Magnetic array current sensor measurement error self-repairing method and system

Abstract

The invention discloses a magnetic array type current sensor measurement error self-repairing method and system, and mainly relates to the technical field of current sensor error measurement and repairing. Comprising the steps of obtaining a linear relation coefficient between output of a magnetic field sensing unit and a measured current; calculating a covariance matrix of the training data, performing singular value decomposition, and determining the number of principal components according to singular values; the magnetic field sensing unit is used for carrying out normalization processing on test data and judging whether error abnormity exists or not by judging whether statistical magnitude exceeds a control threshold value or not; two magnetic field sensing units with normal errors are screened out by combining all the magnetic field sensing units in pairs and comparing the statistical magnitude, and all the abnormal magnetic field sensing units are identified by judging whether the combined statistical magnitude exceeds a threshold value or not. The beneficial effects of the invention are that the self-repairing of the measurement error of the magnetic array type current sensor is realized, and the long-term measurement stability of the magnetic array type current sensor is guaranteed to a certain extent.

Description

Technical Field

[0001] This invention relates to the field of current sensor error measurement and repair technology, specifically a data-driven self-repair method and system for measuring errors of a magnetic array current sensor. Background Technology

[0002] The mainstream current sensing equipment used in power systems includes electromagnetic current transformers, fiber optic current transformers, shunts, Rogowski coils, and zero-flux current sensors. These devices exhibit significant differences in performance regarding current measurement bandwidth, insulation, and cost. For example, electromagnetic current transformers are primarily used for low-frequency (<1kHz) current measurement and demonstrate outstanding long-term measurement stability. However, as voltage levels increase, the complexity of the insulation structure, cost, and weight increase dramatically. Furthermore, they cannot measure DC current. Fiber optic current transformers offer clear performance advantages, enabling both AC and DC measurements. They feature wide bandwidth (DC to 100kHz), good insulation, strong anti-interference capabilities, and light weight. However, the core optical components used for magnetic field sensing are sensitive to environmental factors such as temperature and vibration, and their long-term measurement stability needs further improvement. They also have a higher cost. Shunts, based on the principle of resistance sampling, achieve wide-bandwidth (DC to MHz) current measurement. While measuring current can be done in various ways, it requires connection to an electrical circuit. Due to resistance heating, the measured current value cannot be further increased. Rogowski coils have a natural advantage in high-frequency current measurement, but require complex integration circuits. They also cannot measure DC current and have poor accuracy in low-frequency current measurement. Zero-flux current sensors are based on the zero-flux principle, generating compensating flux through a secondary coil. They achieve magnetic deflection detection based on the Hall effect and fluxgate principles. Through feedback circuit control, they ensure that the magnetic core is in a dynamic equilibrium state of zero flux. They can achieve AC and DC measurements and have high performance characteristics such as wide bandwidth, high precision, and low temperature. However, due to the need for complex feedback control circuits and compensation current, zero-flux current sensors are expensive, have complex structures, and consume a lot of power.

[0003] Magnetic array current sensors measure the magnetic field at multiple points in space within the conductor being measured, and calculate the measured current using a reverse inversion algorithm. Wide bandwidth and redundancy are its two main technical characteristics. Benefiting from the redundancy of the measured magnetic field information, compared to traditional single-source signal current sensing techniques, another yet-to-be-fully-explored technical feature of magnetic array current sensors is their self-monitoring and self-correcting capabilities regarding measurement errors. During long-term operation, influenced by environmental factors (such as temperature) and performance differences between different magnetic field sensing units, the measurement errors of some magnetic field sensing units in a magnetic array current sensor can drift significantly, leading to decreased or even out-of-tolerance accuracy in current measurement.

[0004] Therefore, there is an urgent need for a data-driven self-correction method and system for measurement errors of magnetic array current sensors to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide a data-driven self-repair method and system for measurement errors of magnetic array current sensors. It realizes the self-repair of measurement errors of magnetic array current sensors, and to a certain extent ensures the long-term measurement stability of magnetic array current sensors.

[0006] To achieve the above objectives, the present invention employs the following technical solution: On the one hand, a self-correction method for measurement errors of a magnetic array current sensor is provided, including the following steps: By combining offline and online calibration, the initial linear relationship coefficients between the output of the magnetic field sensing unit and the measured current are obtained, providing basic parameters for subsequent error analysis. Magnetic field data of a magnetic array current sensor under normal operating conditions are collected as training samples. The training data is normalized based on the initial linear relationship coefficients, a data matrix is ​​constructed and its covariance matrix is ​​calculated. The data is projected to the principal component subspace and the residual subspace through singular value decomposition to form a mathematical model that reflects the normal operating characteristics of the sensor. During the long-term operation of the sensor, real-time magnetic field data is collected as test samples. The test data is processed based on the same linear relationship coefficient and normalization method, projected onto the constructed residual subspace, and the Q statistic is calculated and compared with the preset control threshold to preliminarily determine whether there is any measurement error anomaly. When an anomaly is detected, a combined analysis strategy is initiated. By traversing all possible sensor pair combinations, the cumulative Q statistic value of each combination is calculated, and the two benchmark sensors that best meet the normal operating characteristics are selected. Using the benchmark sensor as a reference, each of the remaining sensors is combined with the benchmark sensor to perform Q statistics verification, thereby identifying all abnormal sensors with measurement errors. By removing measurement data from abnormal sensors, and using the measurement values ​​and linear relationship coefficients of the remaining normal sensors, the measured current value is recalculated through a weighted fusion algorithm, thereby achieving self-correction of measurement errors.

[0007] Preferably, the linear relationship coefficients are obtained through offline calibration and online calibration, including: During offline calibration, a high-precision current sensor is used to obtain the reference value of the measured current, and the output data of each magnetic field sensing unit is collected simultaneously to calculate the linear relationship coefficient between the output and the reference current. During online calibration, based on the position information of each magnetic field sensing unit, including coordinates and magnetic sensitivity angle, the measured current is inverted using a current inversion algorithm, and the linear relationship coefficient is corrected.

[0008] Preferably, the normalization process for the training data specifically includes: Based on the mean and standard deviation of the data from each magnetic field sensing unit in the training data, the training data is converted into standardized data with a mean of 0 and a standard deviation of 1 to eliminate the influence of the difference in the magnitude of data from different magnetic field sensing units on the analysis.

[0009] Preferably, determining the number of principal components based on singular values ​​includes: Calculate the cumulative contribution rate of singular values. When the cumulative contribution rate reaches a preset threshold, the corresponding number of singular values ​​is the number of principal components. The preset threshold ranges from 80% to 95%.

[0010] Preferably, the statistics for measurement error are constructed as follows: The statistic is used to characterize the projection amplitude of the test data in the residual subspace, reflecting the degree to which the measured value of the magnetic field sensing unit deviates from the normal linear correlation. Its value is the sum of squares of the components of the test data in the residual subspace.

[0011] Preferably, the determination of the control threshold includes: Threshold parameters are calculated based on the residual features of the training data, and control thresholds are determined by combining the critical values ​​of the normal distribution at a preset confidence level; the preset confidence level ranges from 90% to 99%.

[0012] Preferably, by combining all magnetic field sensing units in pairs and comparing statistical values, two magnetic field sensing units with normal errors are selected, including: Calculate the cumulative value of the statistics of each pair of magnetic field sensing units over the test data time length, and select the pair with the smallest cumulative value as the two magnetic field sensing units with normal error; the smallest cumulative value indicates that the linear correlation of the pair is closest to the normal state.

[0013] Preferably, the step of combining each of the remaining magnetic field sensing units with the two normal units, and identifying all abnormal magnetic field sensing units by whether the combined statistics exceed a threshold, includes: For each magnetic field sensing unit to be tested, it is combined with two normal units to form a ternary combination, and the statistics of the combination are calculated. If the statistics exceed the control threshold, the unit to be tested is judged to be abnormal in error; otherwise, it is judged to be normal in error.

[0014] Preferably, the step of recalculating the measured current using the measurement results from the normal unit includes: Based on the output data of the normal magnetic field sensing unit and its linear relationship coefficient with the measured current, the measured current is inverted through a multi-source data fusion algorithm; the multi-source data fusion algorithm includes weighted least squares method or optimization algorithm based on residual minimization.

[0015] On the other hand, a system is provided for implementing a self-correcting method for measurement errors of a magnetic array current sensor as described above, comprising: The calibration module is used to obtain the linear relationship coefficient between the output of the magnetic field sensing unit and the measured current through offline and online calibration. The training module is used to collect magnetic field data under normal operating conditions as training data, normalize the training data, calculate the covariance matrix and perform singular value decomposition, and determine the number of principal components based on the singular values ​​to separate the principal component subspace and the residual subspace. The monitoring module is used to collect and normalize test data during long-term operation, construct statistics and their control thresholds by combining the residual subspace, and determine whether there are error abnormal units by comparing the statistics with the thresholds. The anomaly identification module is used to filter out two normal units by pairwise combination based on a combination strategy, and then combine the remaining units with the two normal units respectively, and identify all abnormal units by statistical measures. The self-repair module is used to eliminate abnormal units and recalculate the measured current using the measurement results of normal units, thereby achieving error self-repair.

[0016] Preferably, the calibration module includes an offline calibration unit and an online calibration unit; The offline calibration unit is configured to connect to a high-precision reference sensor, synchronously acquire the output of the magnetic field sensing unit and the reference current, and calculate the linear relationship coefficient. The online calibration unit is configured to: acquire the position parameters of the magnetic field sensing unit, invert the measured current based on the magnetic field distribution model, and dynamically correct the linear relationship coefficients.

[0017] Preferably, the training module includes a data preprocessing unit and a subspace separation unit; The data preprocessing unit is configured to: normalize the training data to eliminate differences in data magnitude; The subspace separation unit is configured to: calculate the covariance matrix of the training data and perform singular value decomposition, determine the number of principal components based on the singular values, and separate the principal component subspace and the residual subspace.

[0018] Preferably, the monitoring module includes a test data processing unit and an anomaly early warning unit; The test data processing unit is configured to: normalize the real-time acquired test data and project it onto the residual subspace; The anomaly warning unit is configured to: calculate the statistics of the test data, compare them with the control threshold, and issue a warning that an anomaly exists if the threshold is exceeded.

[0019] Preferably, the anomaly identification module includes a combined analysis unit and an anomaly determination unit; The combined analysis unit is configured to: combine all magnetic field sensing units in pairs, calculate the cumulative statistical value of each combination, and select the two normal units with the smallest cumulative value. The anomaly determination unit is configured to: combine the remaining units with two normal units respectively, calculate the combined statistics and compare them with a threshold to determine the abnormal units.

[0020] Preferably, it also includes an extended application interface for adapting the system to error self-correction scenarios of other array sensors; the array sensors include distributed electric field sensors, temperature sensors, and pressure sensors.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes a self-correction method for magnetic array current sensors. Based on the redundancy characteristics of the magnetic field information measured by the magnetic array current sensor, the correlation between the magnetic field results measured by the magnetic field sensing units is clarified through physical information correlation analysis. Driven by data, the method identifies and eliminates magnetic field sensing units with significant measurement error drift online, thereby achieving self-correction of the measurement error of the magnetic array current sensor and ensuring the long-term measurement stability of the magnetic array current sensor to a certain extent. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of a current sensor based on a circular magnetic array, consisting of eight magnetic field sensing units, measuring current according to an embodiment of the present invention. Figure 3 This is a flowchart illustrating the implementation of a data-driven self-correction method for measurement errors of a magnetic array current sensor according to an embodiment of the present invention. Figure 4 This is a current measurement platform built in a laboratory according to an embodiment of the present invention; Figure 5 The design principle (a) and physical diagram (b) of a single magnetic field sensing unit according to an embodiment of the present invention are shown. Figure 6 The figures (a) and (b) are curves showing the variation of the amplitude and relative amplitude error of the fundamental phasor of the output voltage of all magnetic field sensing units under normal error conditions, according to an embodiment of the present invention. Figure 7 To simulate error drift by adjusting the gain and phase shift of the magnetic field sensing unit according to an embodiment of the present invention, the amplitude relative error and phase error of S1 and S2 both drift, while only the amplitude relative error of S3 drifts, and the measurement errors of other magnetic field sensing units are all in a normal state. The curves (a) showing the change of amplitude and amplitude relative error of the fundamental phasor of the output voltage of different magnetic field sensing units and the curves (b) showing the change of phase and phase error of the fundamental phasor of the output voltage of different magnetic field sensing units. Figure 8 The statistical variation curves of the amplitude (a) and phase (b) of the fundamental phasor of the output voltage of all magnetic field sensing units according to an embodiment of the present invention are shown. Figure 9 In an experiment to simulate error drift by adjusting the gain and phase shift of a magnetic field sensing unit according to an embodiment of the present invention, all magnetic field sensing units are combined in pairs, and the distribution cloud map of the summation of the statistical quantities of the amplitude (a) and phase (b) of the fundamental phasor of the output voltage of different combinations over time is obtained. Figure 10 According to an embodiment of the present invention, in order to simulate error drift test by changing the gain and phase shift of the magnetic field sensing unit, the amplitude and phase of the output voltage of different magnetic field sensing unit combinations are statistical change curves, the amplitude Q statistical change curve (a) and the phase statistical change curve (b). Figure 11 In an experiment to simulate error drift by adjusting the gain and phase shift of a magnetic field sensing unit according to an embodiment of the present invention, the relative error of the current amplitude (a) and the phase error (b) of the magnetic array current sensor before and after eliminating the magnetic field sensing unit with an abnormal error state are shown. Figure 12 To simulate the amplitude relative error change curve (a) and phase error change curve (b) of the magnetic field sensing unit measurement error drift by heating with a soldering iron according to an embodiment of the present invention. Figure 13 The present invention provides a statistical curve (a) of amplitude variation and a statistical curve (b) of phase variation in an error drift test by changing the operating temperature of the magnetic field sensing unit according to an embodiment of the present invention. Figure 14 In an error drift test simulating the operation temperature of a magnetic field sensing unit according to an embodiment of the present invention, the current measurement error of the magnetic array current sensor before and after eliminating the magnetic field sensing unit with an abnormal error state is calculated as follows: amplitude relative error (a) and phase error (b). Figure 15 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0023] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined in this application.

[0024] In this invention, terms such as "upper," "lower," "left," "right," "front," "back," "vertical," "horizontal," "side," and "bottom" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used only to facilitate the description of the structural relationships of the various components or elements of this invention and do not specifically refer to any component or element in this invention. They should not be construed as limiting the invention.

[0025] Example: like Figure 1 As shown, this embodiment provides a self-correction method for measurement errors of a magnetic array current sensor, including: By combining offline and online calibration, the initial linear relationship coefficients between the output of the magnetic field sensing unit and the measured current are obtained, providing basic parameters for subsequent error analysis. Magnetic field data of a magnetic array current sensor under normal operating conditions are collected as training samples. The training data is normalized based on the initial linear relationship coefficients, a data matrix is ​​constructed and its covariance matrix is ​​calculated. The data is projected to the principal component subspace and the residual subspace through singular value decomposition to form a mathematical model that reflects the normal operating characteristics of the sensor. During the long-term operation of the sensor, real-time magnetic field data is collected as test samples. The test data is processed based on the same linear relationship coefficient and normalization method, projected onto the constructed residual subspace, and the Q statistic is calculated and compared with the preset control threshold to preliminarily determine whether there is any measurement error anomaly. When an anomaly is detected, a combined analysis strategy is initiated. By traversing all possible sensor pair combinations, the cumulative Q statistic value of each combination is calculated, and the two benchmark sensors that best meet the normal operating characteristics are selected. Using the benchmark sensor as a reference, each of the remaining sensors is combined with the benchmark sensor to perform Q statistics verification, thereby identifying all abnormal sensors with measurement errors. After removing measurement data from faulty sensors, the measured current value is recalculated using a weighted fusion algorithm based on the remaining normal sensor measurements and their linear relationship coefficients, thus achieving self-correction of measurement errors. like Figure 2As shown, eight magnetic field sensing units are arranged in a circular array around the conductor being measured. The magnetic sensitivity directions of the sensing units are approximately tangential to the circular array to obtain the maximum magnetic field signal, thus achieving a superior signal-to-noise ratio. After the magnetic array current sensor is installed, the relationship between the measured current and the measured magnetic field can be obtained by running the inverse magnetic field problem algorithm software, thereby completing the online calibration of the magnetic array current sensor. The phasor of the measured current at a single frequency (f) and the phasors of the corresponding magnetic induction intensity measured by all magnetic field sensing units satisfy the following relationship: (1) In the formula, Let N be the magnetic flux density phasor measured by the k-th magnetic field sensing unit, and N be the number of magnetic field sensing units. The measured current phasor; for and The proportional coefficient between them.

[0026] The coefficients in equation (1) The calibration can be determined through two methods: offline calibration and online calibration. Offline calibration involves using a high-precision current sensor to measure the phasor of the current being measured and obtaining the phasor of the magnetic induction intensity measured by all magnetic field sensing units, thereby calculating the coefficients. For online calibration, the position information (including coordinates and magnetic angle) of all magnetic field sensing units is known. Based on the current inversion algorithm, the conductor position is calculated from the measured magnetic induction intensity, and then the coefficients are calculated. .

[0027] As shown in equation (1), if the measurement error of a single or multiple magnetic field sensing units drifts significantly, the measurement accuracy of the magnetic array current sensor will inevitably deteriorate or even exceed the tolerance. Therefore, it is necessary to evaluate the error state of each magnetic field sensing unit to ensure the long-term accuracy and reliability of the current measurement results. like Figure 1 As shown, all magnetic field sensing units simultaneously measure the magnetic field around the same measured current; therefore, the magnetic fields they measure have "common origin." Based on this simple physical information constraint, it can be deduced that the magnetic fields measured by all magnetic field sensing units satisfy a linear correlation, which can be described by the following formula: (2) (3) In the formula and These represent the amplitude and phase of the magnetic induction intensity measured by the phasor theory of the k-th magnetic field sensing unit, respectively. and The amplitude and phase of the measured current phasor; The proportionality coefficient between the amplitude of the magnetic flux density phasor theoretical result measured by the k-th magnetic field sensing unit and the amplitude of the measured current phasor needs to be specified. The coefficients defined in equation (1) The theoretical value, and It was obtained through experimental calibration.

[0028] From equations (2) and (3), it can be seen that the amplitude and phase of the theoretical results of the magnetic flux density measured by all magnetic field sensing units are linearly related. The relationship between the actual magnetic flux density measured by the k-th magnetic field sensing unit and the corresponding theoretical result is expressed as follows: (4) (5) In the formula, and These represent the relative amplitude error and phase error of the k-th magnetic field sensing unit, respectively. Their values ​​are obtained through calibration and remain unchanged during long-term measurements. and Δ These are the changes (or drifts) in the relative amplitude error and phase error of the k-th magnetic field sensing unit, respectively. Their magnitudes are affected by the operating environment of the magnetic field sensing unit. Therefore, the error drifts of all magnetic field sensing units are random, independent and uncorrelated.

[0029] As can be seen from equation (4) or equation (5), the magnetic induction intensity measured by the k-th magnetic field sensing unit includes two parts, the first part being... or ,because and Constant and unchanging, while the corresponding measurements of all magnetic field sensing units or The results are linearly correlated; therefore, the first part of the magnetic field results measured by all magnetic field sensing units satisfies a linear correlation. The second part is... or Their values ​​are determined by the error drift of the magnetic field sensing unit. Since the error drift of the magnetic field sensing unit is random and independent, the second part of the magnetic field results measured by all magnetic field sensing units is independent and uncorrelated.

[0030] When a magnetic field sensing unit in a magnetic array current sensor experiences measurement error drift, the linear correlation between the magnetic fields measured by all the magnetic field sensing units will change. If this change in linear correlation can be detected, it can be determined whether a magnetic field sensing unit in the magnetic array current sensor has an abnormal error state. In this embodiment of the invention, the following steps are used to achieve highly sensitive detection of changes in linear correlation: After the magnetic array current sensor is installed and calibrated, all magnetic field sensing units are in a normal error state. Magnetic field data measured by all magnetic field sensing units over a period of time are acquired and used to construct a training data matrix SL×N (S is a matrix composed of the amplitude or phase of the magnetic field measured by all magnetic field sensing units over a period of time; L is the length of the magnetic field data measured by a single magnetic field sensing unit; N is the number of magnetic field sensing units). This training data matrix is ​​then normalized as follows: (6) In the formula, All elements are Column vectors; Let Σ be the row vector formed by the average of each column of matrix S; Σ is the diagonal matrix formed by the standard deviation of each column of matrix S.

[0031] Normalized matrix The covariance matrix is ​​calculated using the following formula: (7) In the formula, the operator Cov represents the covariance matrix of the solution matrix.

[0032] The singular value decomposition of the covariance matrix Z is expressed as follows: (8) In the formula, U and V are the left and right matrices, respectively; Λ is the eigenvalue matrix, located on its diagonal. The value on , while all other values ​​are 0.

[0033] The number of pivot elements m is determined using the following criterion: (9) In the formula, To determine the control threshold for the number of principal components, specifically, one can choose... .

[0034] Based on the number of principal components, the principal subspace and the residual subspace can be determined. Specifically, the first m columns of matrix V are established as the principal subspace P, while the (m+1)th to Nth columns of matrix V are established as the residual subspace R.

[0035] During the long-term operation of the magnetic array current sensor, the magnetic field results measured by all its magnetic field sensing units are used as the test data matrix X, and the test data is normalized, i.e.: (10) Where I represents the time length during which each magnetic field sensing unit measures the magnetic field data.

[0036] Furthermore, by combining the principal component subspace and the residual subspace, the test data matrix can be decomposed into principal component components and residual components, that is: (11) In the formula, As a principal component, its expression is as follows: (12) and The residual component is expressed as follows: (13) As can be seen from equation (4) or (5), the magnetic field result measured by each magnetic field sensing unit includes two parts: one part is the magnetic field measured when the magnetic field sensing unit is in a normal error state, and the other part is the increment of the measured magnetic field introduced by error drift after the measurement error state of the magnetic field sensing unit changes. Through principal component decomposition, the magnetic field result measured when the magnetic field sensing unit is in a normal error state is mapped to the principal component. The measured magnetic field increment introduced by error drift is mapped to the residual component. Therefore, by evaluating whether the residual components obtained from the decomposition have changed significantly, highly sensitive detection of changes in the linear correlation between the magnetic fields measured by the magnetic field sensing unit can be achieved.

[0037] In principal component analysis, the Q statistic is generally used to assess whether the residual components have changed. Its expression is as follows: (14) In the formula, Indicates the first Q statistics at each test point; j is used to index the j-th magnetic field sensing unit; This represents the element corresponding to the i-th row and j-th column of the residual component matrix.

[0038] Define the control threshold for the statistic ,Right now: (15) In the formula, ( )and The expressions are as follows: (16) (17) And coefficient Is it a normal distribution at a confidence level of The corresponding critical value.

[0039] When the Q statistic significantly exceeds its control threshold Then it has a probability of The possibility that the linear correlation between the magnetic fields measured by all magnetic field sensing units has changed indicates that there are magnetic field sensing units in the magnetic array current sensor with abnormal measurement errors (i.e., drifting), but they cannot be accurately located. To address this, based on the redundancy of the number of magnetic field sensing units in the magnetic array current sensor (i.e., most magnetic field sensing units have normal measurement error states, while only a few have drifted), this paper proposes an online identification method for magnetic field sensing units with abnormal error states based on a combination strategy. The specific implementation steps of this method are as follows: First, two magnetic field sensing units with normal error states are identified online. Specifically, all magnetic field sensing units are paired. If both magnetic field sensing units are in normal error states, the linear correlation between the magnetic fields measured by them remains unchanged, and the calculated Q statistic will inevitably be very small. However, when a magnetic field sensing unit with an abnormal error state appears in this pair, it will inevitably cause a change in the linear correlation between the magnetic fields measured by the two magnetic field sensing units, leading to a sharp increase in the Q statistic that significantly exceeds the control threshold. Therefore, by calculating the Q statistic for each combination at each measurement point... and for all The test data are summed over the time period, and the combination with the minimum sum of the Q statistics of all combinations is selected as the two magnetic field sensing units with normal error status.

[0040] Then, each of the remaining magnetic field sensing units is combined with two selected magnetic field sensing units with normal error states. If the Q statistic of the combination (containing two magnetic field sensing units with normal error states and one magnetic field sensing unit with an unknown error state) significantly exceeds its control threshold, it can be determined that the measurement error of the magnetic field sensing unit with an unknown error state has drifted. In this way, all magnetic field sensing units with abnormal error states can be located one by one.

[0041] Once all magnetic field sensing units with abnormal error states are identified, the magnetic field measured by the remaining magnetic field sensing units with normal measurement errors is used, and the current is recalculated according to equation (1), thereby realizing the self-repair of the measurement error of the magnetic array current sensor, which to a certain extent ensures the long-term stability of the measurement accuracy of the magnetic array current sensor.

[0042] The feasibility and effectiveness of the proposed self-correction method for measurement errors in a magnetic array current sensor are now illustrated with an example.

[0043] The proposed method was subjected to specific simulation analysis and verification.

[0044] like Figure 3 As shown, firstly, based on the correlation analysis of the inherent physical information of the magnetic field sensing unit cluster, it is concluded that the magnetic field results measured by different magnetic field sensing units satisfy a linear correlation. Then, based on the above linear correlation, principal component analysis theory is used to map the linear correlation satisfied by the magnetic field results measured by different magnetic field sensing units to the residual subspace, thereby realizing the mathematical modeling of the linear correlation. Secondly, an online identification method for magnetic field sensing units with measurement error drift based on a combination strategy is proposed to identify all magnetic field sensing units with abnormal error states online. Finally, the magnetic field results measured by the magnetic field sensing units with abnormal error states are eliminated, and only the magnetic field results measured by the remaining magnetic field sensing units in normal error states are used to calculate the current, thereby ensuring the accuracy of current measurement, that is, realizing the online self-correction of measurement error of magnetic array current sensor.

[0045] like Figure 4 As shown, the designed magnetic array current sensor consists of 8 magnetic field sensing units. Figure 4 The illustration shows all magnetic field sensing units evenly spaced on a circle with a radius of 35mm, with the magnetic sensitivity direction of each unit pointing tangentially to the circle. The design principle and physical diagram of each magnetic field sensing unit are shown below. Figure 4 As shown, a magnetoresistive chip of model TMR2102 manufactured by Multidimensional Technology Co., Ltd. is used as the magnetic sensing element of the magnetic field sensing unit. The linear range of magnetic field measurement of this chip is ±3mT, the sensitivity is 49mV / V / mT, and each magnetoresistive chip is powered by 5V DC.

[0046] like Figure 5As shown in (a), the magnetoresistive element converts the sensed magnetic field into an output voltage, which is amplified by the differential operational amplifier AD8220. The magnetoresistive element employs a Wheatstone full-bridge design. Due to the non-ideal symmetry of the Wheatstone full-bridge structure, its output voltage may have a small DC bias even without an external magnetic field. Therefore, the reference potential of AD8220 is finely adjusted by adjusting the resistor divider composed of R1 and R2 to compensate for the DC bias component of the magnetic field sensing unit's output signal. The output signal of AD8220 is further amplified by an operational amplifier, then fed into an RC circuit, and finally sampled by a digital sampling module. Figure 5 (b) shows the actual fabricated magnetic field sensing unit. In subsequent current measurement experiments, it can be adjusted... Figure 5 In (a), resistors R4 and R5 are used to adjust the gain and phase shift of the magnetic field sensing unit to simulate the amplitude error and phase error drift that occur in the actual operation of the magnetic field sensing unit, respectively.

[0047] like Figure 6 As shown, this is a curve illustrating the variation of the fundamental phasor amplitude and relative amplitude error of the output voltage of all magnetic field sensing units under normal error conditions, according to an embodiment of the present invention (e.g., Figure 6 (a) and the curves showing the phase and phase error of the fundamental phasor of the output voltage of all magnetic field sensing units (as shown in Figure 1). Figure 6 (b) As shown. To simulate the error drift of the magnetic field sensing unit, the resistors R5 of the magnetic field sensing units S1 and S2 are adjusted (as shown). Figure 5 As shown in the figure, adjusting the resistor R2 of S3 changes both its gain and phase shift, while adjusting the resistor R2 of S3 changes only its gain, but the phase shift remains unchanged. The amplitude and phase of the output voltage of each magnetic field sensing unit, as well as the corresponding amplitude relative error and phase error, are shown in the figure. Figure 7 As shown, the sampling interval at this time is 1 second. (From...) Figure 7 It can be seen that, since the amplitude of the measured current is still modulated into a dynamically changing triangular wave, it is consistent with... Figure 6 Similarly, the amplitude and phase of the output voltage of all magnetic field sensing units change dynamically according to triangular and sawtooth waves, respectively. Figure 7 It can be seen that the relative amplitude errors of magnetic field sensing units S1 and S2 increase to -1.34% and -0.88% respectively in the negative error direction, while the amplitude error of magnetic field sensing unit S3 first increases to 1.44% in the positive error direction, and then changes to 0.55% in the negative direction, while the amplitude errors of other magnetic field sensing units remain unchanged. As for the phase errors of each magnetic field sensing unit, the phase errors of S1 and S2 both increase to -0.10 and -0.11 respectively in the negative error direction, while the phase errors of other magnetic field sensing units remain unchanged. Figure 7The amplitude and phase data of the fundamental phasor of the output voltage of all magnetic field sensing units shown are used as the test data matrix X. The Q statistics of the amplitude or phase of the fundamental phasor of the output voltage of all magnetic field sensing units as a function of time are calculated according to equations (10) to (13). The corresponding results are as follows: Figure 7 The magenta curve in the figure is shown; at the same time, the corresponding control threshold Qα is calculated according to equations (14) to (17), as shown in the figure. Figure 7 As shown by the blue dashed line. It should be noted that the Q statistic essentially reflects the incremental change in the amplitude or phase of the magnetic field measured by the magnetic field sensing unit due to the error variation of the sensing unit. Because the amplitude error of the magnetic field sensing unit is characterized as a relative error, when the amplitude of the measured current is dynamically modulated into a triangular wave, the Q statistic will also change dynamically with time. However, the change in the phase error of the magnetic field sensing unit is a constant absolute quantity; therefore, the Q statistic corresponding to the phase is also constant, rather than changing dynamically with time. For example... Figure 8 As shown, the Q statistic calculated from the amplitude or phase of the fundamental phasor of the output voltage of all magnetic field sensing units exceeds the corresponding control threshold. Therefore, it can be determined that there are magnetic field sensing units in the magnetic array current sensor whose error state has changed.

[0048] To accurately locate all magnetic field sensing units where error states change, all magnetic field sensing units are paired, and the Q-statistic of the amplitude or phase of the output voltage of each pair is calculated over time. The results are presented as follows: Figure 9 The cloud map shown is presented. Figure 9 (a) It can be seen that, since the relative amplitude errors of S1, S2, and S3 have all changed, the Q statistic of their combination with other magnetic field sensing units is significantly larger; similarly, since the phase errors of S1 and S2 have changed, Figure 9 In (b), the Q-statistics of their combination with other magnetic field sensing units are also significantly larger. Thus, to a certain extent, it is possible to determine... Figure 9 First, determine which magnetic field sensing units have experienced error shifts. Second, select the combination corresponding to the minimum sum of the Q statistics over time among all pairwise combinations, namely S4 and S5. Combine each of the other magnetic field sensing units with S4 and S5 respectively, and calculate the Q statistics of the amplitude or phase of the output voltage of the above combinations (including two magnetic field sensing units S4 and S5 with normal error states and one magnetic field sensing unit with an unknown error state). The corresponding results are as follows: Figure 10 As shown. By Figure 10 (a) It can be seen that the Q statistics of the output voltage amplitudes of the combinations (S1,S4,S5), (S2,S4,S5), and (S3,S4,S5) significantly exceed the corresponding control thresholds. Therefore, it can be determined that error drift has occurred in S1, S2, and S3. Figure 10(b) It can be seen that the Q statistics of the phase of the output voltages of the combinations (S1,S4,S5) and (S2,S4,S5) significantly exceed the corresponding control thresholds. Therefore, it can be determined that phase error drift has occurred in S1 and S2. Thus, online accurate identification of all magnetic field sensing units with changes in error states has been achieved.

[0049] like Figure 11 As shown, in an experiment simulating error drift by adjusting the gain and phase shift of a magnetic field sensing unit according to an embodiment of the present invention, the relative error of the current amplitude of a magnetic array current sensor before and after eliminating a magnetic field sensing unit with an abnormal error state is (e.g., ...). Figure 11 (a) and phase error (as shown) Figure 11 (b) shows that by eliminating all magnetic field sensing units with drifting (amplitude or phase) errors, and only using magnetic field sensing units with normal error states to measure the magnetic field results, and recalculating the current using equation (1), the self-correction of measurement errors in the magnetic array current sensor is achieved. Figure 11 (a) It can be seen that since the relative amplitude errors of S1, S2, and S3 have all drifted, and when the current is calculated using the results measured by all magnetic field sensing units, the relative amplitude error of the measured current (magenta curve) has also drifted, first increasing to 0.3% in the positive error direction, and then increasing to -0.4% in the negative error direction; however, after removing the contributions of S1, S2, and S3, the relative amplitude error of the measured current (blue curve) remains within the error range of ±0.1%. Regarding phase error, as... Figure 11 As shown in (b), when the magnetic field results measured by all magnetic field sensing units are used to calculate the current, the phase error drift of the measured current reaches -0.03 rad. However, after eliminating the magnetic field sensing units with abnormal error states, the phase error of the magnetic array current sensor is always within the range of ±1 mrad.

[0050] like Figure 12 As shown, this is a curve illustrating the relative error change in the amplitude of the measurement error drift of a magnetic field sensing unit, simulated by heating with a soldering iron according to an embodiment of the present invention (e.g., Figure 12 (a) shown) and the phase error variation curve (as shown) Figure 12 (b) As shown. Figure 12 As shown in (a), the relative amplitude error of S6 first increases to -0.35% in the negative error direction, and then gradually returns to the initial state. This is because the magnetoresistive chip has a negative temperature coefficient of sensitivity. In addition, the operating temperature of the two magnetic field sensing units (S5 and S7) adjacent to S6 also changes to a certain extent, and their relative amplitude errors also fluctuate slightly accordingly. Figure 12As shown in (b), the phase error of all magnetic field sensing units remains unchanged, which indicates that the phase error of the magnetoresistive chip is not affected by temperature changes.

[0051] like Figure 13 The figure shows the Q-statistic variation curve of the amplitude in an error drift simulation test by changing the operating temperature of the magnetic field sensing unit according to an embodiment of the present invention (e.g., Figure 13 (a) and the Q-statistic change curve of the phase (as shown in Figure 1). Figure 13 (b) shown); by Figure 13 (a) It can be seen that the Q-statistic of the amplitude of combination (S6,S2,S4) significantly exceeds the control threshold, indicating that the relative amplitude error of S6 has significantly drifted. Meanwhile, the Q-statistics of the amplitudes of combinations (S5,S2,S4) and (S7,S2,S4) slightly exceed the control threshold, indicating that S5 and S7 have also experienced some degree of amplitude error drift. Therefore, the magnitude of the error drift can be reflected by the degree to which the Q-statistic exceeds the control threshold. Figure 13 (b) It can be seen that the Q statistics of the phase of all combined output voltages are basically near the control threshold and there is no obvious fluctuation. Therefore, it can be determined that the phase error of all magnetic field sensing units has not drifted.

[0052] Before and after removing the magnetic field sensing units (S5, S6, and S7) with abnormal error states, the current measurement error of the magnetic array current sensor is as follows: Figure 14 As shown. By Figure 14 (a) It can be seen that when the measured magnetic field results of all magnetic field sensing units are used to calculate the current, the relative error of the current measured by the magnetic array current sensor first increases in the negative direction (the maximum error drift reaches 0.08%), and then recovers in the positive direction. This trend is consistent with the trend of the relative error of S6 (see Figure 12 After eliminating all magnetic field sensing units whose error states have changed, the relative error of the current amplitude measured by the magnetic array current sensor remains within ±0.02%. Figure 14 (b) It can be seen that since temperature has no effect on the phase error of the magnetic field sensing unit, the phase error of the current measured by the magnetic array current sensor remains unchanged regardless of whether the magnetic field sensing unit with abnormal error state is removed.

[0053] like Figure 15 As shown, this embodiment also provides a self-correcting system for measurement errors of a magnetic array current sensor, including: The calibration module is used to obtain the linear relationship coefficient between the output of the magnetic field sensing unit and the measured current through offline and online calibration. The training module is used to collect magnetic field data under normal operating conditions as training data, normalize the training data, calculate the covariance matrix and perform singular value decomposition, and determine the number of principal components based on the singular values ​​to separate the principal component subspace and the residual subspace. The monitoring module is used to collect and normalize test data during long-term operation, construct statistics and their control thresholds by combining the residual subspace, and determine whether there are error abnormal units by comparing the statistics with the thresholds. The anomaly identification module is used to filter out two normal units by pairwise combination based on a combination strategy, and then combine the remaining units with the two normal units respectively, and identify all abnormal units by statistical measures. The self-repair module is used to eliminate abnormal units and recalculate the measured current using the measurement results of normal units, thereby achieving error self-repair.

[0054] The calibration module includes an offline calibration unit and an online calibration unit. The offline calibration unit is configured to connect to a high-precision reference sensor, synchronously acquire the output of the magnetic field sensing unit and the reference current, and calculate the linear relationship coefficient. The online calibration unit is configured to: acquire the position parameters of the magnetic field sensing unit, invert the measured current based on the magnetic field distribution model, and dynamically correct the linear relationship coefficients.

[0055] The training module includes a data preprocessing unit and a subspace separation unit; The data preprocessing unit is configured to normalize the training data to eliminate differences in data magnitude. The subspace separation unit is configured as follows: calculate the covariance matrix of the training data and perform singular value decomposition, determine the number of principal components based on the singular values, and separate the principal component subspace and the residual subspace.

[0056] The monitoring module includes a test data processing unit and an anomaly early warning unit; The test data processing unit is configured to: normalize the real-time acquired test data and project it onto the residual subspace; The abnormal warning unit is configured to: calculate the statistics of the test data, compare them with the control threshold, and issue a warning that there is an abnormal unit if the threshold is exceeded.

[0057] The anomaly identification module includes a combined analysis unit and an anomaly determination unit; The combined analysis unit is configured as follows: all magnetic field sensing units are combined in pairs, the cumulative statistical value of each combination is calculated, and the two normal units with the smallest cumulative value are selected. The anomaly detection unit is configured as follows: combine the remaining units with two normal units respectively, calculate the combined statistics and compare them with the threshold to determine the abnormal units.

[0058] It also includes an extended application interface for adapting the system to error self-correction scenarios of other array sensors; the array sensors include distributed electric field sensors, temperature sensors, and pressure sensors.

[0059] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for self-correcting measurement errors of a magnetic array current sensor, characterized in that, Includes the following steps: The linear relationship coefficients between the output of each magnetic field sensing unit and the measured current are obtained by combining offline and online calibration. Magnetic field data of a magnetic array current sensor under normal operating conditions are collected as training samples. The training data is normalized based on the linear relationship coefficients, a data matrix is ​​constructed and its covariance matrix is ​​calculated. Singular value decomposition is performed on the covariance matrix, and the number of principal components is determined based on the singular values, thereby separating the principal component subspace and the residual subspace, and establishing a mathematical model of the normal operating characteristics of the sensor. During the long-term operation of the sensor, magnetic field data is collected in real time as test samples. The test data is processed based on the same linear relationship coefficient and normalization method. The test data is projected onto the residual subspace, the Q statistic is calculated, and the Q statistic is compared with a preset control threshold to determine whether there is a magnetic field sensing unit with abnormal measurement error. When an anomaly is detected, a combination analysis strategy is adopted to iterate through all pairwise combinations of magnetic field sensing units, calculate the cumulative value of the Q statistic of each combination over time, and select the two magnetic field sensing units with the smallest cumulative Q statistic as the reference sensors with normal error. Then, each of the remaining magnetic field sensing units is combined with the two reference sensors respectively, and the Q statistic of the combination is calculated. If the value exceeds the control threshold, the magnetic field sensing unit is determined to be abnormal. The measurement data of the magnetic field sensing units identified as abnormal are removed, and the measured values ​​of the remaining normal magnetic field sensing units and their linear relationship coefficients are used to recalculate the measured current value through a weighted fusion algorithm, thereby achieving self-correction of measurement errors.

2. The self-correcting method for measurement errors of a magnetic array current sensor according to claim 1, characterized in that, Linearity coefficients were obtained through offline and online calibration, including: During offline calibration, a high-precision current sensor is used to obtain the reference value of the measured current, and the output data of each magnetic field sensing unit is collected simultaneously to calculate the linear relationship coefficient between the output and the reference current. During online calibration, based on the position information of each magnetic field sensing unit, including coordinates and magnetic sensitivity angle, the measured current is inverted using a current inversion algorithm, and the linear relationship coefficient is corrected.

3. The self-correction method for measurement error of a magnetic array current sensor according to claim 1, characterized in that, The normalization process for the training data specifically involves: Based on the mean and standard deviation of the data from each magnetic field sensing unit in the training data, the training data is converted into standardized data with a mean of 0 and a standard deviation of 1 to eliminate the influence of the difference in the magnitude of data from different magnetic field sensing units on the analysis.

4. The self-correction method for measurement error of a magnetic array current sensor according to claim 3, characterized in that, The method of determining the number of principal components based on singular values ​​includes: Calculate the cumulative contribution rate of singular values. When the cumulative contribution rate reaches a preset threshold, the corresponding number of singular values ​​is the number of principal components. The preset threshold ranges from 80% to 95%.

5. The self-correction method for measurement error of a magnetic array current sensor according to claim 1, characterized in that, The Q statistic is calculated as follows: The statistic is used to characterize the projection amplitude of the test data in the residual subspace, reflecting the degree to which the measured value of the magnetic field sensing unit deviates from the normal linear correlation. Its value is the sum of squares of the components of the test data in the residual subspace.

6. The self-correction method for measurement error of a magnetic array current sensor according to claim 1, characterized in that, The determination of the control threshold includes: Threshold parameters are calculated based on the residual features of the training data, and control thresholds are determined by combining the critical values ​​of the normal distribution at a preset confidence level; the preset confidence level ranges from 90% to 99%.

7. The self-correcting method for measurement errors of a magnetic array current sensor according to claim 1, characterized in that, By iterating through all pairwise combinations of magnetic field sensing units, the two magnetic field sensing units with the smallest cumulative Q-statistic value are selected as the baseline sensors with normal error, including: Calculate the cumulative value of the statistics of each pair of magnetic field sensing units over the test data time length, and select the pair with the smallest cumulative value as the two magnetic field sensing units with normal error; the smallest cumulative value indicates that the linear correlation of the pair is closest to the normal state.

8. The self-correction method for measurement error of a magnetic array current sensor according to claim 1, characterized in that, The Q statistic for the calculated combination includes: For each magnetic field sensing unit to be tested, it is combined with two normal units to form a ternary combination, and the statistics of the combination are calculated. If the statistics exceed the control threshold, the unit to be tested is determined to have an abnormal error; otherwise, it is determined to have a normal error.

9. The self-correction method for measurement error of a magnetic array current sensor according to claim 1, characterized in that, The recalculation of the measured current value includes: Based on the output data of the normal magnetic field sensing unit and its linear relationship coefficient with the measured current, the measured current is inverted through a multi-source data fusion algorithm; the multi-source data fusion algorithm includes weighted least squares method or optimization algorithm based on residual minimization.

10. A system for implementing a self-correcting method for measurement errors of a magnetic array current sensor as described in any one of claims 1-9, characterized in that, include: The calibration module is used to obtain the linear relationship coefficient between the output of the magnetic field sensing unit and the measured current through offline and online calibration. The training module is used to collect magnetic field data under normal operating conditions as training data, normalize the training data, calculate the covariance matrix and perform singular value decomposition, and determine the number of principal components based on the singular values ​​to separate the principal component subspace and the residual subspace. The monitoring module is used to collect and normalize test data during long-term operation, construct statistics and their control thresholds by combining the residual subspace, and determine whether there are error abnormal units by comparing the statistics with the thresholds. The anomaly detection module is used to filter out two normal units by pairwise combination based on a combination strategy, and then combine the remaining units with the two normal units respectively, and identify all abnormal units by statistical measures. The self-repair module is used to eliminate abnormal units and recalculate the measured current using the measurement results of normal units, thereby achieving error self-repair.

11. The self-correcting system for measurement errors of a magnetic array current sensor according to claim 10, characterized in that, The calibration module includes an offline calibration unit and an online calibration unit; The offline calibration unit is configured to connect to a high-precision reference sensor, synchronously acquire the output of the magnetic field sensing unit and the reference current, and calculate the linear relationship coefficient. The online calibration unit is configured to: acquire the position parameters of the magnetic field sensing unit, invert the measured current based on the magnetic field distribution model, and dynamically correct the linear relationship coefficients.

12. The self-correcting system for measurement errors of a magnetic array current sensor according to claim 10, characterized in that, The training module includes a data preprocessing unit and a subspace separation unit; The data preprocessing unit is configured to: normalize the training data to eliminate differences in data magnitude; The subspace separation unit is configured to: calculate the covariance matrix of the training data and perform singular value decomposition, determine the number of principal components based on the singular values, and separate the principal component subspace and the residual subspace.

13. The self-correcting system for measurement errors of a magnetic array current sensor according to claim 10, characterized in that, The monitoring module includes a test data processing unit and an anomaly early warning unit; The test data processing unit is configured to: normalize the real-time acquired test data and project it onto the residual subspace; The anomaly warning unit is configured to: calculate the statistics of the test data, compare them with the control threshold, and issue a warning that an anomaly exists if the threshold is exceeded.

14. The self-correcting system for measurement errors of a magnetic array current sensor according to claim 10, characterized in that, The anomaly identification module includes a combined analysis unit and an anomaly determination unit; The combined analysis unit is configured to: combine all magnetic field sensing units in pairs, calculate the cumulative statistical value of each combination, and select the two normal units with the smallest cumulative value. The anomaly determination unit is configured to: combine the remaining units with two normal units respectively, calculate the statistics of the combination and compare them with the threshold to determine the abnormal unit.

15. A self-correcting system for measurement errors of a magnetic array current sensor according to claim 10, characterized in that, It also includes an extended application interface for adapting the system to error self-correction scenarios of other array sensors; the array sensors include distributed electric field sensors, temperature sensors, and pressure sensors.

Images