Group correlation CVT error state discrimination method

Abstract

The invention discloses a group correlation CVT error state discrimination method, and belongs to the technical field of voltage transformer detection, and the method comprises the steps: collecting secondary loop voltage signals of a plurality of voltage transformers on a same-phase bus, and carrying out the processing of the signals; constructing a feature recognition matrix by using the processed secondary loop voltage signals of each voltage transformer, calculating a magnitude T of the feature recognition matrix, analyzing whether there is a voltage transformer with an abnormal error state in the group of voltage transformers according to the magnitude of the magnitude T, and determining the voltage transformer with the abnormal error state; performing dimension elimination processing on the magnitude T of the feature recognition matrix, inputting the processed magnitude T as a recognition feature into the state evaluation model for training, recognizing an abnormal T value, and positioning a voltage transformer with an abnormal error state; and evaluating the error state of the voltage transformer based on the trained state evaluation model. According to the invention, the error state of the voltage transformer is accurately evaluated.

Description

Technical Field

[0001] The invention relates to the technical field of voltage transformer detection, and in particular to a group-correlation CVT error state discrimination method. Background Art

[0002] Voltage transformers (VTs) play a key role in power networks. Their main task is to reduce the high voltage in the transmission line to a lower voltage level that can be used by standard instruments and protection devices. This conversion process is essential to maintaining the safety and reliability of the power system. In order to ensure the continuous and stable operation of the system, the working condition of the voltage transformer must be monitored and checked regularly to promptly detect and resolve potential problems or faults that may cause system instability. Traditionally, the status assessment of voltage transformers usually uses fixed threshold-based techniques or single algorithm machine learning models. Although these methods are easy to implement, they have some disadvantages. For example, threshold techniques may ignore small fluctuations in the data, while individual machine learning models may be unable to cope with complex and nonlinear data patterns, thereby affecting the accuracy of detection. In order to improve these problems, researchers have developed a variety of more advanced technical means in recent years to improve the accuracy of voltage transformer status detection.

[0003] In the prior art, Zhao Peng et al., according to the time series theory, divided the primary voltage fluctuation into non-stationary fluctuation caused by deterministic factors and small slow and stable fluctuation caused by random factors. At the same time, they proposed a stabilization method based on differential spacing adaptation and a difference evaluation index based on the consistency of in-phase measurement, respectively weakening the impact of non-stationary fluctuations and small slow and stable fluctuations in the primary voltage fluctuation on the state assessment. However, when there is an abnormal error in the voltage transformer, the difference index can effectively identify it, but it has not yet considered the situation where multiple voltage transformers are abnormal at the same time. Therefore, the accuracy of voltage transformer state detection is not high. Summary of the invention

[0004] In view of this, the present invention aims at the deficiencies of the prior art and provides a group-correlation CVT error state discrimination method.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a group correlation CVT error state discrimination method, comprising the following steps:

[0006] S1, collects secondary circuit voltage signals of multiple voltage transformers on the same-phase bus and processes the signals;

[0007] S2. Construct a feature recognition matrix using the processed secondary circuit voltage signals of each voltage transformer, calculate the value T of the feature recognition matrix, analyze whether there is a voltage transformer with an abnormal error state in the group of voltage transformers according to the value T, and determine the voltage transformer with the abnormal error state;

[0008] S3, performing dimension elimination processing on the value T of the feature recognition matrix, and inputting the processed value T as the recognition feature into the state assessment model for training, identifying abnormal T values, and locating the voltage transformer with abnormal error state;

[0009] S4. Based on the trained state evaluation model, the error state of the voltage transformer is evaluated.

[0010] Furthermore, in S1, the voltage signal of the secondary circuit of the voltage transformer is processed, including:

[0011] S11. Processing of outliers and missing values ​​of the signal

[0012] The signal is processed according to the 3sigma principle, and the formula is as follows:

[0013]

[0014] In the formula, represents the amplitude of the signal at the i-th moment of the m-th transformer after processing, represents the signal amplitude of the mth transformer at the i-th moment, Indicates the average value of the signal collected over a period of time;

[0015] S12, detrending the signal processed in step S11

[0016]

[0017] In the formula, is a detrended signal. express The derivative of .

[0018] Furthermore, in S2, the feature recognition matrix constructed is:

[0019]

[0020] In the formula, k m represents the transformation ratio of the mth voltage transformer, Indicates the output signal value of the secondary side of the mth voltage transformer;

[0021] Sum the absolute values ​​of each row of the feature recognition matrix to get the value T, which is expressed as:

[0022]

[0023] If the T value in each row is extremely small, it means that the error status of all voltage transformers is normal;

[0024] If the T value of each row increases significantly, it means that there is a voltage transformer with an abnormal error state. At this time, compare the T values ​​of each row. If the T value of a row is the largest, it means that the characteristic voltage transformer of this row is the voltage transformer with an abnormal error state.

[0025] Furthermore, in S3, the value T of the feature recognition matrix is ​​dimensionally eliminated:

[0026] S31. Simulate multiple error conditions, calculate the T value of each row of the feature recognition matrix under each error condition, and calculate the mean of the T value of each row under each error condition.

[0027] S32. Find the average value of T value corresponding to the normal error state in each error case That is, the mean value in step S31 Minimum value of:

[0028]

[0029] S33, divide each row of T values ​​by The standardized T i ', the formula is:

[0030]

[0031] Where, T i is the T value of the i-th row, T′ i is the T value after processing the i-th row.

[0032] Furthermore, the state assessment model is any one of SVM, DBSCAN and neural network.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The group-correlation CVT error state discrimination method of the present invention first collects output signals of multiple voltage transformers on the same bus, constructs an identification matrix, calculates the magnitude of the identification matrix to determine whether there is a transformer in an abnormal error state in the group of transformers and locates it, then constructs the magnitude of the matrix, improves the discrimination between features by creating new features, and then trains a state evaluation model. Using the trained state evaluation model, accurate evaluation of the error state of the voltage transformer is achieved.

[0035] The present invention uses the T value of the sum of the absolute values ​​of the elements in each row of the characteristic matrix to characterize the error states of different voltage transformers, and conveniently and quickly identifies the voltage transformer with abnormal error states. The algorithm is simple and the accuracy is high.

[0036] The present invention designs a unique method for removing the dimension of T values ​​under different error states, which can retain the distribution of T values ​​under different states and remove the influence of the dimension on the data regularity, so that the state assessment model can accurately learn the T value change law during abnormal errors, thereby improving the assessment accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a method flow chart of an embodiment of the present invention;

[0038] Figure 2 is a box plot of T values ​​corresponding to each mutual inductor in an embodiment of the present invention;

[0039] Figure 3 is a box plot of T values ​​corresponding to each mutual inductor after scaling at the same ratio in an embodiment of the present invention;

[0040] Figure 4 is a box plot of T values ​​corresponding to each mutual inductor after normalization processing in an embodiment of the present invention;

[0041] Figure 5 It is a box plot of T values ​​corresponding to each mutual inductor after standardization in the embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0043] Example 1

[0044] like Figure 1 As shown, a group correlation CVT error state discrimination method includes the following steps:

[0045] S1, collects secondary circuit voltage signals of multiple voltage transformers on the same-phase bus and processes the signals;

[0046] Processing of voltage transformer secondary circuit voltage signal, including:

[0047] S11. Processing of outliers and missing values ​​of the signal

[0048] The signal is processed according to the 3sigma principle, and the formula is as follows:

[0049]

[0050] In the formula, represents the amplitude of the signal at the i-th moment of the m-th transformer after processing, represents the signal amplitude of the mth transformer at the i-th moment, Indicates the average value of the signal collected over a period of time;

[0051] S12, detrending the signal processed in step S11

[0052]

[0053] In the formula, is a detrended signal. express The derivative of .

[0054] S2. Construct a feature recognition matrix using the processed secondary circuit voltage signals of each voltage transformer, calculate the value T of the feature recognition matrix, analyze whether there is a voltage transformer with an abnormal error state in the group of voltage transformers according to the value T, and determine the voltage transformer with the abnormal error state;

[0055] Usually, in order to meet the requirements of substation for signal stability and accuracy, multiple voltage transformers are connected to the busbar on the same phase. Therefore, the voltage values ​​measured by multiple voltage transformers are the same. The voltage transformer measurement value under standard conditions can be expressed as:

[0056]

[0057] In the formula, It represents the primary voltage value measured by the nth voltage transformer, and m represents that there are m voltage transformers on the same phase bus;

[0058] The output signals of multiple transformers are constructed into an output matrix and expressed as:

[0059]

[0060] In the formula, Indicates the output signal value of the secondary side of the voltage transformer, k m represents the transformation ratio of the mth voltage transformer, α m Represents the error of the mth transformer.

[0061] At the same time, the output signals of each voltage transformer are basically the same, and the main difference is caused by the different errors of each transformer. Therefore, the differences of the group transformers can be represented according to this characteristic, thereby reflecting the difference in the error state of each voltage transformer. The constructed feature recognition matrix is:

[0062]

[0063] In the formula, k m represents the transformation ratio of the mth voltage transformer, Indicates the output signal value of the secondary side of the mth voltage transformer;

[0064] Sum the absolute values ​​of each row of the feature recognition matrix to get the value T, which is expressed as:

[0065]

[0066] According to the redundancy design under actual working conditions, if one of the voltage transformers has an abnormal error, the corresponding T value will increase significantly; conversely, if the error state of all voltage transformers is normal, the corresponding T value will be extremely small. Therefore, the T value can reflect the error state of each transformer to a certain extent. Moreover, in the feature recognition matrix, each row element is the difference between a certain voltage transformer and the remaining voltage transformers, that is, the first row is the difference between the first voltage transformer and other voltage transformers, the second row is the difference between the second voltage transformer and other voltage transformers, ..., the mth row is the difference between the mth voltage transformer and other voltage transformers, therefore, the mth voltage transformer is called the characteristic voltage transformer of the mth row.

[0067] It can be seen that the method of judging the error state of the voltage transformer by the T value is:

[0068] If the T value in each row is extremely small, it means that the error status of all voltage transformers is normal;

[0069] If the T value of each row increases significantly, it means that there is a voltage transformer with an abnormal error state. At this time, compare the T values ​​of each row. If the T value of a row is the largest, it means that the characteristic voltage transformer in this row is the voltage transformer with an abnormal error state (for example, assuming U2 is abnormal, the absolute value of each element in the second row is large, while only one element in other rows has a large absolute value. Therefore, U2 can be judged to be abnormal).

[0070] S3, performing dimension elimination processing on the value T of the feature recognition matrix, and inputting the processed value T as the recognition feature into the state assessment model for training, identifying abnormal T values, and locating the voltage transformer with abnormal error state;

[0071] Due to the difference in the dimensions of T values ​​under different mutual inductor error states, it is difficult for the subsequent state assessment model to learn the changing rules of T values ​​during abnormal errors. Therefore, it is necessary to perform dimension elimination before identifying the T value. However, conventional normalization and standardization processing will inevitably affect the distribution of the original data. This patent proposes an improved proportional standardization method that can eliminate the dimensions while retaining the distribution of the values.

[0072] Method for dimension elimination of the value T of the feature recognition matrix:

[0073] S31. Simulate multiple error conditions, calculate the T value of each row of the feature recognition matrix under each error condition, and calculate the mean of the T value of each row under each error condition.

[0074] S32. Find the average value of T value corresponding to the normal error state in each error case That is, the mean value in step S31 Minimum value of:

[0075]

[0076] S33, divide each row of T values ​​by The standardized T i ', the formula is:

[0077]

[0078] Where, T i is the T value of the i-th row, T′ i is the T value after processing the i-th row;

[0079] The processed T value can effectively eliminate the influence of dimension, mainly eliminating the T value dimension of the normal voltage transformer, enhancing the discrimination of T value under different error states, and providing a certain data basis for the subsequent identification model.

[0080] S4. Based on the trained state evaluation model, the error state of the voltage transformer is evaluated. The state evaluation model can be selected from existing classification models, such as SVM, DBSCAN and neural network.

[0081] Application example: The evaluation status of 5 voltage transformers on the same busbar of a substation, where the ratio difference of voltage transformers 1 and 2 is 0.3% and 0.2% respectively. 100 cross-evaluations are performed under the same error state, and the T value box-line diagram corresponding to each transformer is obtained as shown below: Figure 2 shown.

[0082] According to the above figure, the characteristic value T value proposed by the method of the present invention can distinguish the voltage transformer under different error states, and Figure 3 , Figure 4 , Figure 5 As shown, compared with the traditional normalization and standardization methods, the proportional scaling method proposed in the present invention can also effectively remove the dimension while retaining the original distribution of the data.

[0083] The accuracy of voltage transformer error state assessment through T value identification is shown in the following table:

[0084] Evaluation Metrics Original T value Normalized T value Standardized T value Scaling T value in the same way Accuracy 75.26％ 80.47％ 70.54％ 94.83％ .

[0085] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, a person skilled in the art can still modify or make equivalent substitutions to the specific implementations of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention are within the scope of protection of the claims of the present invention to be approved.

Claims

1. A method for distinguishing CVT error states based on group correlation, characterized in that: The following steps are involved: S1, collects the secondary circuit voltage signals of multiple voltage transformers on the same-phase bus and processes the signals; S2. Construct a feature recognition matrix using the processed secondary circuit voltage signals of each voltage transformer, calculate the value T of the feature recognition matrix, analyze whether there is a voltage transformer with an abnormal error state in the group of voltage transformers according to the value T, and determine the voltage transformer with the abnormal error state; S3, performing dimension elimination processing on the value T of the feature recognition matrix, and inputting the processed value T as the recognition feature into the state assessment model for training, identifying abnormal T values, and locating the voltage transformer with abnormal error state; S4. Based on the trained state evaluation model, the error state of the voltage transformer is evaluated.

2. The CVT error state discrimination method based on group correlation according to claim 1, characterized in that: In S1, the voltage signal of the secondary circuit of the voltage transformer is processed, including: S11. Processing of outliers and missing values ​​of the signal The signal is processed according to the 3sigma principle, and the formula is as follows: In the formula, represents the amplitude of the signal at the i-th moment of the m-th transformer after processing, represents the signal amplitude of the mth transformer at the i-th moment, Indicates the average value of the signal collected over a period of time; S12, detrending the signal processed in step S11 In the formula, is a detrended signal. express The derivative of .

3. The CVT error state discrimination method based on group correlation according to claim 1, characterized in that: In S2, the constructed feature recognition matrix is: In the formula, k m represents the transformation ratio of the mth voltage transformer, Indicates the output signal value of the secondary side of the mth voltage transformer; Sum the absolute values ​​of each row of the feature recognition matrix to get the value T, which is expressed as: If the T value in each row is extremely small, it means that the error status of all voltage transformers is normal; If the T value of each row increases significantly, it means that there is a voltage transformer with an abnormal error state. At this time, compare the T values ​​of each row. If the T value of a row is the largest, it means that the characteristic voltage transformer of this row is the voltage transformer with an abnormal error state.

4. The CVT error state discrimination method based on group correlation according to claim 1, characterized in that: In S3, the value T of the feature recognition matrix is ​​dimensionally eliminated: S31. Simulate multiple error conditions, calculate the T value of each row of the feature recognition matrix under each error condition, and calculate the mean of the T value of each row under each error condition. S32. Find the average value of T value corresponding to the normal error state in each error case That is, the mean value in step S31 Minimum value of: S33, divide each row of T values ​​by The standardized T i ', the formula is: Where, T i is the T value of the i-th row, T i ' is the T value after processing the i-th row.

5. The CVT error state discrimination method based on group correlation according to claim 1, characterized in that: The state assessment model is any one of SVM, DBSCAN and neural network.

Images