC characteristics according to any one of claims 1 to 7 comprises: a DC steady-state test module for measuring the DC resistance value at each tap position in response to applying an adjustable DC voltage to the on-load tap changer and calculating the DC resistance deviation; An AC transient test module is used to respond to a swept-frequency AC signal within a preset frequency range injected into the on-load tap-changer, collect the voltage and current transient waveforms at the moment of switching action, and calculate the swept-frequency impedance phase angle, as well as the arc energy intensity and contact jitter variance in the transient waveform; a feature fusion module, configured to form a fusion feature vector from the DC resistance deviation, the swept frequency impedance phase angle, the arc energy feature, and the contact jitter variance; The fault diagnosis module is used to input the fused feature vector into a pre-trained deep belief network diagnosis model and output a fault probability distribution.

Images