Real-time state evaluation method and system for voltage transformer

Abstract

The invention provides a real-time state evaluation method and system for a voltage transformer, and relates to the field of measurement and monitoring of a power system. Secondary side historical operation data of the voltage transformer are collected, and data preprocessing is carried out; a KNN-ELM model is adopted, an evaluation threshold value is preset, current operation data are input into the optimal KNN-ELM model, a predicted value of secondary output of the current three-phase voltage transformer is obtained, a current evaluation index is calculated for comparison and judgment, the health state of the voltage transformer equipment is evaluated, an evaluation threshold value self-updating strategy of an online model is established, and the online model of the three-phase voltage transformer equipment is updated according to the evaluation threshold value self-updating strategy. And further judging whether a consistency drift fault occurs or not. According to the embodiment of the invention, the accuracy and reliability of the collected historical data and the current operation data are ensured, the deviation of model training and prediction is avoided, the real-time state evaluation of the voltage transformer is effectively carried out, the abnormal state of equipment is found and processed in time, and the safety and reliability of a power system are improved.

Description

Technical Field

[0001] The present invention relates to the field of measurement and monitoring of electric power systems, and in particular to a real-time state evaluation method and system for voltage transformers. Background Art

[0002] With the development of smart grids and the continuous expansion of the scale of power systems, the importance of voltage transformers in ensuring the safe operation of power systems and the accuracy of electric energy metering has become increasingly prominent. At present, voltage transformer error detection mainly relies on periodic offline calibration, which is not only time-consuming and labor-intensive, but also difficult to timely capture the error changes of the transformer during operation. In addition, the offline detection method is even more difficult to achieve centralized monitoring of large-scale transformers, especially in high-voltage power grids. Due to the complexity of the detection equipment and the harsh operating environment, the offline calibration mode can no longer meet the current real-time monitoring needs of the power grid. Therefore, there is an urgent need for an online monitoring method that can evaluate the status of the voltage transformer in real time, thereby identifying operating anomalies and improving the stability of equipment operation. Summary of the invention

[0003] The main purpose of the present invention is to provide a real-time status evaluation method and system for a voltage transformer, so as to solve the problem in the prior art that it is difficult to detect abnormal operation of the transformer in a timely manner.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is: a real-time state evaluation method for a voltage transformer, comprising the following steps:

[0005] S1: Collect historical operation data of voltage transformer, perform data preprocessing, and save as original data set;

[0006] The historical operation data is data characterizing the standard operation state of the transformer, including the secondary output of the three-phase voltage transformer when the equipment is in a normal state;

[0007] S2: Build a KNN-ELM model, use the original data set for training, and save the optimal KNN-ELM model;

[0008] S3: Input the current operating data of the voltage transformer into the optimal KNN-ELM model, obtain the predicted value of the current secondary output of the three-phase voltage transformer, and calculate the current evaluation index;

[0009] S4: If the current evaluation index is greater than the evaluation threshold, the current voltage transformer is in an abnormal state, and enters the system warning and records; the system warning needs to be divided into alarm triggering levels.

[0010] In the preferred solution, if the current evaluation index is not greater than the evaluation threshold, it is a normal state, and the evaluation threshold is updated, specifically:

[0011] Continue to collect online operation data in real time, calculate the evaluation threshold in real time through the weighted sliding time window, and dynamically adjust it. According to the divided alarm trigger level, determine whether a consistency drift fault occurs, including:

[0012] The evaluation threshold is calculated in real time through a weighted sliding time window. The formula is:

[0013]

[0014] Where t is the current time; Q q (t+1) is the statistical threshold at the next moment; Q' is the confidence quantile in the sliding window at the previous moment; Q q is the initial threshold; α is the adjustment weight; Q(t) is the evaluation statistic at the current moment; P is the cumulative number of samples.

[0015] In the preferred solution, in step S3, it is determined whether a consistency drift fault occurs, and the change rate of the evaluation index change ratio is monitored to assist in determining whether the mutual inductor group deteriorates at the same time. The formula is as follows:

[0016]

[0017] In the formula, R mean,i It represents the relative change ratio of the mean of the i-th window to the i-1-th window; T i and T i-1 are the dynamic means of the i-th and i-1-th windows respectively, and n is the total number of windows;

[0018]

[0019] In the formula, ΔR mean,i It represents the rate of change of the change ratio of the i-th window, max() is the maximum value of the two, ξ is a very small positive number, and n is the total number of windows;

[0020]

[0021] Where, represents the relative change ratio of the standard deviation of the i-th window to the i-1-th window; S i and S i-1 are the dynamic standard deviations of the i-th and i-1-th windows respectively, and n is the total number of windows;

[0022]

[0023] In the formula, ΔR std,i It represents the rate of change of the i-th window relative to the i-1-th window, that is, R std,i With R std,i-1 The rate of change of , n is the total number of windows.

[0024] In the preferred embodiment, the data preprocessing in step S1 includes:

[0025] A1: Determine whether the collected data is missing and mark the missing values ​​as NaN;

[0026] A2: Use the 3σ principle to eliminate data. The eliminated data is marked as NaN. The formula is:

[0027]

[0028] Where xi is the i-th value in the data set, and N is the total number of data points;

[0029] A3: Use PCHIP to interpolate the marked NaNs. The formula is:

[0030]

[0031] Where z0 and z1 are the known time points before and after the interpolation point, f0 and f1 are the known data values ​​corresponding to these time points; f 0′ and f 1′ is the corresponding first-order derivative, and the interpolated data are saved to the original data set;

[0032] A4: Organize the output data of the same group of three-phase voltage transformers in the interpolated original data set into a matrix form K to obtain the corresponding input vector:

[0033]

[0034] In the formula, w x (t), w y (t) and w z (t) represents the output of the three-phase voltage transformer at time t; M is the sequence length; Next, multiply the matrix K by the sparse matrix Q (3×6) , as the input Z of the model:

[0035]

[0036] It can be calculated that

[0037] [Z x |Z y |Z z ]=[W y (t)W z (t)|W x (t)W x (t)|W z (t)W y (t)];

[0038] Simplifying further:

[0039] [X x |X y |X z ]=K=[W x (t)|W y (t)|W z (t)];

[0040] In the formula, Z x , Z y and Z z are the input feature matrices of the three-phase inductor respectively; Q converts K into the input form of the prediction model, W x (t), W y (t) and W z (t) is the corresponding input vector.

[0041] In the preferred solution, in step S2, the KNN-ELM model is constructed by combining the KNN and ELM models, wherein:

[0042] The KNN model calculates the Euclidean distance between samples in the feature space of the input vector, identifies neighboring samples in the historical data that are most similar to the current input data features, sorts them in sequence, and constructs the corresponding feature vector;

[0043] Using the ELM model to perform regression prediction on the characteristic vector, to obtain a predicted value of the secondary output voltage of the three-phase voltage transformer at each predicted moment;

[0044] The ELM model introduces a penalty factor, and the formula is as follows:

[0045]

[0046] Where: β* is the least squares solution of weight β; I is the diagonal unit matrix; C is the regularization coefficient, h is the hidden node output of ELM; h T is the transposed matrix of h.

[0047] In a preferred embodiment, the calculation of the evaluation threshold comprises the following steps:

[0048] During model training, the output of the KNN-ELM model is matched one by one with the original data set saved after interpolation, and a subtraction operation is performed to obtain the prediction residual. The squared prediction residual is selected as the evaluation indicator and the evaluation standard of the offline model. The formula is:

[0049]

[0050] In the formula, e i is the three-phase voltage transformer at t i The moment prediction residual vector; for e iThe transposed matrix of i is the corresponding squared prediction residual, i.e., the evaluation index;

[0051] According to the obtained prediction residuals, the distribution of the residual evaluation index is estimated using the Gamma probability model, and the control chart method is used to select the appropriate high quantile as the evaluation threshold for the evaluation index.

[0052] In the preferred solution, in step S4, the system warning divides the alarm according to the frequency of the deviation, and divides it into different alarm trigger levels, including warning and abnormal states, and the process is as follows:

[0053] 1) If the Q(t) of the current online running data is greater than the adaptive updated judgment threshold Q q (t), then start marking;

[0054] 2) Continue to calculate the evaluation index Q(t) of the online running data for the next time interval and Q(t) at the time of stopping updating E For comparison, if n consecutive E The time interval is greater than Q E , it is marked as a warning state and a log is recorded;

[0055] 3) After marking as warning state, if E time intervals, or 3*n consecutive E There are at least 2*n time intervals E If all the discrimination indicators of a certain number are greater than the judgment threshold, it is marked as an abnormal state, triggering a fault alarm and alerting the operation and maintenance personnel;

[0056] Among them, n E The value of is set to: the maximum number of consecutive missing values ​​and consecutive values ​​greater than three times the standard deviation in the historical measurement data collected in step S1;

[0057] The log contains timestamp, judgment threshold, evaluation index, alarm status (warning or abnormality) and duration.

[0058] A real-time status assessment system for a voltage transformer, comprising:

[0059] A data acquisition module is used to collect historical operation data of the voltage transformer, perform data preprocessing, and save it as an original data set; the historical operation data is data characterizing the standard operation state of the transformer, including the secondary output of the three-phase voltage transformer when the equipment is in a normal state;

[0060] Model training module, used to build the KNN-ELM model, train it with the original data set, and save the optimal KNN-ELM model;

[0061] The model evaluation module is used to input the current operation data of the voltage transformer into the optimal KNN-ELM model, obtain the predicted value of the current secondary output of the three-phase voltage transformer, and calculate the current evaluation index;

[0062] The early warning trigger module is used to, if the current evaluation index is greater than the evaluation threshold, the current voltage transformer is in an abnormal state, enter the system early warning and record; the system early warning needs to be divided into alarm trigger levels.

[0063] The present invention provides a real-time status evaluation method for a voltage transformer, which uses the historical data of the transformer as a benchmark to construct a virtual standard, compares the current operating data with the virtual standard, and evaluates the status of the voltage transformer: a regression model is constructed by integrating the KNN-ELM model, and an evaluation index is constructed as a benchmark standard according to the output result of the model. The technical effects of the present invention are as follows:

[0064] 1) Based on the historical data of the secondary side output of the three-phase voltage transformer, the present invention uses the KNN-ELM model and evaluation indicators to establish a virtual standard through the output synergy correlation between the three-phase voltage transformers. This method converts the equipment performance degradation of the three-phase voltage transformer into an abnormality detection problem, compares the real-time collected voltage transformer output data with the established virtual standard indicators, and determines whether the voltage transformer is in an abnormal state at the current moment. Compared with the traditional offline calibration mode, the present invention extracts the statistics of the existing voltage transformer output data and analyzes the change rules to determine whether the transformer is abnormal at the current moment, thereby reducing the operation and maintenance costs and reducing the system operation risks.

[0065] 2) The present invention adopts a sliding time window update method to adaptively adjust the judgment threshold, which can reduce the negative impact of unbalanced fluctuations in the power grid and improve the sensitivity of anomaly detection. At the same time, this adaptive update mechanism ensures the stability and accuracy of monitoring, so that relatively accurate error detection can still be achieved in a complex power grid environment.

[0066] 3) The present invention provides an alarm triggering identification process, divides the alarm level according to the deviation frequency, prompts the operation and maintenance personnel whether further inspection is needed, and records relevant data in real time to form a monitoring log.

[0067] 4) The present invention uses the standard deviation and the rate of change of the mean threshold change ratio to assist in determining whether a consistency fault occurs in the mutual inductor.

[0068] 5) The present invention performs abnormality detection on the voltage transformer in operation in an online manner, which can timely discover early abnormalities and reduce the dependence on manual offline calibration, thereby greatly reducing the frequency and cost of operation and maintenance.

[0069] 6) The present invention uses the historical operation data of the transformer itself as a benchmark, without relying on complex hardware equipment. Therefore, the method of the present invention is applicable to voltage transformers of different voltage types and has high versatility. This feature provides support for the scalability of the system, enabling it to be applied to various power grid monitoring scenarios, further enhancing the practical value of the technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0071] Figure 1 It is a schematic diagram of the model process framework of the present invention;

[0072] Figure 2 is a flow chart of the state assessment method of the present invention;

[0073] Figure 3 It is a comparison display diagram of the output voltage of the present invention before and after PCHIP interpolation;

[0074] Figure 4 This is a diagram showing the online monitoring effect under the adaptive threshold of the present invention;

[0075] Figure 5 This is a diagram showing the online monitoring effect of the present invention under performance degradation;

[0076] Figure 6 The present invention adopts Figure 5 Corresponding transformer auxiliary indicator diagram. DETAILED DESCRIPTION

[0077] Example 1

[0078] like Figure 1-6 As shown, a real-time status evaluation method for a voltage transformer comprises the following steps:

[0079] S1: Collect historical operation data of voltage transformer, perform data preprocessing, and save as original data set;

[0080] The historical operation data is data that characterizes the standard operation status of the transformer, including the secondary output of the three-phase voltage transformer when the equipment is in a normal state.

[0081] S2: Build a KNN-ELM model, use the original data set for training, and save the optimal KNN-ELM model.

[0082] S3: Input the current operating data of the voltage transformer into the optimal KNN-ELM model, obtain the predicted value of the current three-phase voltage transformer secondary output, and calculate the current evaluation index.

[0083] S4: If the current evaluation index is greater than the evaluation threshold, the current voltage transformer is in an abnormal state, and the system enters the early warning and records it; the system early warning needs to be divided into alarm trigger levels.

[0084] If the current evaluation index is not greater than the evaluation threshold, it is in a normal state and the evaluation threshold is updated. Specifically, the online operation data is collected in real time, and the evaluation threshold is calculated in real time through a weighted sliding time window, and dynamically adjusted. According to the level of the divided alarm trigger, it is determined whether a consistency drift fault occurs, including:

[0085] The evaluation threshold is calculated in real time through a weighted sliding time window. The formula is:

[0086]

[0087] Where t is the current time; Q q (t+1) is the statistical threshold at the next moment; Q' is the confidence quantile in the sliding window at the previous moment; Q q is the initial threshold; α is the adjustment weight; Q(t) is the evaluation statistic at the current moment; P is the cumulative number of samples.

[0088] like Figure 1-2 As shown, this embodiment collects historical operating data of the voltage transformer, trains the model after data preprocessing, saves the optimal KNN-ELM model, presets the evaluation threshold, inputs the current operating data into the optimal KNN-ELM model, obtains the current predicted value of the secondary output of the three-phase voltage transformer, and calculates the current evaluation index for comparison and judgment, evaluates the health status of the voltage transformer equipment, and establishes an online model evaluation threshold self-update strategy. As the real-time data collected online continues to accumulate, the evaluation threshold in step 2 is dynamically adjusted; the alarm trigger level is divided, and further judgment is made as to whether a consistency drift fault occurs. This embodiment ensures the accuracy and reliability of the collected historical data and current operating data to avoid deviations in model training and prediction, effectively performs real-time status evaluation of the voltage transformer, promptly discovers and handles abnormal equipment status, and improves the safety and reliability of the power system.

[0089] In step S1, historical operation data of the voltage transformer is collected. The collection category of historical data is the secondary output of the three-phase voltage transformer when the equipment is in a normal state, so as to represent the standard operation state of the transformer.

[0090] In the preferred embodiment, the data preprocessing in step S1 includes:

[0091] A1: Determine whether the collected data is missing and mark the missing values ​​as NaN.

[0092] A2: Noise is unavoidable in the actual data collection process, so the mean μ and standard deviation σ of the collected data are calculated, and the data is eliminated using the 3σ principle. The eliminated data is marked as NaN. The formula is:

[0093]

[0094] Where xi is the i-th value in the data set and N is the total number of data points.

[0095] The data outside [μ-3σ,μ+3σ] are regarded as noise outliers, and the deleted data are marked as NaN. A3: PCHIP is used to interpolate the marked NaN, and the formula is:

[0096]

[0097] Where z0 and z1 are the known time points before and after the interpolation point, f0 and f1 are the known data values ​​corresponding to these time points; f 0′ and f 1′ is the corresponding first-order derivative, and the interpolated data is saved to the original data set.

[0098] like Figure 3 The figure shows the comparison of the output voltage before and after PCHIP interpolation. PCHIP constructs a piecewise cubic polynomial on each interval, so that the value and derivative of the interpolation result at each node remain consistent, maintaining the smoothness and monotonicity of the data and avoiding the oscillation problem.

[0099] A4: After processing the data in steps A1 to A3, the output data of the same group of three-phase voltage transformers in the interpolated original data set are organized into a matrix form K to obtain the corresponding input vector:

[0100]

[0101] In the formula, w x (t), w y (t) and w z (t) represent the output of the three-phase voltage transformer at time t; M is the sequence length; the essence of the matrix transformation is to splice three one-dimensional data together.

[0102] Next, multiply the matrix K by the sparse matrix Q (3×6) , as the input Z of the model:

[0103]

[0104] It can be calculated that

[0105] [Z x |Z y |Zz ]=[W y (t)W z (t)|W x (t)W z (t)|W x (t)W y (t)] (7);

[0106] Simplifying further:

[0107] [X x |X y |X z ]=K=[W x (t)|W y (t)|W z (t)] (8);

[0108] In the formula, Z x , Z y and Z z are the input feature matrices of the three-phase inductor respectively; Q converts K into the input form of the prediction model, W x (t), W y (t) and W z (t) is the corresponding input vector.

[0109] Before using the processed data for model training, it is recommended to conduct further data validation and visual analysis to confirm whether the data is consistent with expectations.

[0110] Through the above steps, the data can be preprocessed effectively, laying a solid foundation for the subsequent KNN-ELM model training and real-time status evaluation. This embodiment adopts the KNN-ELM model.

[0111] In the preferred solution, in step S2, the KNN-ELM model is constructed by combining the KNN and ELM models, wherein:

[0112] The KNN model calculates the Euclidean distance between samples in the feature space of the input vector, identifies neighboring samples in the historical data that are most similar to the current input data features, sorts them in sequence, and constructs the corresponding feature vector;

[0113] The characteristic vector expression is:

[0114] x i =g([d i,1 ,θ i,1 ,x i,1 ],[d i,2 ,θ i,2 ,x i,2 ],…,[d i,k ,θ i,k ,xi,k ]) (9);

[0115] Where d and θ are the relative polar coordinates of the training sample and the new sample respectively; x is the label value of the training sample.

[0116] The ELM model is then used to perform regression prediction on the feature vector to obtain the predicted value of the secondary output voltage of the three-phase voltage transformer at each predicted moment.

[0117] The ELM model introduces a penalty factor, the formula is as follows:

[0118]

[0119] Where β* is the least squares solution of weight β; I is the diagonal unit matrix; C is the regularization coefficient; and h is the hidden node output of ELM.

[0120] When building a KNN-ELM model, it is necessary to carefully select hyperparameters such as the K value (the number of neighboring samples), the number of hidden nodes of ELM, and the regularization coefficient C to optimize the performance of the model.

[0121] After obtaining the output of the KNN model and the ELM model, they are used to calculate the evaluation threshold. In real-time status evaluation, the evaluation threshold is used to determine whether the current status of the device is normal. If the current evaluation index exceeds the threshold, a warning or alarm may be triggered.

[0122] In the KNN-ELM model of this embodiment, the ELM model receives the feature vector extracted by the KNN model as input, and performs regression prediction to obtain the secondary output voltage value of the three-phase voltage transformer at each predicted moment.

[0123] In a preferred embodiment, the calculation of the evaluation threshold comprises the following steps:

[0124] During model training, the output of the KNN-ELM model is matched one by one with the original data set saved after interpolation, and a subtraction operation is performed to obtain the prediction residual. The squared prediction residual is selected as the evaluation indicator and the evaluation standard of the offline model. The formula is:

[0125]

[0126] In the formula, e i is the three-phase voltage transformer at t i The moment prediction residual vector; for e i The transposed matrix of i is the corresponding squared prediction residual, which is the evaluation indicator.

[0127] According to the obtained prediction residual, the distribution of the residual evaluation index is estimated using the Gamma probability model, and the control chart method is used to select the appropriate high quantile as the evaluation threshold for the evaluation index Q. The evaluation threshold is recorded as Q q .

[0128] like Figure 4 As shown, Figure a is a diagram of the evaluation index and evaluation threshold established by the KNN-ELM model of this embodiment when offline, and Figure b is a diagram of the real-time online test results of the KNN-ELM model obtained by using the threshold in Figure a.

[0129] The self-updating strategy of the evaluation threshold is used to reduce the impact of the fluctuation of the three-phase imbalance in the power grid on the transformer evaluation statistic Q. This strategy calculates the evaluation threshold in real time through a weighted sliding time window to achieve adaptive adjustment, updates the current threshold by giving higher weights to the newer data in the sliding window, and slowly expands the normal data to Q through the accumulation of P. q The weight influence of (t+1).

[0130] In the preferred solution, the system warning in step S4 divides the alarm into different alarm trigger levels according to the frequency of deviation, including two states: warning and abnormal. The process is as follows:

[0131] 1) If the Q(t) of the current online running data is greater than the adaptive updated judgment threshold Q q (t), then start marking and set Q q (t) is marked as Q E .

[0132] 2) Continue to calculate the evaluation index Q(t) of the online operation data of the next time interval. If n consecutive E The time interval is greater than Q E , it is marked as a warning state and a log is recorded.

[0133] 3) After marking as warning state, if E time intervals, or 3*n consecutive E There are at least 2*n time intervals E The discriminant index of the quantity is greater than Q E , it is marked as an abnormal state, triggering a fault alarm and alerting the operation and maintenance personnel.

[0134] Among them, n E The value of is set to: the maximum number of consecutive missing values ​​and consecutive values ​​greater than three times the standard deviation in the historical measurement data collected in step S1.

[0135] The log contains timestamp, judgment threshold, evaluation indicator, alarm status (warning or abnormality) and duration.

[0136] This embodiment provides an alarm triggering identification process, divides the alarm level according to the deviation frequency, prompts the operation and maintenance personnel whether further inspection is needed, and records relevant data in real time to form a monitoring log.

[0137] In the preferred solution, in step S3, it is determined whether a consistency drift fault occurs, and the change rate of the evaluation index change ratio is monitored to assist in determining whether the mutual inductor group deteriorates at the same time. The formula is as follows:

[0138]

[0139] In the formula, R mean,i It represents the relative change ratio of the mean of the i-th window to the i-1-th window; T i and T i-1 are the dynamic means of the i-th and i-1-th windows respectively, and n is the total number of windows;

[0140]

[0141] In the formula, ΔR mean,i It represents the rate of change of the change ratio of the i-th window, max() is the maximum value of the two, ξ is a very small positive number, and n is the total number of windows;

[0142]

[0143] In the formula, R std,i It represents the relative change ratio of the standard deviation of the i-th window to the i-1-th window; S i and S i-1 are the dynamic standard deviations of the i-th and i-1-th windows respectively, and n is the total number of windows;

[0144]

[0145] In the formula, ΔR std,i It represents the rate of change of the i-th window relative to the i-1-th window, that is, R std,i With R std,i-1 The rate of change of , n is the total number of windows, and ξ is a very small positive number to prevent the denominator from being zero.

[0146] The degradation drift of the transformer increases over time, which is manifested as the accumulation of system errors and the increase of measurement ratio errors caused by the slow change of specific parameters. When the equipment is running independently, it is a very small probability event that two-phase and three-phase degradation drifts at the same time and maintains the same degradation rate. Therefore, when the model first determines that the transformer is faulty, the transformer group fault is set as a single-phase fault, and the value and trend of the change rate index and acceleration index at this time are recorded. On this basis, the change ratio R mean,i , R std,i The rate of change of the ratio ΔRmean,i , ΔR std,i , and compare it with the change ratio index and the change rate index of the change ratio at the time of the first fault to assist in judging two-phase or three-phase faults.

[0147] In order to determine the evaluation effect of the real-time status, a fault data that changes linearly with time is added to the normal data, and the model is used for real-time evaluation. The evaluation results are entered into Figure 5 As shown in Figure 2, it can be seen that the real-time index of the fault data is significantly greater than the evaluation threshold, showing a good judgment result. Figure 5-6 As shown, the change ratio R mean , R std and the rate of change of relative change ratio ΔR mean , ΔR std In order to show the trend more clearly, the figure captures a local range. It can be observed that before the fault occurs, ΔR mean , ΔR std It fluctuates randomly around 0, and after the failure gradient begins at the 6000th point, it presents a nearly stable straight line trend, that is, and Δk is a constant, n E is the number of fault data, is the mean change rate of fault data, is the standard deviation change rate of the fault. The change ratio of the mean and standard deviation is R mean , R std The initial alternating oscillation turns into a trend in which the standard deviation significantly follows the increase of the mean, that is, and υ is a threshold coefficient greater than 1, is the mean of the fault data, is the standard deviation of the failure data, is the mean of normal values, is the standard deviation of the normal value. A reference indicator for auxiliary judgment is as follows:

[0148] Table 1 Reference index judgment conditions and categories

[0149]

[0150] As shown in Table 1, Δq is a constant. is the mean of the data in the next time window, is the standard deviation of the data in the next time window, The rate of change of the mean of the data in the next time window, is the rate of change of the standard deviation of the data in the next time window. The above judgment conditions will only be established when they are met at the same time.

[0151] This embodiment is based on the historical data of the secondary side output of the three-phase voltage transformer, through the output synergy correlation between the three-phase voltage transformers, and uses the KNN-ELM model and evaluation indicators to establish a virtual standard, converting the equipment performance degradation of the three-phase voltage transformer into an abnormality detection problem, and comparing the real-time collected voltage transformer output data with the established virtual standard indicators to determine whether the voltage transformer at the current moment is in an abnormal state. 1. Compared with the traditional offline calibration mode, by extracting the statistics of the existing voltage transformer output data and analyzing the change law, it is determined whether the transformer in operation at the current moment is abnormal, which reduces the operation and maintenance costs and reduces the system operation risk. 2. By performing abnormality detection on the operating voltage transformer in an online manner, early abnormalities can be discovered in time, reducing the reliance on manual offline calibration, thereby greatly reducing the frequency and cost of operation and maintenance. 3. Based on the historical operation data of the transformer itself, there is no need to rely on complex hardware equipment, so the method described in the present invention is applicable to voltage transformers of different voltage types and has high versatility. This feature provides support for the scalability of the system, enabling it to be applied to various power grid monitoring scenarios, further enhancing the practical value of the technology.

[0152] Example 2

[0153] In combination with Example 1, a real-time status assessment system for a voltage transformer is further described, including:

[0154] The data acquisition module is used to collect the historical operation data of the voltage transformer, perform data preprocessing, and save it as an original data set; the historical operation data is data that characterizes the standard operating state of the transformer, including the secondary output of the three-phase voltage transformer when the equipment is in normal state.

[0155] The model training module is used to build the KNN-ELM model, train it with the original data set, and save the optimal KNN-ELM model.

[0156] The model evaluation module is used to input the current operating data of the voltage transformer into the optimal KNN-ELM model, obtain the predicted value of the current secondary output of the three-phase voltage transformer, and calculate the current evaluation index.

[0157] The early warning trigger module is used to indicate that if the current evaluation index is greater than the evaluation threshold, the current voltage transformer is in an abnormal state, and the system early warning is entered and recorded; the system early warning needs to be divided into alarm trigger levels.

[0158] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A real-time status assessment method for a voltage transformer, characterized in that: The following steps are involved: S1: Collect historical operation data of voltage transformer, perform data preprocessing, and save as original data set; The historical operation data is data characterizing the standard operation state of the transformer, including the secondary output of the three-phase voltage transformer when the equipment is in a normal state; S2: Build a KNN-ELM model, use the original data set for training, and save the optimal KNN-ELM model; S3: Input the current operating data of the voltage transformer into the optimal KNN-ELM model, obtain the current predicted value of the secondary output of the three-phase voltage transformer, and calculate the current evaluation index; S4: If the current evaluation index is greater than the evaluation threshold, the current voltage transformer is in an abnormal state, and enters the system warning and records; the system warning needs to be divided into alarm triggering levels.

2. The real-time status assessment method for a voltage transformer according to claim 1, characterized in that: If the current evaluation index is not greater than the evaluation threshold, it is a normal state, and the evaluation threshold is updated, specifically: Continue to collect online operation data in real time, calculate the evaluation threshold in real time through the weighted sliding time window, and dynamically adjust it. According to the divided alarm trigger level, determine whether a consistency drift fault occurs, including: The evaluation threshold is calculated in real time through a weighted sliding time window. The formula is: Where t is the current time; Q q (t+1) is the statistical threshold at the next moment; Q' is the confidence quantile in the sliding window at the previous moment; Q q is the initial threshold; α is the adjustment weight; Q(t) is the evaluation statistic at the current moment; P is the cumulative number of samples.

3. The real-time status assessment method for a voltage transformer according to claim 2, characterized in that: In step S3, it is determined whether a consistency drift fault occurs. In order to assist in determining whether the mutual inductor group deteriorates at the same time by monitoring the change rate of the evaluation index change ratio, the formula is as follows: In the formula, R mean,i It represents the relative change ratio of the mean of the i-th window to the i-1-th window; T i and T i-1 are the dynamic means of the i-th and i-1-th windows respectively, and n is the total number of windows; In the formula, ΔR mean,i It represents the rate of change of the change ratio of the i-th window, max() is the maximum value of the two, ξ is a very small positive number, and n is the total number of windows; In the formula, R std,i It represents the relative change ratio of the standard deviation of the i-th window to the i-1-th window; S i and S i-1 are the dynamic standard deviations of the i-th and i-1-th windows respectively, and n is the total number of windows; In the formula, ΔR std,i It represents the rate of change of the i-th window relative to the i-1-th window, that is, R std,i With R std,i-1 The rate of change of , n is the total number of windows.

4. The real-time status assessment method for a voltage transformer according to claim 1, characterized in that: The data preprocessing in step S1 includes: A1: Determine whether the collected data is missing and mark the missing values ​​as NaN; A2: Use the 3σ principle to eliminate data. The eliminated data is marked as NaN. The formula is: Where xi is the i-th value in the data set, and N is the total number of data points; A3: Use PCHIP to interpolate the marked NaNs. The formula is: Where z0 and z1 are the known time points before and after the interpolation point, f0 and f1 are the known data values ​​corresponding to these time points; f 0′ and f 1′ is the corresponding first-order derivative, and the interpolated data are saved to the original data set; A4: Organize the output data of the same group of three-phase voltage transformers in the interpolated original data set into a matrix form K to obtain the corresponding input vector: In the formula, w x (t), w y (t) and w z (t) represents the output of the three-phase voltage transformer at time t; M is the sequence length; Next, multiply the matrix K by the sparse matrix Q (3×6) , as the input Z of the model: It can be calculated that [Z x |Z y |Z z ]=[W y (t)W z (t)|W x (t)W x (t)|W z (t)W y (t)]; Simplifying further: [X x |X y |X z ]=K=[W x (t)|W y (t)|W z (t)]; In the formula, Z x , Z y and Z z are the input feature matrices of the three-phase inductor respectively; Q converts K into the input form of the prediction model, W x (t), W y (t) and W z (t) is the corresponding input vector.

5. The real-time status assessment method for a voltage transformer according to claim 1, characterized in that: In step S2, the KNN-ELM model is constructed by combining the KNN and ELM models, where: The KNN model calculates the Euclidean distance between samples in the feature space of the input vector, identifies neighboring samples in the historical data that are most similar to the current input data features, sorts them in sequence, and constructs the corresponding feature vector; Using the ELM model to perform regression prediction on the characteristic vector, to obtain a predicted value of the secondary output voltage of the three-phase voltage transformer at each predicted moment; The ELM model introduces a penalty factor, and the formula is as follows: Where: β* is the least squares solution of weight β; I is the diagonal unit matrix; C is the regularization coefficient, h is the hidden node output of ELM; h T is the transposed matrix of h.

6. The real-time status assessment method for a voltage transformer according to claim 1, characterized in that: The calculation of the evaluation threshold comprises the following steps: During model training, the output of the KNN-ELM model is matched one by one with the original data set saved after interpolation, and a subtraction operation is performed to obtain the prediction residual. The squared prediction residual is selected as the evaluation indicator and the evaluation standard of the offline model. The formula is: In the formula, e i is the three-phase voltage transformer at t i The moment prediction residual vector; for e i The transposed matrix of i is the corresponding squared prediction residual, i.e., the evaluation index; According to the obtained prediction residuals, the distribution of the residual evaluation index is estimated using the Gamma probability model, and the control chart method is used to select the appropriate high quantile as the evaluation threshold for the evaluation index.

7. The real-time status assessment method for a voltage transformer according to claim 1, characterized in that: In step S4, the system warning divides the alarm into different alarm trigger levels according to the frequency of the deviation, including two states: warning and abnormal. The process is as follows: 1) If the Q(t) of the current online running data is greater than the adaptive updated judgment threshold Q q (t), then start marking; 2) Continue to calculate the evaluation index Q(t) of the online running data for the next time interval and Q(t) at the time of stopping updating E For comparison, if n consecutive E The time interval is greater than Q E , it is marked as a warning state and a log is recorded; 3) After marking as warning state, if E time intervals, or 3*n consecutive E There are at least 2*n time intervals E If all the discrimination indicators of a certain number are greater than the judgment threshold, it is marked as an abnormal state, triggering a fault alarm and alerting the operation and maintenance personnel; Among them, n E The value of is set to: the maximum number of consecutive missing values ​​and consecutive values ​​greater than three times the standard deviation in the historical measurement data collected in step S1; The log contains timestamp, judgment threshold, evaluation index, alarm status (warning or abnormality) and duration.

8. A real-time status assessment system for a voltage transformer, characterized in that: include: The data acquisition module is used to collect the historical operation data of the voltage transformer, perform data preprocessing, and save it as an original data set; The historical operation data is data characterizing the standard operation state of the transformer, including the secondary output of the three-phase voltage transformer when the equipment is in a normal state; Model training module, used to build the KNN-ELM model, train it with the original data set, and save the optimal KNN-ELM model; The model evaluation module is used to input the current operation data of the voltage transformer into the optimal KNN-ELM model, obtain the predicted value of the current secondary output of the three-phase voltage transformer, and calculate the current evaluation index; The early warning trigger module is used to, if the current evaluation index is greater than the evaluation threshold, the current voltage transformer is in an abnormal state, enter the system early warning and record; the system early warning needs to be divided into alarm trigger levels.

Images