On-load tap-changing switch life prediction method and system

Abstract

The present invention provides a method and system for predicting the electrical life of an on-load tap-changer. The prediction system includes a breaking test unit, a modeling unit, and a life prediction unit. The corresponding prediction method is specifically as follows: based on three-phase cross-interconnection, breaking test data is obtained, and an incremental sequence of contact resistance values ​​is calculated, and the degradation process of the contact resistance increment is simulated based on the Wiener process; according to the simulated degradation process of the contact resistance increment, an electrical life distribution model is established, and the electrical life prediction result of the on-load tap-changer is obtained through Bayesian parameter estimation. The present invention can adopt a three-phase cross-interconnection electrical life detection method to obtain corresponding breaking test data, reduce the deviation between the collected parameters and the actual operating conditions, and construct an electrical life distribution model under the action of multi-stress coupling by simulating the degradation process of the contact resistance increment, and further combine it with Bayesian parameter estimation to solve it, thereby ensuring prediction accuracy while shortening the test cycle.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment life prediction, and in particular to a method and system for predicting the life of an on-load tap-changing switch. Background Art

[0002] In the process of building a new power system with a high proportion of renewable energy access, on-load tap-changing distribution transformers, with their millisecond-level load regulation capabilities, are becoming a key component in improving distribution network voltage compliance and power supply reliability. However, on-load tap-changers, as their core actuators, must frequently perform winding tap-switching operations under load. This exposes the contact system to long-term arc erosion, mechanical wear, and thermal stress, resulting in significant electrical life degradation. To ensure the proper operation of on-load tap-changing distribution transformers, accurate prediction of the remaining life of the on-load tap-changers is necessary, allowing for timely repair and replacement.

[0003] Predicting the remaining life of an on-load tap changer typically requires establishing a mapping relationship between tap changer switching operations and the degree of equipment degradation. When capturing this mapping relationship through detection operations, obtaining accurate mechanical life test results over the entire life cycle requires at least tens of thousands of mechanical operations, resulting in a lengthy detection cycle. Furthermore, because the tap changer's tap changer switching process involves a combined mechanical-electrical mechanism, existing offline detection methods are limited by the complexity of the tap changer's multi-contact linkage mechanism and the dynamic switching characteristics of the process. This makes it difficult to construct a dynamic electrical stress loading system equivalent to actual operating conditions, and single-phase simulation platforms are unable to reproduce the electromagnetic coupling characteristics of three-phase systems. Consequently, key parameters such as the contact erosion rate and contact resistance growth pattern obtained during the detection process deviate significantly from field operating data. This makes it difficult to accurately simulate the degradation trajectory of contact materials under alternating electrothermal stress, resulting in low accuracy in life prediction. Summary of the Invention

[0004] The present invention aims to overcome the shortcomings of the prior art, which require at least tens of thousands of mechanical operations to collect corresponding samples, resulting in long test cycles, high testing costs, significant deviations between the parameters obtained during the testing process and field operating data, making it difficult to accurately simulate degradation trajectories and resulting in low accuracy in life prediction. A method and system for predicting the electrical life of an on-load tap changer are provided. The method uses a three-phase cross-connected electrical life detection method to obtain corresponding interruption test data. This method can effectively reproduce the voltage gradient, circulating current impact, and arc energy distribution characteristics of the tap changer in actual grid operation, reducing the deviation between the collected parameters and actual operating conditions and improving the accuracy of subsequent life prediction. Furthermore, based on this, the degradation process of the contact resistance increment is simulated based on the collected interruption test data to establish a nonlinear mapping relationship between the contact resistance increment and the number of operations, thereby realizing the construction of an electrical life distribution model under the action of multiple stress coupling. Furthermore, the method combines prior knowledge and the collected test data with a Bayesian framework to achieve electrical life prediction under small sample conditions, eliminating the need for full life cycle testing operations, shortening the test cycle, and reducing testing costs while ensuring prediction accuracy.

[0005] The purpose of the present invention is achieved through the following technical solutions.

[0006] The life prediction method of on-load tap-changer includes:

[0007] Conduct electrical life testing of on-load tap changers based on three-phase cross-connection and obtain breaking test data;

[0008] Based on the breaking test data, the incremental sequence of contact resistance values ​​is calculated;

[0009] Based on the Wiener process and combined with the incremental sequence of contact resistance values, the degradation process of contact resistance increment is simulated;

[0010] According to the degradation process of simulated contact resistance increment, an electrical life distribution model is established;

[0011] Based on the electrical life distribution model, the electrical life prediction results of the on-load tap changer are obtained through Bayesian parameter estimation.

[0012] Furthermore, the on-load tap changer electrical life detection based on three-phase cross-connection and obtaining breaking test data includes:

[0013] The on-load tap changers are grouped and interconnected, and the voltage level tap changers corresponding to the on-load tap changers are reconstructed;

[0014] Based on the phase setting of the test power supply, a test circuit is established with the reconstructed voltage range tap changer;

[0015] Connect the test circuit to the test transformer, perform a breaking test by switching the voltage position tap changer, and collect breaking test data for each breaking test.

[0016] Furthermore, the Wiener process is combined with the incremental sequence of the contact resistance value to simulate the degradation process of the contact resistance increment, including:

[0017] Based on the incremental sequence of contact resistance values, independence check and normal distribution check are performed on the contact resistance increments;

[0018] Set the drift coefficient and diffusion coefficient based on the contact resistance increment that passes the calibration;

[0019] The degradation process of the contact resistance increment is simulated based on the drift coefficient and diffusion coefficient.

[0020] Furthermore, the electrical life distribution model is established according to the degradation process of the contact resistance increment, including:

[0021] Set the failure threshold of the on-load tap changer. Based on the first arrival time theory, the time it takes to reach the failure threshold for the first time is used as the electrical life identification parameter.

[0022] Combined with the degradation process of contact resistance increment, the probability density function of electrical life identification parameters is set and the electrical life distribution model is established.

[0023] Furthermore, the electrical life prediction result of the on-load tap changer is obtained by Bayesian parameter estimation based on the electrical life distribution model, including:

[0024] Set the prior distribution of drift coefficient and diffusion coefficient;

[0025] Based on the prior distribution, posterior sampling is performed in combination with the incremental sequence of the calibrated contact resistance values ​​to obtain the posterior means of the drift coefficient and the diffusion coefficient;

[0026] Based on the posterior mean, the electrical life distribution model is solved to obtain the predicted value of the electrical life of the on-load tap changer.

[0027] Furthermore, it also includes:

[0028] When solving the electrical life distribution model, the Mahalanobis distance detection index is also used to identify outliers in the solution results.

[0029] Furthermore, the setting of the prior distribution of the drift coefficient and the diffusion coefficient includes:

[0030] Identify the type of on-load tap changer and obtain historical operating data of the same type of on-load tap changer;

[0031] Based on historical operating data, the prior distribution of the drift coefficient and diffusion coefficient is set.

[0032] Furthermore, the step of calculating the incremental sequence of contact resistance values ​​based on the breaking test data includes:

[0033] Based on the breaking test data, the contact resistance value of the on-load tap changer after each breaking test is obtained;

[0034] Based on the contact resistance values ​​of adjacent breaking tests, an incremental sequence of contact resistance values ​​is constructed.

[0035] Furthermore, it also includes:

[0036] A recursive Bayesian update formula is set, and the electrical life prediction results of the on-load tap changer are corrected in real time based on the recursive Bayesian update formula.

[0037] The on-load tap-changer life prediction system includes:

[0038] The breaking test unit is used to establish the breaking test wiring of the on-load tap changer and collect the breaking test data during the breaking test;

[0039] The modeling unit is connected to the breaking test unit and is used to simulate the degradation process of the contact resistance increment according to the breaking test data and establish an electrical life distribution model accordingly;

[0040] The life prediction unit is connected to the modeling unit and is used to solve the electrical life distribution model according to Bayesian estimation to obtain the electrical life prediction result of the on-load tap-changer.

[0041] The beneficial effects of the present invention are:

[0042] A three-phase cross-connected electrical life test method is used to obtain corresponding interruption test data. This method utilizes the principle of three-phase power exchange and can simulate the dynamic operating conditions of on-load tap changers in three-phase systems. This accurately reproduces the voltage gradient, circulating current surge, and arc energy distribution characteristics of the corresponding tap changers in actual grid operation. This reduces the deviation between the collected data during the test and the actual operating conditions, and improves the accuracy of subsequent life prediction. Furthermore, a Wiener process is proposed to simulate the degradation process of the contact resistance increment based on the collected interruption test data. This establishes a nonlinear mapping relationship between the contact resistance increment and the number of operations, thereby constructing an electrical life distribution model under multiple stress coupling. Furthermore, a Bayesian framework is used to integrate prior knowledge and collected test data to achieve electrical life prediction. This eliminates the need for full life cycle testing operations. High-precision electrical life prediction results can be achieved using small sample data obtained through a small number of test operations, effectively shortening the test cycle and reducing testing costs while ensuring prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of a process of the present invention;

[0044] Figure 2 This is a wiring diagram of a detection circuit constructed in an embodiment of the present invention;

[0045] Figure 3 The present invention is a schematic structural diagram of an on-load tap-changer electrical life prediction system.

[0046] Among them: 1. Breaking test unit, 2. Modeling unit, 3. Life prediction unit. DETAILED DESCRIPTION

[0047] The present invention will be further described below with reference to the accompanying drawings and examples.

[0048] Example: The tap-changing process of an on-load tap changer involves a combined mechanical and electrical mechanism. During tap-changing, the moving contact must bridge adjacent tap positions using transition resistors to suppress circulating current shock while maintaining continuous main circuit power. This process utilizes a vacuum interrupter to achieve millisecond-level arc interruption. During this process, the moving and static contacts of the on-load tap changer must withstand the three-phase coupled stresses of inter-position voltage difference, transient circulating current, and arc thermal erosion. However, traditional single-phase testing platforms for on-load tap changer life testing can only simulate a single electrical stress and cannot replicate the 120° phase difference effect of the three-phase power supply. The complexity of the multi-contact linkage mechanism of the tap changer makes it difficult to replicate the dynamic interaction of "mechanical wear + arc erosion" during offline testing using single-phase testing platforms, resulting in significant deviations from field operating parameters.

[0049] Furthermore, the design life of a switch is typically over 10,000 operations. Traditional testing requires mechanical operations that cover 100% of the design life. This results in a long test cycle and the simultaneous monitoring of multiple parameters, resulting in extremely high equipment wear and labor costs. After completing the relevant life test operations and obtaining the relevant life data, life prediction is often performed using point estimation methods such as maximum likelihood estimation. However, this method relies on a large sample size of data covering the entire life cycle. Reducing the sample size can lead to an inability to effectively integrate historical experience or prior knowledge, unstable parameter estimation, and low accuracy in life prediction results.

[0050] In order to solve the various problems existing in the above-mentioned on-load tap-changing switch electrical life prediction means, this embodiment proposes an on-load tap-changing switch electrical life prediction method, such as Figure 1 Shown, including:

[0051] Conduct electrical life testing of on-load tap changers based on three-phase cross-connection and obtain breaking test data;

[0052] Based on the breaking test data, the incremental sequence of contact resistance values ​​is calculated;

[0053] Based on the Wiener process and combined with the incremental sequence of contact resistance values, the degradation process of contact resistance increment is simulated;

[0054] According to the degradation process of simulated contact resistance increment, an electrical life distribution model is established;

[0055] Based on the electrical life distribution model, the electrical life prediction results of the on-load tap changer are obtained through Bayesian parameter estimation.

[0056] First, the principle of three-phase power exchange is used to realize the dynamic operating condition simulation of the on-load tap-changer in the three-phase system, so as to accurately reproduce the voltage gradient, circulating current impact and arc energy distribution characteristics of the tap changer corresponding to the on-load tap-changer in the actual operation of the power grid, so as to obtain the breaking test data required for life prediction.

[0057] Specifically, the on-load tap changer electrical life detection based on three-phase cross-connection and obtaining breaking test data includes:

[0058] The on-load tap changers are grouped and interconnected, and the voltage level tap changers corresponding to the on-load tap changers are reconstructed;

[0059] Based on the phase setting of the test power supply, a test circuit is established with the reconstructed voltage range tap changer;

[0060] Connect the test circuit to the test transformer, perform a breaking test by switching the voltage position tap changer, and collect breaking test data for each breaking test.

[0061] Taking the 9-speed tap changer of a 400kVA on-load tap changer with a rated voltage of 10kV and a short-circuit impedance of 4% as an example, by reconstructing the three-phase connection topology of the tap changer and utilizing the phase difference and amplitude characteristics of the three-phase power supply, dynamic voltage gradients and transient circulating currents equivalent to those in actual grid operation are synthesized under laboratory conditions. At the same time, the short-circuit impedance characteristics of the transformer are simulated through a closed-loop circuit to achieve precise control of arc energy distribution.

[0062] In the process of reconstructing the three-phase connection topology of the tap changer, the 120° phase difference of the three-phase power supply can be used to construct an equivalent inter-stage voltage to generate a dynamic voltage gradient. By synchronously applying the rated current and dynamic voltage, the coupling effect of contact arc erosion and mechanical wear is simulated to achieve the loading of current-voltage composite stress and complete the dynamic operating condition simulation of the on-load tap changer in the three-phase system.

[0063] Among them, for the reconstruction of the three-phase wiring topology of the tap changer, the voltage level tap changers of the on-load tap changer are first physically reconstructed. Specifically, the odd-numbered levels are first integrated, and the 1, 3, 5, 7, and 9 level taps are welded into the first group of common terminals through copper bars. Then the even-numbered levels are integrated, and the 2, 4, 6, and 8 level taps are welded into the second group of common terminals. Epoxy resin insulating partitions are installed between the first and second groups of common terminals to ensure electrical isolation between the level groups and achieve insulation reinforcement.

[0064] Then, based on the phase configuration of the test power supply, the test loop is established. The test power supply should be an ABC three-phase power supply, and the established test loop includes an A-phase loop, a B-phase loop, and a C-phase loop.

[0065] The A-phase circuit consists of the test power supply phase A, the test power supply phase B, the first group of common terminals and the second group of common terminals. The first group of common terminals is connected to the test power supply phase A with a phase angle of 0°, and the second group of common terminals is connected to the test power supply phase B with a phase angle of -120°.

[0066] The B-phase circuit consists of the test power supply B-phase, the test power supply C-phase, the first group of common terminals and the second group of common terminals. The first group of common terminals is connected to the test power supply B-phase with a phase angle of -120°, and the second group of common terminals is connected to the test power supply C-phase with a phase angle of 120°.

[0067] The C-phase circuit consists of the test power supply C-phase, the test power supply A-phase, the first group of common terminals and the second group of common terminals. The first group of common terminals is connected to the test power supply C-phase with a phase angle of 120°, and the second group of common terminals is connected to the test power supply A-phase with a phase angle of 0°.

[0068] According to the above wiring principle, the three phases of the test power supply are connected to the taps of the voltage level tap changers of the on-load tap changer according to the cross wiring.

[0069] Finally, the test circuit is connected to the test transformer. The overall detection circuit is constructed by connecting the output terminals D / y of the test circuit to the high-voltage terminals of the test transformer and then short-circuiting the low-voltage terminals of the test transformer.

[0070] After the construction is completed, the detection circuit wiring is as follows Figure 2 shown.

[0071] After the detection circuit is connected, the test current is set. In this embodiment, the test current is set to a rated value of 23A, and the test voltage is set according to the short-circuit impedance characteristics. .

[0072] Test voltage The calculation method is:

[0073] ;

[0074] in, is the rated voltage, is the short-circuit impedance percentage.

[0075] Under the test current and test voltage test environment, taking the A phase 1 gear switching to 2 gear as an example, the voltage gradient changes for:

[0076] .

[0077] It can be seen that the three-phase cross-interconnection topology achieved through the 120° phase difference of the ABC three-phase power supply can ensure that the voltage gradient when the adjacent gears of the reproduced on-load tap-changer are switched can be close to the actual working conditions. At the same time, by synchronously applying the rated current of 23A and the dynamic voltage gradient of 692V, the coupling effect of contact arc erosion and mechanical wear can be more accurately restored, ensuring the accuracy of the subsequent disconnection test data collected during the disconnection test by opening the voltage gear tap changers of the on-load tap-changer.

[0078] Considering that different types of on-load tap changers have different design lifespans, the sample data required to support electrical life prediction also varies. To reduce the number of tests while ensuring the accuracy of electrical life prediction, the following are also required:

[0079] Obtain device information of the on-load tap-changing switch, and obtain the design life of the on-load tap-changing switch based on the device information;

[0080] The number of breaking tests for the electrical life detection of the on-load tap-changer shall be set according to the design life of the on-load tap-changer.

[0081] In this embodiment, 15% of the breaking times corresponding to the design life is used as the breaking test times required to be performed in each test.

[0082] After completing the breaking test to obtain breaking test data to support subsequent electrical life prediction, the increment of the contact resistance of the on-load tap changer contacts can be regarded as a random process with drift characteristics based on the obtained breaking test data. The Wiener process is used to describe its degradation trajectory. Combined with the principle that when the contact resistance increment reaches the failure threshold, the corresponding first arrival time is the electrical life prediction value, a corresponding electrical life distribution model is established. Then, through Bayesian parameter estimation, model parameter identification under small sample conditions is realized to achieve the prediction of electrical life.

[0083] The method of simulating the degradation process of the contact resistance increment based on the Wiener process and combining the incremental sequence of the contact resistance value includes:

[0084] Based on the incremental sequence of contact resistance values, independence check and normal distribution check are performed on the contact resistance increments;

[0085] Set the drift coefficient and diffusion coefficient based on the contact resistance increment that passes the calibration;

[0086] The degradation process of the contact resistance increment is simulated based on the drift coefficient and diffusion coefficient.

[0087] Based on the collected breaking test data, the incremental sequence of contact resistance values ​​can be obtained by sequentially comparing the contact resistance values ​​of two adjacent breaking tests. Then, the independence check and normal distribution check of each contact resistance increment in each sequence are performed on the incremental sequences of multiple contact resistance values ​​integrated from multiple tests.

[0088] Among them, for one of the tests, based on the breaking test data, the incremental sequence of contact resistance values ​​is calculated, including:

[0089] Based on the breaking test data, the contact resistance value of the on-load tap changer after each breaking test is obtained;

[0090] Based on the contact resistance values ​​of adjacent breaking tests, an incremental sequence of contact resistance values ​​is constructed.

[0091] The independence check mentioned is based on the concept of independent random variables in probability theory. If the data points in the contact resistance increment sequence are independent of each other, it means that the change in contact resistance at each moment is not affected by the previous moment and can be regarded as an independent random event.

[0092] Specifically, the autocorrelation function is used to calculate the autocorrelation coefficient of the resistance increment sequence. Then, based on the preset significance level, the critical value of the autocorrelation coefficient is obtained according to the delay order and sequence length. If the autocorrelation coefficient is less than the critical value, it can be determined that the contact resistance increment sequences are independent at the corresponding delay order and pass the independence test. If there is a delay order that makes the autocorrelation coefficient greater than or equal to the critical value, the contact resistance increment sequence is considered to not meet the independence requirement and requires reprocessing or data collection.

[0093] The normal distribution check is carried out based on the central limit theorem. When the factors affecting the contact resistance change are numerous and independent of each other, the distribution of the contact resistance increment will approach the normal distribution.

[0094] Specifically, the contact resistance increment series are grouped and a frequency histogram is drawn to observe the distribution of the data, and the normal distribution is further verified by the normal distribution hypothesis test method.

[0095] Through independence verification and normal distribution verification, the contact resistance increment data that conforms to statistical laws can be screened out.

[0096] After determining the contact resistance increment that has passed the calibration, the drift coefficient and the diffusion coefficient can be set according to the statistical law of the contact resistance increment.

[0097] The drift coefficient reflects the average trend of the contact resistance increment over time, and embodies the characteristic of the contact resistance gradually increasing or decreasing due to factors such as physical and chemical changes under normal working conditions. Specifically, the mean value of the contact resistance increment that can be verified is the drift coefficient.

[0098] The diffusion coefficient describes the degree of fluctuation of the contact resistance increment around the average trend, reflecting the impact of environmental factors, accidental factors, etc. on the contact resistance change. In order to significantly simplify the mathematical operations and parameter estimation process while ensuring the accuracy of the model, this embodiment can specifically use the variance of the contact resistance increment to be verified as the diffusion coefficient.

[0099] The Wiener process is a continuous-time random process with independent increments and normal distribution characteristics, which is consistent with the verified contact resistance increment characteristics. Based on the drift coefficient and diffusion coefficient set according to the statistical change law of the contact resistance increment, it can effectively describe the random degradation process of the contact resistance over time.

[0100] The degradation process of the contact resistance increment simulated in this embodiment is:

[0101] ;

[0102] in, Is the running time equivalent to the number of interruptions The contact resistance increment after is the drift coefficient, which characterizes the average rate of contact material wear, is the diffusion coefficient, reflecting the random fluctuation intensity of arc erosion, is a standard Wiener process, satisfying , is the initial contact resistance increment.

[0103] Based on the degradation process of the contact resistance increment simulated above, an electrical life distribution model is established, including:

[0104] Set the failure threshold of the on-load tap changer. Based on the first arrival time theory, the time it takes to reach the failure threshold for the first time is used as the electrical life identification parameter.

[0105] Combined with the degradation process of contact resistance increment, the probability density function of electrical life identification parameters is set and the electrical life distribution model is established.

[0106] During long-term operation, the contact resistance of an on-load tap changer gradually increases. When this contact resistance reaches a certain level, it can affect the switch's normal operation and even cause failure. This critical value is known as the failure threshold. The first-reach time theory describes the time it takes for a variable to first reach a specific boundary value in a random process. This embodiment specifically utilizes this theory, using the time it takes to first reach the failure threshold as the electrical life identification parameter to directly reflect the duration of the on-load tap changer's operation from commissioning to failure.

[0107] Then, based on the time when the failure threshold is first reached, the random degradation characteristics of the contact resistance increment are linked to the electrical life by constructing a probability density function, thereby establishing an electrical life distribution model to describe the possibility of the electrical life taking values ​​at different times, that is, the probability of the on-load tap-changer failing at different times, and comprehensively reflecting the probability distribution of the electrical life of the on-load tap-changer.

[0108] The failure threshold in this implementation is set as:

[0109] ;

[0110] in, The maximum permissible contact resistance value is set according to the relevant design specifications of the on-load tap changer. It is the initial contact resistance value of the contacts of the newly-used on-load tap-changer.

[0111] Based on the first arrival time theory, the time to first reach the failure threshold is defined as:

[0112] ;

[0113] in, is the time when the failure threshold is first reached, is the lower bound function, which means that when the condition is met All Among the values, take the minimum value.

[0114] Based on the set failure threshold, the time to first reach the failure threshold, and the simulated degradation process of the contact resistance increment, a corresponding probability density function can be constructed. The probability density function obeys the inverse Gaussian distribution, and its expression is:

[0115] .

[0116] The probability density function is the expression of the electrical lifetime distribution model mentioned in this embodiment.

[0117] On the basis of the established electrical lifetime distribution model, the hierarchical Bayesian estimation method is used to realize the model parameter identification under small sample conditions, so as to solve the electrical lifetime distribution model and realize the prediction of electrical lifetime.

[0118] Traditional methods for determining drift and diffusion coefficients often rely on simple statistical calculations or empirical values, making it difficult to fully leverage the complex relationships between prior information and sample data. Bayesian parameter estimation, however, can more accurately capture parameter uncertainty by combining prior distributions with sample data, resulting in parameter estimates that are more realistic. Substituting these precise parameters into the electrical life distribution model significantly improves the accuracy of on-load tap changer electrical life predictions and reduces prediction errors.

[0119] Specifically, the electrical life prediction result of the on-load tap changer is obtained by Bayesian parameter estimation based on the electrical life distribution model, including:

[0120] Set the prior distribution of drift coefficient and diffusion coefficient;

[0121] Based on the prior distribution, posterior sampling is performed in combination with the incremental sequence of the calibrated contact resistance values ​​to obtain the posterior means of the drift coefficient and the diffusion coefficient;

[0122] Based on the posterior mean, the electrical life distribution model is solved to obtain the predicted value of the electrical life of the on-load tap changer.

[0123] The core idea of ​​Bayesian parameter estimation is to combine prior information with sample data and infer model parameters through posterior distribution. In the life distribution model of on-load tap changer, the drift coefficient and diffusion coefficient It is a key parameter that determines the trend and fluctuation characteristics of the contact resistance incremental degradation process.

[0124] The prior distribution reflects the subjective cognition or historical experience information of the drift coefficient and the diffusion coefficient before obtaining the sample data. On this basis, the prior distribution of the drift coefficient and the diffusion coefficient is set, including:

[0125] Identify the type of on-load tap changer and obtain historical operating data of the same type of on-load tap changer;

[0126] Based on historical operating data, the prior distribution of the drift coefficient and diffusion coefficient is set.

[0127] The prior distributions of the drift coefficient and diffusion coefficient set in this embodiment are:

[0128] ;

[0129] in, is the mean drift coefficient of historical data statistics, is the prior variance of the drift coefficient, is a normal distribution, is the inverse Gamma distribution to control the distribution shape of the diffusion coefficient, and These are all hyperparameters and are set according to the device type of the on-load tap changer.

[0130] In this embodiment, the initial drift coefficient in the prior distribution is set based on the historical data of the on-load tap changer of the same equipment type. , , , .

[0131] Then use the contact resistance value increment sequence of n times of verification test For sample data, preprocess it and first pass the unit root test to verify whether it conforms to the Wiener process characteristics. After passing, it is converted into a format suitable for Bayesian analysis, and then combined with the prior distribution to obtain the posterior distribution:

[0132] ;

[0133] in, For a given sample data Under the condition of , the joint posterior distribution of drift coefficient and diffusion coefficient is, The contact resistance value increment sequence of n tests given after verification The sample data set is composed of is the prior distribution of the drift coefficient, is the prior distribution of the diffusion coefficient, For the The collection time of sample data, For the moment The collected contact resistance increment.

[0134] Posterior sampling is then performed based on the Markov Chain Monte Carlo (MCMC) algorithm. Specifically, the initial drift coefficient and diffusion coefficient are set, new drift coefficients and diffusion coefficients are generated based on the posterior distribution, the corresponding acceptance probability is calculated, and the decision on whether to accept the new parameter value is made based on the acceptance probability. A series of samples are generated through continuous iteration, and the generated samples constitute the posterior distribution sample set of the drift coefficient and diffusion coefficient.

[0135] In this embodiment, the chain length of the Markov Chain Monte Carlo (MCMC) algorithm in the posterior sampling is set to 20,000 times, the corresponding initial iteration burn-in period is 5,000 times, and the convergence condition is that the corresponding Gelman-Rubin statistic is less than a preset threshold of 1.1.

[0136] Then, according to the drift coefficient samples and diffusion coefficient samples in the posterior distribution sample set, the corresponding posterior means are calculated respectively. and .

[0137] The posterior distribution can be used to integrate prior information and sample information to more accurately describe the distribution of parameters. The posterior mean obtained from the posterior distribution is further used as the optimal estimate of the parameter to solve the electrical life distribution model.

[0138] The solution for the posterior mean in the electrical lifetime distribution model includes:

[0139] Based on the probability density function corresponding to the electrical life distribution model, the corresponding electrical life mean formula is obtained. At the same time, considering the influence of the diffusion coefficient on the electrical life, the corresponding diffusion coefficient correction factor is added to the electrical life mean formula.

[0140] The corrected electrical life mean formula is:

[0141] ;

[0142] in, is the mean value of electrical life, i.e. the predicted value of electrical life.

[0143] When constructing the mean value formula of electrical life, the corresponding confidence interval is further constructed, and its expression is:

[0144] ;

[0145] in, for Confidence interval.

[0146] Substituting the calculated posterior mean into the corrected electrical life mean formula and the corresponding confidence interval expression, the specific electrical life prediction value and the corresponding confidence interval can be obtained.

[0147] During the solution process, the Mahalanobis distance detection indicator is also used to identify outliers in the solution results.

[0148] Specifically, the calculation expression of the Mahalanobis distance detection index is:

[0149] ;

[0150] in, For the The Mahalanobis distance detection index value corresponding to the sample data is For the The contact resistance increment corresponding to the sample data is: For the The collection time corresponding to the sample data.

[0151] By comparing the calculated Mahalanobis distance detection index value with the preset abnormality threshold, if the Mahalanobis distance detection index value corresponding to a sample data exceeds the preset abnormality threshold, it is judged as abnormal data and the retest program needs to be started to re-detect, calculate and analyze the relevant data.

[0152] The preset abnormal threshold in this embodiment is , is the 0.95 quantile of the chi-square distribution with 1 degree of freedom.

[0153] The forecast results were further revised in subsequent runs, including:

[0154] A recursive Bayesian update formula is set, and the electrical life prediction results of the on-load tap changer are corrected in real time based on the recursive Bayesian update formula.

[0155] The principle of recursive Bayesian updating is to use each newly acquired data as a sample and the posterior distribution obtained from the previous update as the prior distribution for the next update, so as to achieve gradual optimization of the parameter estimation results.

[0156] When new incremental contact resistance data is generated, the recursive Bayesian update mechanism is used to combine the new data with the posterior distribution of the previous moment to recalculate the posterior distribution of the drift coefficient and diffusion coefficient. The new posterior distribution integrates historical data information and the latest observation data, which can more accurately reflect the current degradation state of the on-load tap changer, and thus correct the electrical life prediction results.

[0157] The Bayesian update formula used in this embodiment is expressed as:

[0158] ;

[0159] in, is a sample data set composed of the incremental sequence of the latest contact resistance values. It is the historical data of the incremental series of contact resistance values.

[0160] The feasibility of the on-load tap changer electrical life prediction method described in this embodiment is evaluated using mean absolute percentage error and coverage probability.

[0161] The calculation expression of mean absolute percentage error is: ;

[0162] The calculation expression of coverage probability is: ;

[0163] in, To evaluate the sample size, that is, the total number of life data used to evaluate the effectiveness of the electrical life prediction method, For the The actual life of the evaluation sample, For the The predicted lifespan of the evaluation samples, For the The prediction confidence interval corresponding to the evaluation samples, is the assignment function, in When the actual life value of an evaluation sample falls within the corresponding prediction confidence interval, it is assigned a value of 1; otherwise, it is assigned a value of 0.

[0164] Taking three on-load tap-changers OLTC-01, OLTC-02, and OLTC-03 as evaluation samples, their actual lifespans, as measured through actual measurements and equipment operation records, were 86,542, 79,331, and 92,147 times, respectively. The electrical lifespan prediction method provided in this embodiment was used to predict these three devices. A comparison of the predicted results with the actual lifespans is shown in Table 1.

[0165] Table 1 Comparison between electrical life prediction results and actual life

[0166]

[0167] Data statistics were performed based on the prediction results. The statistical results showed that the corresponding mean absolute percentage error and coverage probability were: MAPE=2.45±0.62%, CP=93.7%, which can effectively verify the effectiveness of the electrical life prediction method described in this embodiment.

[0168] It can be seen that the on-load tap-changer electrical life prediction method given in this embodiment can not only reduce the number of tests by 85% compared with the traditional full-cycle experimental method, but also integrate prior knowledge and experimental data through the Bayesian framework, and still maintain a prediction error of <3% under small sample conditions, which not only reduces the experimental cost but also ensures the prediction accuracy.

[0169] In another aspect of this embodiment, a system for predicting the life of an on-load tap-changer is provided. Figure 3 Shown, including:

[0170] The breaking test unit is used to establish the breaking test wiring of the on-load tap changer and collect the breaking test data during the breaking test;

[0171] The modeling unit is connected to the breaking test unit and is used to simulate the degradation process of the contact resistance increment according to the breaking test data and establish an electrical life distribution model accordingly;

[0172] The life prediction unit is connected to the modeling unit and is used to solve the electrical life distribution model according to Bayesian estimation to obtain the electrical life prediction result of the on-load tap-changer.

[0173] Among them, the breaking test unit provides a breaking test environment for the on-load tap-changer, including a test power supply and a test transformer. After the three-phase wiring topology of the on-load tap-changer to be tested is reconstructed, it is connected to the breaking test unit to establish the test circuit, and the breaking test of the on-load tap-changer can be carried out.

[0174] The breaking test unit is also equipped with a variety of sensors, such as current sensors, voltage sensors, contact resistance testers, etc., which can collect breaking test data in real time during the breaking test process.

[0175] The modeling unit and life prediction unit can use high-performance computers, servers, microcontrollers and other data analysis and processing devices, which carry corresponding algorithms such as the degradation process of contact resistance increment simulation based on Wiener process, electrical life prediction based on Bayesian parameter estimation, and electrical life prediction correction. Through data processing and analysis of breaking test data, the life prediction of the on-load tap changer can be effectively realized.

[0176] The embodiment described above is only a preferred solution of the present invention and does not limit the present invention in any form. Other variations and modifications are possible without exceeding the technical solution described in the claims.

Claims

1. A method for predicting the electrical life of an on-load tap-changer, characterized in that: include: Conduct electrical life testing of on-load tap changers based on three-phase cross-connection and obtain breaking test data; Specifically, the on-load tap changers are grouped and interconnected, and the voltage level tap changers corresponding to the on-load tap changers are reconstructed; Based on the phase setting of the test power supply, a test circuit is established with the reconstructed voltage range tap changer; Connect the test circuit to the test transformer, perform a breaking test by switching the voltage position tap changer, and collect breaking test data for each breaking test; Based on the breaking test data, the incremental sequence of contact resistance values ​​is calculated; Specifically, based on the breaking test data, the contact resistance value of the on-load tap changer after each breaking test is obtained; Based on the contact resistance values ​​of adjacent breaking tests, an incremental sequence of contact resistance values ​​is constructed; Based on the Wiener process and combined with the incremental sequence of contact resistance values, the degradation process of contact resistance increment is simulated; Specifically, based on the incremental sequence of the contact resistance value, the independence check and normal distribution check of the contact resistance increment are performed; Set the drift coefficient and diffusion coefficient based on the contact resistance increment that passes the calibration; The degradation process of contact resistance increment is simulated based on drift coefficient and diffusion coefficient; According to the degradation process of simulated contact resistance increment, an electrical life distribution model is established; Specifically, a failure threshold of the on-load tap changer is set, and based on the first arrival time theory, the time when the failure threshold is first reached is used as the electrical life identification parameter; Combined with the degradation process of contact resistance increment, the probability density function of electrical life identification parameters is set and the electrical life distribution model is established; Based on the electrical life distribution model, the electrical life prediction results of the on-load tap changer are obtained through Bayesian parameter estimation. Specifically, the prior distribution of drift coefficient and diffusion coefficient is set; Based on the prior distribution, posterior sampling is performed in combination with the incremental sequence of the calibrated contact resistance values ​​to obtain the posterior means of the drift coefficient and the diffusion coefficient; Based on the posterior mean, the electrical life distribution model is solved to obtain the predicted value of the electrical life of the on-load tap changer.

2. The method for predicting the electrical life of an on-load tap-changer according to claim 1, characterized in that: The prior distribution of the drift coefficient and the diffusion coefficient is set, including: Identify the type of on-load tap changer and obtain historical operating data of the same type of on-load tap changer; Based on historical operating data, the prior distribution of the drift coefficient and diffusion coefficient is set.

3. The method for predicting the electrical life of an on-load tap-changer according to claim 1, characterized in that: Also includes: When solving the electrical life distribution model, the Mahalanobis distance detection index is also used to identify outliers in the solution results.

4. The method for predicting the electrical life of an on-load tap-changer according to claim 1, characterized in that: Also includes: A recursive Bayesian update formula is set, and the electrical life prediction results of the on-load tap changer are corrected in real time based on the recursive Bayesian update formula.

5. An on-load tap-changer electrical life prediction system, configured to execute the on-load tap-changer electrical life prediction method according to any one of claims 1 to 4, characterized in that: include: The breaking test unit is used to establish the breaking test wiring of the on-load tap changer and collect the breaking test data during the breaking test; The modeling unit is connected to the breaking test unit and is used to simulate the degradation process of the contact resistance increment according to the breaking test data and establish an electrical life distribution model accordingly; The life prediction unit is connected to the modeling unit and is used to solve the electrical life distribution model according to Bayesian estimation to obtain the electrical life prediction result of the on-load tap-changer.

Images