A method for detecting the electrical performance of a high-precision current sensor

Abstract

The present invention relates to the fields of electrical engineering and sensor technology, and provides a method for detecting the electrical performance of a high-precision current sensor, including collecting the original output signal of the current sensor and performing multi-dimensional signal preprocessing; extracting multi-dimensional feature parameters from the preprocessed signal based on a dynamic feature extraction algorithm; establishing a dynamic compensation model according to the multi-dimensional feature parameters to generate a compensation control signal; injecting the compensation control signal into the current sensor under test to form a closed-loop feedback, and collecting the output signal after compensation; calculating the sensor performance degradation parameters according to the signal differences before and after compensation; and generating an electrical performance evaluation report based on the performance degradation parameters. The present invention can improve the measurement accuracy and reliability of the sensor, extend its service life, reduce the maintenance cost, and provide a strong guarantee for the safe and stable operation of power systems and electronic devices.

Description

Technical Field

[0001] The present invention relates to the technical fields of electrical engineering and sensor technology, and more specifically, the present invention relates to a method for detecting the electrical performance of a high-precision current sensor. Background Art

[0002] In modern power systems and electronic devices, high-precision current sensors are widely used in key links such as current measurement, fault detection, and system control. Existing methods for detecting current sensors mainly rely on static performance tests, and evaluate their performance by comparing the differences between the sensor outputs and standard signals. These methods usually include simple signal acquisition and preliminary error analysis, but often ignore the complex characteristics of sensors in dynamic environments, such as non-linear errors, temperature drift, and performance degradation during long-term use. Traditional detection technologies are difficult to accurately extract multi-dimensional characteristic parameters when dealing with dynamic signals, and are also unable to adjust the compensation model in real time to adapt to the dynamic changes of sensors. In addition, existing technologies lack effective multi-physical field coupling models when evaluating the performance degradation of sensors, and are unable to accurately predict the remaining service life of sensors.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the existing technologies: existing detection methods cannot comprehensively reflect the true performance of current sensors under dynamic conditions, lacking in-depth analysis and real-time compensation capabilities for the dynamic characteristics of sensors; at the same time, the evaluation of sensor performance degradation is not precise enough, and the failure time cannot be effectively predicted, resulting in difficulties in taking maintenance measures in advance in practical applications, affecting the stability and reliability of power systems. Summary of the Invention

[0004] The present invention provides a method for detecting the electrical performance of a high-precision current sensor, including:

[0005] S1. Collect the original output signal of the current sensor and perform multi-dimensional signal preprocessing;

[0006] S2. Extract multi-dimensional characteristic parameters from the preprocessed signal based on a dynamic feature extraction algorithm;

[0007] S3. Establish a dynamic compensation model according to the multi-dimensional characteristic parameters and generate a compensation control signal;

[0008] S4. Inject the compensation control signal into the current sensor under test to form a closed-loop feedback, and collect the output signal after compensation;

[0009] S5. Calculate the sensor performance degradation parameter according to the signal difference before and after compensation;

[0010] S6. Generate an electrical performance evaluation report based on the performance degradation parameter.

[0011] Furthermore, step S2 includes:

[0012] S21. Decompose the preprocessed signal using an improved time-frequency joint decomposition algorithm to obtain a time-domain feature vector and a frequency-domain feature vector , where represents the i-th time-domain feature parameter, represents the j-th frequency-domain feature parameter;

[0013] S22. Calculate the feature fusion weight matrix using the dynamic correlation degree analysis algorithm, where represents the correlation coefficient between the time-domain feature and the frequency-domain feature ;

[0014] S23. Calculate the multi-dimensional feature parameter matrix through a non-linear feature fusion formula.

[0015] Furthermore, step S3 includes:

[0016] S31. Construct a compensation function with a memory effect:

[0017]

[0018] where is the compensation control signal at time t, K is the dynamic gain coefficient, is the historical compensation influence factor, is the time interval, is the modulation angular frequency, is the phase compensation amount;

[0019] S32. Adjust the parameter group in real time through an online learning algorithm to match the spectral characteristics of the compensation signal with the non-linear error of the sensor.

[0020] Furthermore, the calculation of the performance degradation parameter in step S5 includes constructing an error evolution model of multi-physical field coupling:

[0021] S51. Define the dynamic error parameter:

[0022]

[0023] where is the composite error parameter at time t, is the compensated signal, is the original signal, is the change rate of the multi-dimensional feature parameter at time is the instantaneous error weight, is the cumulative error weight;

[0024] S52. Establish a degradation equation with multi-factor coupling:

[0025]

[0026] where D is the performance degradation parameter, is the aging attenuation coefficient, is the time decay factor, is the mutation sensitivity coefficient, is the temperature coupling coefficient, is the change in environmental temperature;

[0027] S53. Solve the degradation equation by the fourth-order Runge-Kutta method to obtain the time-varying performance degradation curve .

[0028] Furthermore, step S53 also includes:

[0029] S531. Construct a temperature compensation model as shown in the following formula:

[0030]

[0031] where T is the real-time temperature value, is the material thermal characteristic coefficient;

[0032] S532. Define the rules for dividing the degradation stage:

[0033] When it is determined to be in a healthy state;

[0034] When a warning state is triggered;

[0035] When it is determined to be in a failure state.

[0036] Furthermore, the dynamic correlation degree analysis algorithm in step S22 includes:

[0037] S221. Calculate the time-frequency feature mutual information entropy:

[0038]

[0039] where is the probability distribution function;

[0040] S222. Calculate the dynamic weight through a sliding time window:

[0041]

[0042] where, is the weight update rate coefficient, is the update time interval.

[0043] Furthermore, step S1 includes:

[0044] S11. Implement multi-modal signal enhancement:

[0045] Use an adaptive notch filter to eliminate power frequency interference

[0046] Use the empirical mode decomposition algorithm to separate the signal's intrinsic mode

[0047] S12. Perform environmental disturbance compensation:

[0048] Synchronously collect the three-dimensional electromagnetic field intensity vector Correct the original signal through the field strength compensation formula:

[0049]

[0050] where are the compensation coefficients for each axis, is the reference field strength value.

[0051] Furthermore, the online learning algorithm in step S32 includes:

[0052] S321. Construct a dual-objective optimization function as shown in the following formula:

[0053]

[0054]

[0055] where is the weight coefficient, is the average tracking error, is the maximum fluctuation amount of the compensation signal;

[0056] S322. Use the constrained particle swarm optimization algorithm to perform directional search in the parameter space and update the parameter group every 5 seconds.

[0057] Furthermore, step S4 includes:

[0058] S41. Design an anti-aliasing injection circuit, including:

[0059] A programmable gain instrumentation amplifier with a gain range of 60 dB to 100 dB;

[0060] A digital isolation device with an isolation voltage ≥ 2500 Vrms;

[0061] S42. Implement multi-rate signal synchronization:

[0062] Collect the original signal at a sampling rate of 200 kHz;

[0063] Inject a compensation signal at a 50 kHz update rate;

[0064] Implement time alignment of signals with different rates through an interpolation algorithm.

[0065] Furthermore, step S6 includes:

[0066] S61. Construct a three-dimensional health state map, where the three-dimensional health state map includes a time axis, a performance axis, and an environment axis. The time axis is the cumulative working time;

[0067] The performance axis is the degradation parameter;

[0068] The environment axis is a composite index of temperature, humidity, and vibration;

[0069] S62. Predict the remaining life through a deep residual network, as shown in the following formula:

[0070] where

[0071]

[0072] is the trained neural network model, is the failure threshold, is the degradation acceleration, is the performance degradation parameter at time t.

[0073] According to the above embodiments of the present invention, it has at least the following beneficial effects: The high-precision electrical performance detection method of the current sensor of the present invention can achieve a comprehensive evaluation of the dynamic performance of the current sensor. Through multi-dimensional signal preprocessing and dynamic feature extraction algorithms, it can accurately extract time-domain and frequency-domain characteristic parameters, providing richer information for subsequent performance analysis. Combining the dynamic compensation model and the closed-loop feedback mechanism can adjust the compensation signal in real time, effectively reducing the non-linear error and dynamic deviation of the sensor and improving the measurement accuracy. In addition, based on the error evolution model and performance degradation parameter calculation of multi-physical field coupling, the performance degradation degree of the sensor can be accurately evaluated, and its health state can be intuitively displayed through a three-dimensional health state map, providing a scientific basis for equipment maintenance.

[0074] At the same time, the present invention can also predict the remaining service life of the sensor through a deep residual network, providing strong support for the preventive maintenance of power systems and electronic devices. This method can not only improve the reliability of the current sensor, but also reduce the system downtime and maintenance costs caused by sensor failures. Brief Description of the Drawings

[0075] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understandable. In the drawings, several embodiments of the present invention are shown by way of example and not limitation, wherein:

[0076] Figure 1 It is a schematic flow chart of a method for detecting the electrical performance of a high-precision current sensor provided by an embodiment of the present invention. Detailed implementation manners

[0077] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and do not limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to convey the scope of the present invention fully to those skilled in the art.

[0078] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, device, equipment, method, or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: completely hardware, completely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0079] It should be noted that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0080] Refer to the following Figure 1 , Figure 1 It is a schematic flow chart of a method for detecting the electrical performance of a high-precision current sensor provided by an embodiment of the present invention. As Figure 1 shown, a method 100 for detecting the electrical performance of a high-precision current sensor includes:

[0081] S1. Collect the original output signal of the current sensor and perform multi-dimensional signal preprocessing;

[0082] S2. Extract multi-dimensional feature parameters from the preprocessed signal based on a dynamic feature extraction algorithm;

[0083] S3. Establish a dynamic compensation model according to the multi-dimensional feature parameters and generate a compensation control signal;

[0084] S4. Inject the compensation control signal into the current sensor under test to form a closed-loop feedback and collect the output signal after compensation;

[0085] S5. Calculate the sensor performance degradation parameter according to the signal difference before and after compensation;

[0086] S6. Generate an electrical performance evaluation report based on the performance degradation parameters.

[0087] It should be noted that the electrical performance detection method of the high-precision current sensor of the present invention first involves collecting the original output signal of the current sensor and performing multi-dimensional signal preprocessing. Here, multi-dimensional signal preprocessing refers to performing a series of processing operations on the collected original signal to remove factors affecting signal quality such as noise and interference, and at the same time enhancing the useful features of the signal. For example, the original output signal refers to the electrical signal generated by the current sensor in the actual working environment, and these signals may be affected by various factors such as electromagnetic interference and temperature changes. And multi-dimensional signal preprocessing is to improve the signal-to-noise ratio of the signal to make it more suitable for subsequent feature extraction and analysis.

[0088] Specifically, multi-dimensional signal preprocessing can include implementing multi-modal signal enhancement and performing environmental perturbation compensation. Multi-modal signal enhancement refers to using an adaptive notch filter to eliminate power frequency interference and using the empirical mode decomposition algorithm to separate the intrinsic mode of the signal. Power frequency interference usually refers to the interference of alternating current of 50Hz or 60Hz, which is relatively common in the power system. The adaptive notch filter can dynamically adjust the filter parameters according to the frequency characteristics of the signal, so as to effectively remove power frequency interference. The empirical mode decomposition algorithm is an adaptive signal processing method, which can decompose a complex signal into several intrinsic mode functions, and these intrinsic mode functions can better reflect the local characteristics of the signal. Environmental perturbation compensation is to synchronously collect the three-dimensional electromagnetic field intensity vector and correct the original signal using the field strength compensation formula. The three-dimensional electromagnetic field intensity vector includes the electromagnetic field intensity components in three directions. By measuring these components and combining the compensation formula, the original signal can be corrected to reduce the influence of environmental electromagnetic field changes on the signal.

[0089] Preferably, the notch frequency of the adaptive notch filter in multi-modal signal enhancement can be adjusted according to the frequency of the actual power frequency interference. For example, in an environment with 50Hz power frequency interference, the notch frequency can be set to 50Hz. The decomposition process of the empirical mode decomposition algorithm can be optimized according to the complexity of the signal to improve the decomposition efficiency and accuracy. In environmental perturbation compensation, the compensation coefficients of each axis can be calibrated according to the installation position of the sensor and the characteristics of the environmental electromagnetic field to ensure the accuracy of the compensation. In addition, other signal enhancement methods such as wavelet transform can also be used as a supplement to the adaptive notch filter to further improve the anti-interference ability of the signal.

[0090] In some embodiments, step S2 includes:

[0091] S21. Decompose the preprocessed signal using an improved time-frequency joint decomposition algorithm to obtain a time-domain feature vector and a frequency-domain feature vector , where represents the i-th time-domain feature parameter, represents the j-th frequency-domain feature parameter;

[0092] S22. Calculate the feature fusion weight matrix based on the dynamic correlation degree analysis algorithm , where represents the time-domain feature and the frequency-domain feature correlation coefficient;

[0093] S23. Calculate the multi-dimensional feature parameter matrix through the non-linear feature fusion formula .

[0094] It should be noted that in the present invention, step S2 includes decomposing the preprocessed signal by using an improved time-frequency joint decomposition algorithm to obtain a time-domain feature vector and a frequency-domain feature vector, calculating a feature fusion weight matrix based on the dynamic correlation degree analysis algorithm, and finally calculating a multi-dimensional feature parameter matrix through the non-linear feature fusion formula. The time-frequency joint decomposition algorithm here is a signal processing method that can analyze the signal in both the time domain and the frequency domain simultaneously, so as to extract the features of the signal more comprehensively. The time-domain feature vector and the frequency-domain feature vector respectively describe the characteristics of the signal in time and frequency, while the dynamic correlation degree analysis algorithm is used to calculate the correlation degree between these features to determine their weights in the fusion process. The calculation of this weight matrix is crucial for subsequent feature fusion because it determines the importance of different features in the final feature parameter matrix.

[0095] Specifically, the improved time-frequency joint decomposition algorithm is optimized on the basis of the traditional time-frequency analysis method to better adapt to the characteristics of the current sensor signal. For example, the time-domain feature vector can include parameters such as the amplitude, mean, and variance of the signal, which reflect the variation law of the signal in time. The frequency-domain feature vector can include parameters such as the frequency distribution and power spectral density of the signal, which reveal the characteristics of the signal in frequency. The dynamic correlation degree analysis algorithm determines the correlation coefficient between the time-domain feature and the frequency-domain feature by calculating the mutual information entropy between them. The mutual information entropy is an index to measure the mutual dependence degree between two variables, which can help to determine which features have strong correlations. When calculating the feature fusion weight matrix, the weights can be assigned according to these correlation coefficients, so that the fused feature parameter matrix can more accurately reflect the comprehensive characteristics of the signal.

[0096] Preferably, the time-frequency joint decomposition algorithm can adopt the combination of wavelet transform and Hilbert-Huang transform. Wavelet transform can provide good time-frequency resolution, while Hilbert-Huang transform is more suitable for processing non-stationary signals. The combination of the two can better extract the characteristics of the current sensor signal. When calculating the dynamic correlation degree, the concept of a sliding time window can be introduced to adapt to the dynamic changes of the signal. The size of the sliding time window can be adjusted according to the frequency characteristics of the signal. For example, for high-frequency signals, the time window can be set smaller to capture the changes of the signal more quickly. In addition, other dynamic correlation degree analysis methods, such as correlation degree analysis based on neural networks, can be adopted to further improve the accuracy and adaptability of weight calculation.

[0097] In some embodiments, step S3 includes:

[0098] S31. Construct a compensation function with a memory effect:

[0099]

[0100] where is the compensation control signal at time t, K is the dynamic gain coefficient, is the historical compensation influence factor, is the time interval, is the modulation angular frequency, is the phase compensation amount;

[0101] S32. Real-time adjust the parameter group through an online learning algorithm to make the compensation signal match the spectral characteristics of the sensor non-linear error.

[0102] It should be noted that in the present invention, step S3 involves constructing a compensation function with a memory effect and real-time adjusting the parameter group through an online learning algorithm to make the compensation signal match the spectral characteristics of the sensor non-linear error. Here, the compensation function is a mathematical model used to generate a compensation control signal to correct the measurement error of the current sensor. The memory effect means that the compensation function can consider historical compensation information to better adapt to the dynamic changes of the sensor. The online learning algorithm is an adaptive algorithm that can dynamically adjust the parameters of the compensation function according to real-time data to ensure the matching degree between the compensation signal and the sensor error characteristics.

[0103] Specifically, the construction of the compensation function takes into account multiple parameters, including the dynamic gain coefficient, historical compensation influence factor, time interval, modulation angular frequency, and phase compensation amount, etc. The dynamic gain coefficient is used to adjust the amplitude of the compensation signal to adapt to different error magnitudes; the historical compensation influence factor reflects the influence of past compensation on current compensation, which helps the compensation function better remember the historical state; the time interval and modulation angular frequency are used to control the time characteristics and frequency characteristics of the compensation signal; the phase compensation amount is used to correct the phase deviation of the signal. The online learning algorithm adjusts these parameters by constructing a two-objective optimization function, and the optimization objectives usually include minimizing the average tracking error and limiting the maximum fluctuation amount of the compensation signal. The constrained particle swarm optimization algorithm is a commonly used online learning algorithm, which can perform efficient search in the parameter space to find the optimal parameter combination.

[0104] Preferably, the dynamic gain coefficient can be dynamically adjusted according to the error range of the sensor. For example, when the error is large, the gain is increased to accelerate the compensation speed. The historical compensation influence factor can be determined through experiments to ensure that the compensation function can effectively utilize historical information. The time interval can be set according to the sampling rate of the sensor. For example, when the sampling rate is 1 kHz, the time interval can be set to 1 ms. The modulation angular frequency can be selected according to the operating frequency range of the sensor to ensure that the frequency characteristics of the compensation signal match those of the error signal. The phase compensation amount can be determined through a phase analysis algorithm to correct the phase deviation of the sensor output signal. In addition, other online learning algorithms, such as adaptive filtering algorithms or deep learning algorithms, can also be used to further improve the accuracy and adaptability of parameter adjustment.

[0105] In some embodiments, the calculation of the performance degradation parameter in step S5 includes constructing an error evolution model of multi-physical field coupling:

[0106] S51. Define the dynamic error parameter:

[0107]

[0108] Among them, is the composite error parameter at time t, is the compensated signal, is the original signal, is the change rate of the multi-dimensional feature parameter at time is the instantaneous error weight, is the cumulative error weight;

[0109] S52. Establish a degradation equation of multi-factor coupling:

[0110]

[0111] where D is the performance degradation parameter, is the aging attenuation coefficient, is the time attenuation factor, is the mutation sensitivity coefficient, is the temperature coupling coefficient, is the change in ambient temperature;

[0112] S53. Solve the degradation equation by the fourth-order Runge-Kutta method to obtain the time-varying performance degradation curve .

[0113] It should be noted that in the present invention, step S5 involves constructing an error evolution model for multi-physical field coupling to calculate the performance degradation parameter. The multi-physical field coupling here refers to considering the performance changes of the sensor under the combined action of multiple physical factors (such as temperature, time, mutation factors, etc.). The performance degradation parameter is an index to measure the decline of the sensor performance over time. By establishing an error evolution model, the dynamic error changes of the sensor can be quantitatively analyzed, thereby providing a basis for evaluating its health status.

[0114] Specifically, the definition of the dynamic error parameter combines the instantaneous error and the cumulative error, where the instantaneous error weight and the cumulative error weight are used to balance the importance of the two. The instantaneous error weight reflects the influence of the error at the current moment, while the cumulative error weight takes into account the cumulative effect of the error over time. The degradation equation with multi-factor coupling further considers factors such as the aging attenuation coefficient, the time attenuation factor, the mutation sensitivity coefficient, and the temperature coupling coefficient. These coefficients respectively correspond to the aging, time dependence, sensitivity to mutations, and the influence of temperature changes on the performance of the sensor. By solving the degradation equation by the fourth-order Runge-Kutta method, a time-varying performance degradation curve can be obtained, which can intuitively show the change trend of the sensor performance over time.

[0115] Preferably, the temperature compensation model can be adjusted according to the working ambient temperature of the sensor, and the material thermal characteristic coefficient can be obtained through experimental measurement to ensure the accuracy of the model. The degradation stage division rule can be set according to actual application requirements. For example, the thresholds for the healthy state, the warning state, and the failure state can be determined according to the performance requirements and safety standards of the sensor. In addition, other physical factors (such as humidity, vibration, etc.) can be introduced to further improve the multi-physical field coupling model to more comprehensively evaluate the performance degradation of the sensor.

[0116] In some embodiments, step S53 further includes:

[0117] S531. Construct a temperature compensation model as shown in the following formula:

[0118]

[0119] where T is the real-time temperature value, is the thermal characteristic coefficient of the material;

[0120] S532. Define the rules for dividing the degradation stage:

[0121] When it is determined to be in a healthy state;

[0122] When it triggers a warning state;

[0123] When it is determined to be in a failure state.

[0124] It should be noted that in the present invention, step S53 further refines the calculation process of the performance degradation parameters, especially introducing a temperature compensation model and rules for dividing the degradation stage. The temperature compensation model is used to correct the influence of temperature changes on the performance degradation parameters to ensure the accuracy of performance evaluation under different ambient temperatures. The rules for dividing the degradation stage divide the health state of the sensor into three stages: healthy, warning, and failure according to the numerical range of the performance degradation parameters, so as to take corresponding maintenance measures in a timely manner.

[0125] Specifically, the material thermal characteristic coefficients in the temperature compensation model are parameters determined according to the thermophysical properties of the sensor material. These coefficients can be obtained through experimental measurements and are used to describe the influence of temperature changes on the sensor performance. The real-time temperature value refers to the actual temperature in the sensor working environment and is measured by a temperature sensor. The thresholds (such as and ) in the rules for dividing the degradation stage are preset according to the performance indicators and reliability requirements of the sensor and are used to judge the health state of the sensor. For example, when the performance degradation parameter D is less than D1, the sensor is in a healthy state; when D is between and , it triggers a warning state; when D is greater than or equal to , the sensor is considered to have failed.

[0126] Preferably, the material thermal characteristic coefficients can be determined by conducting a temperature cycle experiment on the sensor. During the experiment, the performance changes of the sensor at different temperatures are recorded, and then these coefficients are obtained by fitting. The real-time temperature value can be collected by a high-precision temperature sensor and its accuracy is ensured through data calibration. The thresholds in the rules for dividing the degradation stage can be adjusted according to the actual application scenario of the sensor. For example, in applications with higher reliability requirements, the value of can be appropriately reduced to give an early warning. In addition, compensation models for other environmental factors (such as humidity, vibration, etc.) can be introduced to further improve the comprehensiveness and accuracy of performance evaluation.

[0127] In some embodiments, the dynamic correlation analysis algorithm in step S22 includes:

[0128] S221. Calculate the time-frequency feature mutual information entropy:

[0129]

[0130] where is the probability distribution function;

[0131] S222. Calculate the dynamic weight through a sliding time window:

[0132]

[0133] where is the weight update rate coefficient, is the update time interval.

[0134] It should be noted that the dynamic correlation analysis algorithm in step S22 of the present invention includes calculating the time-frequency feature mutual information entropy and calculating the dynamic weight through a sliding time window. The time-frequency feature mutual information entropy here is an index that measures the degree of mutual dependence between time-domain features and frequency-domain features, and it can help determine which features have strong correlations. The sliding time window is a dynamic update mechanism used to adjust the weight according to the real-time changes of the signal, so as to adapt to the dynamic characteristics of the signal.

[0135] Specifically, the calculation of the time-frequency feature mutual information entropy is based on the probability distribution function, which reflects the information sharing degree between time-domain features and frequency-domain features. The probability distribution function describes the probability distribution of the occurrence of feature values, and by calculating the mutual information entropy, the correlation degree between two features can be quantified. The size of the sliding time window and the weight update rate coefficient are key parameters in the dynamic correlation analysis algorithm. The size of the sliding time window determines the response speed of the algorithm to signal changes, while the weight update rate coefficient controls the influence degree of new data on the weight. These parameters can be adjusted according to the characteristics of the signal and the application scenario to optimize the performance of the algorithm.

[0136] Preferably, the size of the sliding time window can be selected according to the frequency characteristics of the signal. For high-frequency signals, a smaller time window can capture the dynamic changes of the signal faster; for low-frequency signals, a larger time window can reduce the influence of noise. The weight update rate coefficient can be adjusted according to the stability of the signal. For example, in an environment where the signal changes rapidly, the update rate can be increased to improve the adaptability of the algorithm. In addition, other dynamic weight update mechanisms, such as the weight update method based on an adaptive filter, can also be adopted to further improve the flexibility and accuracy of the algorithm.

[0137] In some embodiments, step S1 includes:

[0138] S11. Implement multi-modal signal enhancement:

[0139] Use an adaptive notch filter to eliminate power frequency interference

[0140] Use the empirical mode decomposition algorithm to separate the signal's intrinsic modes

[0141] S12. Perform environmental disturbance compensation:

[0142] Synchronously collect the three-dimensional electromagnetic field intensity vector Correct the original signal through the field strength compensation formula:

[0143]

[0144] where, are the compensation coefficients for each axis, is the reference field strength value.

[0145] It should be noted that in the present invention, step S1 involves implementing multi-modal signal enhancement and performing environmental disturbance compensation. Here, multi-modal signal enhancement refers to improving the quality and reliability of signals through various signal processing techniques, while environmental disturbance compensation refers to measuring and correcting the influence of external environmental factors on signals to reduce the interference of these factors on the performance detection of current sensors. These operations are the basis for ensuring the accuracy of subsequent signal processing and performance evaluation.

[0146] Specifically, multi-modal signal enhancement includes using an adaptive notch filter to eliminate power frequency interference and using the empirical mode decomposition algorithm to separate the signal's intrinsic modes. Power frequency interference generally refers to the interference generated by alternating current in the power system, and its frequency is generally 50Hz or 60Hz. The adaptive notch filter can dynamically adjust the filtering parameters according to the characteristics of the signal to effectively remove the interference of this specific frequency. The empirical mode decomposition algorithm is an adaptive signal decomposition method that can decompose complex signals into several intrinsic mode functions, and these intrinsic mode functions can better reflect the local characteristics of the signal. In environmental disturbance compensation, the three-dimensional electromagnetic field intensity vector is synchronously collected, and the original signal is corrected using the field strength compensation formula. The three-dimensional electromagnetic field intensity vector includes the electromagnetic field intensity components in three directions. By measuring these components and combining the compensation formula, the original signal can be corrected to reduce the influence of environmental electromagnetic field changes on the signal. The compensation coefficients for each axis are parameters pre-calibrated according to the installation position of the sensor and the characteristics of the environmental electromagnetic field, and are used to adjust the amplitude of the correction.

[0147] Preferably, the notch frequency of the adaptive notch filter can be precisely adjusted according to the frequency of the actual power frequency interference. For example, in an environment with 50 Hz power frequency interference, the notch frequency can be set to 50 Hz, and the notch depth can be adjusted in real time according to the dynamic changes of the signal. The decomposition process of the empirical mode decomposition algorithm can be optimized according to the complexity of the signal. For example, the decomposition efficiency and accuracy can be improved by adjusting the number of decomposition iterations. In environmental disturbance compensation, the axial compensation coefficients can be determined by experimental methods. For example, the output signals of the sensor are measured under different environmental electromagnetic field conditions, and the optimal compensation coefficients are obtained by least squares fitting. In addition, other signal enhancement methods, such as wavelet transform, can be used as a supplement to the adaptive notch filter to further improve the anti-interference ability of the signal.

[0148] In some embodiments, the online learning algorithm in step S32 includes:

[0149] S321. Construct a dual-objective optimization function as shown in the following formula:

[0150]

[0151]

[0152] where is the weight coefficient, is the average tracking error, is the maximum fluctuation amount of the compensation signal;

[0153] S322. Use the constrained particle swarm optimization algorithm to perform a directional search in the parameter space, and update the parameter group every 5 seconds.

[0154] It should be noted that the online learning algorithm in step S32 of the present invention includes constructing a dual-objective optimization function and using the constrained particle swarm optimization algorithm to perform a directional search in the parameter space. The dual-objective optimization function here is a mathematical model used to optimize multiple objectives simultaneously, such as minimizing the average tracking error and limiting the maximum fluctuation amount of the compensation signal. The constrained particle swarm optimization algorithm is an optimization algorithm based on swarm intelligence, which searches for the optimal solution by simulating the behavior of a particle swarm. This algorithm can efficiently find the optimal parameter combination that satisfies the constraint conditions in the parameter space, so as to realize the dynamic adjustment of the compensation model parameters.

[0155] Specifically, the weight coefficients (such as weight coefficient 1 and weight coefficient 2) in the dual-objective optimization function are used to balance the importance of different optimization objectives. The average tracking error refers to the difference between the compensation signal and the actual error. By minimizing this error, the accuracy of compensation can be improved. The maximum fluctuation amount of the compensation signal reflects the stability of the compensation signal. Limiting this fluctuation amount can avoid the impact of excessive fluctuations in the compensation signal on the system. The constrained particle swarm optimization algorithm searches for the optimal solution by defining the update rules for the velocity and position of particles. In the parameter space, each particle represents a potential solution, and the particle swarm gradually approaches the optimal solution through information sharing and individual learning. The parameters of this algorithm, such as the inertia weight and learning factors in the velocity and position update formulas of particles, can be adjusted according to specific problems to improve the search efficiency and convergence speed.

[0156] Preferably, the weight coefficients can be adjusted according to the requirements in actual applications. For example, in scenarios with higher requirements for compensation accuracy, the weight of the average tracking error can be increased; while in scenarios with higher requirements for system stability, the weight of the maximum fluctuation amount of the compensation signal can be increased. The inertia weight of the constrained particle swarm optimization algorithm can be dynamically adjusted. For example, a larger inertia weight can be adopted at the initial stage of the search to accelerate the search speed, and the inertia weight can be reduced at the later stage of the search to improve the search accuracy. The learning factors can be dynamically adjusted according to the diversity and convergence of the group to balance the global search ability and the local search ability. In addition, other optimization algorithms, such as genetic algorithms or simulated annealing algorithms, can be used as alternatives to the constrained particle swarm optimization algorithm to further improve the performance and adaptability of parameter optimization.

[0157] In some embodiments, step S4 includes:

[0158] S41. Design an anti-aliasing injection circuit, including:

[0159] A programmable gain instrumentation amplifier with a gain range of 60 dB to 100 dB;

[0160] A digital isolation device with an isolation voltage ≥ 2500 Vrms;

[0161] S42. Implement multi-rate signal synchronization:

[0162] Collect the original signal at a sampling rate of 200 kHz;

[0163] Inject the compensation signal at an update rate of 50 kHz;

[0164] Achieve time alignment of signals with different rates through an interpolation algorithm.

[0165] It should be noted that in the present invention, step S4 involves designing an anti-aliasing injection circuit and implementing multi-rate signal synchronization. Here, the anti-aliasing injection circuit is a circuit design used to prevent signal aliasing during the sampling process, which can ensure the accurate injection and acquisition of the compensation signal. Multi-rate signal synchronization refers to aligning signals in time at different sampling rates to ensure the temporal consistency between the compensation signal and the original signal. These two operations are key steps for achieving closed-loop feedback and precise performance evaluation.

[0166] Specifically, the anti-aliasing injection circuit includes a programmable gain instrumentation amplifier and a digital isolation device. The programmable gain instrumentation amplifier is used to adjust the amplitude of the compensation signal, and its gain range can be set according to actual application requirements. For example, in the present invention, the gain range is 60 dB to 100 dB. The digital isolation device is used to isolate the electrical connection between the compensation signal and the original signal to prevent interference between signals, and its isolation voltage should meet the system safety requirements. For example, the isolation voltage ≥ 2500 Vrms. In multi-rate signal synchronization, the sampling rate of the original signal and the update rate of the compensation signal can be selected according to the dynamic characteristics of the sensor and system requirements. For example, the original signal is sampled at a rate of 200 kHz, and the compensation signal is injected at an update rate of 50 kHz. Through an interpolation algorithm, the time alignment of signals at different rates can be achieved to ensure that the compensation signal can act accurately on the original signal.

[0167] Preferably, the gain of the programmable gain instrumentation amplifier can be dynamically adjusted according to the error range of the sensor. For example, when the error is large, the gain is increased to improve the compensation effect. The selection of the digital isolation device should consider its isolation performance and reliability. For example, a high-voltage digital isolation chip is used. In multi-rate signal synchronization, the interpolation algorithm can adopt methods such as linear interpolation or spline interpolation. The specific selection depends on the characteristics and accuracy requirements of the signal. For example, for smoothly changing signals, linear interpolation may be sufficient; while for rapidly changing signals, spline interpolation can provide higher accuracy. In addition, other synchronization technologies, such as phase-locked loop synchronization, can also be adopted to further improve the accuracy and stability of signal synchronization.

[0168] In some embodiments, step S6 includes:

[0169] S61. Construct a three-dimensional health status map, where the three-dimensional health status map includes a time axis, a performance axis, and an environment axis.

[0170] The time axis is the cumulative working time.

[0171] The performance axis is the degradation parameter.

[0172] The environment axis is the composite index of temperature, humidity, and vibration.

[0173] S62. Predict the remaining life through a deep residual network, as shown in the following formula:

[0174]

[0175] where is the trained neural network model, is the failure threshold, is the degradation acceleration, is the performance degradation parameter at time t.

[0176] It should be noted that in the present invention, step S6 involves constructing a three-dimensional health state map and predicting the remaining life through a deep residual network. Here, the three-dimensional health state map is a visualization tool for showing the health state of the current sensor. It includes a time axis, a performance axis, and an environment axis, and can intuitively reflect the state of the sensor under different times, performance levels, and environmental conditions. The deep residual network is a deep learning model for predicting the remaining service life of the sensor according to the performance degradation parameter of the sensor. By learning a large amount of sensor data, this model can accurately predict the failure time of the sensor and provide a scientific basis for maintenance and replacement.

[0177] Specifically, the time axis in the three-dimensional health state map represents the cumulative working time of the sensor, the performance axis represents the performance degradation parameter, and the environment axis is a composite index including factors such as temperature, humidity, and vibration. The settings of these axes can be adjusted according to actual application requirements. For example, the unit of the time axis can be hours, days, or months, the range of the performance axis can be set according to the performance index of the sensor, and the composite index of the environment axis can be calculated by weighted summation. The inputs of the deep residual network include the performance degradation parameter, the degradation acceleration, and the failure threshold, etc., and these parameters can be obtained from the actual operation data of the sensor. The output of the network is the remaining service life of the sensor, and the trained model can predict new sensor data.

[0178] Preferably, the three-dimensional health state map can be implemented through visualization software, such as using the Matplotlib library of Python or the D3.js library of JavaScript. In the calculation of the composite index of the environment axis, weights can be assigned according to the influence degree of different environmental factors on the sensor performance. For example, temperature may have a greater influence on the sensor performance, and a higher weight can be assigned. The training of the deep residual network can adopt a large amount of historical sensor data and be carried out in a supervised learning manner. When predicting the remaining life, uncertainty analysis can be introduced, such as estimating the confidence interval of the prediction through the Monte Carlo method. In addition, other deep learning models, such as the recurrent neural network (RNN) or the long short-term memory network (LSTM), can be used as alternatives to the deep residual network to further improve the accuracy and adaptability of the prediction.

[0179] The above-mentioned embodiments of the present invention have the following beneficial effects: By collecting the original output signal of the current sensor and performing multi-dimensional signal preprocessing, power frequency interference and environmental disturbances can be effectively eliminated, and the signal quality can be improved. The improved time-frequency joint decomposition algorithm is used to extract multi-dimensional characteristic parameters, which can more comprehensively reflect the dynamic characteristics of the sensor. Based on the dynamic correlation analysis algorithm, the characteristic fusion weight matrix is calculated, which can further optimize the fusion effect of the characteristic parameters and provide more accurate inputs for the subsequent establishment of the compensation model. A compensation function with a memory effect is constructed, and the parameters are adjusted in real time through an online learning algorithm, so that the spectral characteristics of the compensation signal can be accurately matched with the non-linear error of the sensor, thereby improving the measurement accuracy and stability of the sensor. In addition, by calculating the performance degradation parameters through the multi-physical field coupling error evolution model, the performance changes of the sensor can be accurately evaluated, providing a scientific basis for the maintenance and replacement of the equipment.

[0180] In the closed-loop feedback process, designing an anti-aliasing injection circuit and implementing multi-rate signal synchronization can ensure the accurate injection and acquisition of the compensation signal, further improving the reliability of the system. By solving the degradation equation with the fourth-order Runge-Kutta method, a time-varying performance degradation curve can be obtained, providing a more intuitive reference for the health status monitoring of the sensor. Constructing a three-dimensional health status map and a remaining life prediction model based on a deep residual network can realize the visual display and life prediction of the sensor's health status, providing strong support for the preventive maintenance of power systems and electronic devices. These improvement measures can jointly improve the performance evaluation accuracy and reliability of the current sensor, and reduce the system downtime and maintenance costs caused by sensor failures.

[0181] Furthermore, the storage medium of the embodiment of the present application stores program instructions capable of implementing all the above methods. Among them, the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, or terminal devices such as computers, servers, mobile phones, and tablets.

[0182] The above description is only some preferred embodiments of the present invention and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the embodiments of the present invention.

Claims

1. A method for detecting the electrical performance of a high-precision current sensor, characterized in that It includes the following steps: S1. Collect the original output signal of the current sensor and perform multi-dimensional signal preprocessing; S2. Extract multi-dimensional feature parameters from the preprocessed signal based on the dynamic feature extraction algorithm; S3. Establish a dynamic compensation model according to the multi-dimensional feature parameters and generate a compensation control signal; S4. Inject the compensation control signal into the measured current sensor to form a closed-loop feedback and collect the compensated output signal; S5. Calculate the sensor performance degradation parameter according to the signal difference before and after compensation; S6. Generate an electrical performance evaluation report based on the performance degradation parameter; Among them, the calculation of the performance degradation parameter in S5 includes constructing an error evolution model of multi-physical field coupling: S51. Define the dynamic error parameter as shown in the following formula: ； Among them, is the composite error parameter at time t, is the compensated signal, is the original signal, is the change rate of the multi-dimensional feature parameter at time is the instantaneous error weight, is the cumulative error weight; S52. Establish a degradation equation of multi-factor coupling as shown in the following formula: ； where D is the performance degradation parameter, is the aging attenuation coefficient, is the time attenuation factor, is the mutation sensitivity coefficient, is the temperature coupling coefficient, is the environmental temperature change; S53. Solve the degradation equation by the fourth-order Runge-Kutta method to obtain the time-varying performance degradation curve .

2. The method according to claim 1, wherein Step S2 includes: S21. Decompose the preprocessed signal using an improved time-frequency joint decomposition algorithm to obtain a time-domain feature vector and a frequency-domain feature vector , where represents the i-th time-domain feature parameter, represents the j-th frequency-domain feature parameter; S22. Calculate the feature fusion weight matrix based on the dynamic correlation analysis algorithm , where represents the time-domain feature and the frequency-domain feature correlation coefficient; S23. Calculate the multi-dimensional feature parameter matrix through the non-linear feature fusion formula.

3. The method according to claim 1, wherein Step S3 includes: S31. Construct a compensation function with memory effect as shown in the following formula: ； Among them, is the compensation control signal at time t, K is the dynamic gain coefficient, is the historical compensation influence factor, is the time interval, is the modulation angular frequency, is the phase compensation amount; S32. Adjust the parameter set in real time through an online learning algorithm to match the spectral characteristics of the compensation signal and the sensor non-linear error.

4. The method according to claim 3, characterized in that Step S53 also includes: S531. Construct a temperature compensation model as shown in the following formula: ； where T is the real-time temperature value, is the material thermal characteristic coefficient; S532. Define the degradation stage division rule: When it is determined to be in a healthy state; When it triggers the warning state; When it is determined to be a failure state.

5. The method according to claim 4, wherein The dynamic correlation analysis algorithm of step S22 includes: S221. Calculate the time-frequency feature mutual information entropy as shown in the following formula: ； wherein, is a probability distribution function; is the time-domain feature and the frequency-domain feature of the joint probability distribution, is the time-domain feature of the marginal probability distribution, is the frequency-domain feature of the marginal probability distribution; S222. Calculate the dynamic weight through the sliding time window: ； Among them, is the weight update rate coefficient, is the update time interval.

6. The method according to claim 1, wherein Step S1 includes: S11. Implement multi-modal signal enhancement, including: Adopt an adaptive notch filter to eliminate power frequency interference; Use the empirical mode decomposition algorithm to separate the signal intrinsic mode; S12. Perform environmental disturbance compensation, including: Synchronously collect the three-dimensional electromagnetic field strength vector And correct the original signal through the field strength compensation formula as shown in the following formula: ； Among them, is the compensation coefficient in the i-axis direction, is the reference field strength value, is the corrected signal after environmental compensation, is the original signal, is the field strength value in the i-axis direction.

7. The method according to claim 3, wherein The online learning algorithm of step S32 includes: S321. Construct a double-objective optimization function as shown in the following formula: ； ； Among them, is the weight coefficient, is the average tracking error, is the maximum fluctuation amount of the compensation signal; S322. Adopt the constrained particle swarm optimization algorithm to perform directional search in the parameter space and update the parameter group every 5 seconds.

8. The method according to claim 7, characterized in that Step S4 includes: S41. Design an anti-aliasing injection circuit, and the anti-aliasing injection circuit includes: A programmable gain instrumentation amplifier with a gain range of 60dB to 100dB; A digital isolation device with an isolation voltage ≥ 2500Vrms; S42. Implement multi-rate signal synchronization: Collect the original signal at a sampling rate of 200kHz; Inject the compensation signal at an update rate of 50kHz; Realize the time alignment of signals with different rates through the interpolation algorithm.

9. The method according to claim 8, wherein Step S6 includes: S61. Construct a three-dimensional health state map, and the three-dimensional health state map includes a time axis, a performance axis, and an environment axis, wherein, the time axis is the cumulative working time; the performance axis is the degradation parameter; the environment axis is the composite index of temperature, humidity, and vibration; S62. Predict the remaining life through the deep residual network as shown in the following formula: ； Among them, is a trained neural network model, is the failure threshold, is the degradation acceleration, is the predicted remaining life, is the performance degradation parameter at time t.

Images