Mutual inductor verification master-slave clock error correction method

Abstract

The invention provides a mutual inductor verification master-slave clock error correction method, which comprises the following steps of: analyzing a specific reason for generating a deviation between a master clock and a slave clock in a mutual inductor verification system, and completing mathematical modeling of the error between the master clock and the slave clock according to the reason for the error; obtaining an initial deviation between the master clock and the slave clock of the mutual inductor; on the basis, the error of the master clock and the slave clock is corrected by using a Kalman filtering and RBF neural network adaptive control (RBF-PID) method, so that the deviation of the master clock and the slave clock is reduced, and the verification result of the mutual inductor is more accurate. Example analysis shows that the Kalman filtering has good filtering performance, and the RBF-PID algorithm has good adaptability and high convergence speed. An experiment and simulation combined example verifies that the method has good convergence precision and is suitable for actual production in the power industry.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power system automation, and particularly relates to a method for constructing a master-slave clock error model and adaptively controlling and adjusting a neural network in a mutual inductor calibration. BACKGROUND

[0002] A synchronous time system is an important part of a digital substation. The protection devices and control devices of the intermediate layer of the digital substation, especially the merging units of the process layer, cannot be separated from the synchronous time information, and ensuring the accurate synchronization of the devices of the digital substation is an important basis for the stable operation of the digital substation. As the most important information source in the digital substation, it is crucial to ensure the time synchronization of the current transformer. The traditional clock synchronization technology mainly uses the global positioning system (GPS) to time each device. This method is greatly affected by environmental factors and the hardware performance of the receiver, and has poor accuracy. On the basis of GPS timing, IRIG-B coding is introduced to form a pulse string by transmitting the second pulse and time information, which can reduce the influence of environmental factors and hardware performance to a certain extent, but this method has hidden dangers in terms of safety. In recent years, with the continuous progress of network information technology, the clock synchronization technology has gradually changed from code time to network time. Network time usually follows the IEEE 1588 protocol, also known as the precise time protocol (PTP). This method has lower requirements for hardware, but is greatly affected by network transmission delay. When facing an environment with large network fluctuations and obvious noise interference, this method has great limitations.

[0003] In view of the problems existing in the clock synchronization technology in the mutual inductor calibration process, a master-slave clock error correction algorithm for mutual inductor calibration is proposed in the present application. SUMMARY

[0004] The present application aims to at least solve one of the above technical problems or at least provide a useful commercial option. To this end, the present application proposes a master-slave clock error correction method for mutual inductor calibration.

[0005] The specific technical solution of the present application is to propose a master-slave clock error correction method for mutual inductor calibration, which comprises the following steps:

[0006] Step A. Analyze the specific reasons for the deviation of the master-slave clock in the mutual inductor calibration system, and complete the mathematical modeling of the master-slave clock error according to the error reasons;

[0007] Step B. Obtain the initial deviation between the master-slave clocks of the mutual inductor;

[0008] Step C. The method of Kalman filter and RBF-PID is used to correct the master-slave clock error, and more accurate transformer calibration results are obtained.

[0009] For step A, it comprises: analyzing the specific reasons for the deviation of the master-slave clock in the transformer calibration system, and completing the mathematical modeling of the master-slave clock error according to the error reasons, comprising the following steps:

[0010] Step A1: analyze the specific process of transformer calibration, and determine that the error causes are concentrated in two aspects: theoretical error and transmission error;

[0011] Step A2: analyze the temperature drift phenomenon of GPS crystal oscillator, and establish a theoretical error model of the master-slave clock of the transformer calibration;

[0012] Step A3: analyze the transmission process of the collected data in the transformer calibration process, and establish a transmission error model of the master-slave clock of the transformer calibration.

[0013] For step B, it comprises: building a transformer calibration platform and obtaining the initial error of the master-slave clock.

[0014] For step C, it comprises: using Kalman filter and RBF-PID to correct the master-slave clock error, comprising the following steps:

[0015] Step C1: use Kalman filter to filter out the process noise in the error model, and obtain effective waveform information;

[0016] Step C2: based on step C1, use RBF-PID algorithm to correct the master-slave error to obtain more accurate transformer calibration results.

[0017] The present application is a kind of transformer calibration master-slave clock error correction method, which has good adaptability to the transformer calibration process in substation. Simulation experiment verifies the performance index of this algorithm. In the face of the situation, this algorithm has good convergence, and the convergence speed is fast. And through the physical experiment, the feasibility and practicability of this algorithm are verified. After the correction of the master-slave clock error, it meets the requirements of 0.1 accurate level of transformer calibration. At the same time, this algorithm is well combined with embedded system, simple operation, low cost, and has good application prospect. The additional advantages of the present application will be given in the following description, which becomes more obvious through the following description. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 It is the overall framework of the transformer edge cloud collaborative calibration system.

[0019] Figure 2 It is the data transmission process of the monitoring system and the cloud server.

[0020] Figure 3 For a mutual inductor verification platform.

[0021] Figure 4 For the structure block diagram of the master-slave clock error acquisition circuit.

[0022] Figure 5 For the master-slave clock error acquisition result.

[0023] Figure 6 For the RBF-PID algorithm result block diagram.

[0024] Figure 7 For the master-slave time error before and after filtering.

[0025] Figure 8 For the master-slave clock error correction result.

[0026] Figure 9 For the mutual inductor output waveform before and after the master-slave clock error correction.

[0027] Figure 10 For the mutual inductor amplitude-frequency characteristic before and after the master-slave clock error correction. DETAILED DESCRIPTION

[0028] The application will be further described in detail below with reference to the accompanying drawings.

[0029] A mutual inductor verification master-slave clock error correction method, comprising the following steps:

[0030] Step A. Analyze the specific reasons for the deviation of the master-slave clock in the mutual inductor verification system, and complete the mathematical modeling of the master-slave clock error according to the error reason.

[0031] Step A1: Analyze the specific process of mutual inductor verification, and determine that the error causes are concentrated in two aspects: theoretical error and transmission error;

[0032] The structure of the mutual inductor edge cloud collaborative verification system built is as shown in Figure 1 Each monitoring system establishes communication with the cloud server through WiFi, transmits mutual inductor acquisition data to the cloud using TCP protocol, and then exports it to the edge computing side for subsequent operation. Selecting a standard mutual inductor can verify the remaining mutual inductors to be verified.

[0033] In this process, although each monitoring system has introduced a GPS module for synchronization time service, there is still a clock error. The master-slave clock error between the standard mutual inductor and the mutual inductor to be verified will affect the accuracy of mutual inductor verification.

[0034] Analyzing the above process, the causes of master-slave clock error are mainly concentrated in two aspects, one is the error of GPS crystal oscillator itself, that is, the theoretical error, and the other is the link delay error in the process of collecting data of the mutual inductor and transmitting to the cloud through TCP protocol, that is, the transmission error. In the following, the specific process of mutual inductor verification will be analyzed from the two aspects, and the master-slave clock error model will be established.

[0035] Step A2: analyze the temperature drift phenomenon of GPS crystal oscillator, and establish the theoretical error model of master-slave clock of mutual inductor verification.

[0036] For the temperature drift phenomenon of GPS crystal oscillator, the following theoretical error model is established.

[0037] Define the clock of the standard mutual inductor monitoring system as the master clock, and the clock of the mutual inductor monitoring system to be verified as the slave clock. Let t be the standard time, at this time the master clock time is m(t), and m(t) = t, and the slave clock time is s(t), then the master-slave clock error is

[0038] θ(t) = s(t) - t (1)

[0039] The fundamental reason for the theoretical error of master-slave clock is clock drift, that is, there is a deviation in the rate of pulse generation of different crystal oscillators, so the frequency deviation of master-slave clock is defined as

[0040]

[0041] In the formula, τ is the time interval, and τ→0.

[0042] From formula (2), we can get

[0043]

[0044] Under stable conditions, the frequency deviation of master-slave clock can be considered as a constant, so formula (3) can be discretized. In a fixed time interval ΔT, the slave clock can be represented as

[0045]

[0046] In the formula, s[k] is the slave clock time at time k, and s[k+1] is the slave clock time at time k+1.

[0047] Therefore, the state equation of the discrete master-slave clock error is

[0048]

[0049] In the formula, ω θ [k+1] is the clock error noise at time k+1, with a mean of 0 and a variance of

[0050] The state equation of the master-slave clock frequency deviation can be expressed as

[0051]

[0052] wherein, is the master-slave clock frequency deviation noise at k+1 time, with mean value of 0 and variance of

[0053] Equations (5) and (6) together constitute the theoretical error model of the master-slave clock.

[0054] Step A3: Analyze the transmission process of the collected data in the transformer calibration process, and establish the transmission error model of the transformer calibration master-slave clock.

[0055] Each monitoring system and the cloud server perform data transmission using the TCP protocol. The TCP protocol is a connection-oriented and reliable communication protocol. The TCP transmission process needs to go through three handshakes and four handwavings. To simplify the calculation, the transmission model is simplified, as shown in Figure 2 .

[0056] In the figure, T1 is the time when the monitoring system sends a connection request message to the server, T2 is the time when the server receives the request connection message, T3 is the time when the server sends an acknowledgment request message (ACK), and T4 is the time when the monitoring system receives the ACK message. T5 is the time when the monitoring system sends an ACK message and a request to open a connection channel message to the server, T6 is the time when the server receives the request to open a connection channel message, T7 is the time when the server sends an ACK message, and T8 is the time when the monitoring system receives the ACK message.

[0057] Under stable conditions, the link error between the monitoring system and the cloud server can be regarded as a constant value, so the TCP transmission process is regarded as two symmetrical parts, which have been marked in Figure 3 . For any part, the error analysis has the following relationship

[0058]

[0059] wherein, D jf is the transmission link delay from the monitoring system to the cloud server, D fj is the transmission link delay from the cloud server to the monitoring system, and ε is the initial clock error between the monitoring system and the server.

[0060] For the problem of transformer calibration, only the clock synchronization between the standard transformer monitoring system and the calibrated transformer monitoring system needs to be considered, that is, the relative synchronization of the master and slave clocks, and all monitoring systems satisfy the relationship of formula (7) when transmitting with the cloud server, only the transmission link delay and initial clock error between different monitoring systems and the cloud server are different, so the transmission model between the master and slave clocks can be obtained as

[0061]

[0062] In the formula, Δt1 is the time difference between the master clock sending a connection request to the cloud server and the slave clock sending a connection request to the cloud server, Δt2 is the time difference between the cloud server receiving the master clock connection request and receiving the slave clock connection request, Δt3 is the time difference between the cloud server sending a confirmation request to the master clock and sending a confirmation request to the slave clock, and Δt4 is the time difference between the master clock receiving the confirmation request and the slave clock receiving the confirmation request.D sm D is the transmission link delay difference from the master clock to the cloud server and from the slave clock to the cloud server, ms D is the transmission link delay difference from the cloud server to the master clock and from the cloud server to the slave clock, and θ is the master-slave clock error.

[0063] From formula (8), we can get

[0064]

[0065] Discretizing formula (9), we get

[0066]

[0067] In actual situations, the transmission link is asymmetric, that is, ms ≠D sm , which can be regarded as an independent and identically distributed Gaussian process, and their mean is set to 0 and the variance is At the same time, the time stamp recorded during the transmission process will also have a certain degree of error, so Δt1 and Δt4 are regarded as Gaussian processes with mean 0 and variance s 2, and Δt2 and Δt3 are regarded as Gaussian processes with mean 0 and variance . Therefore, formula (10) can be further rewritten to obtain the measurement equation of the master-slave clock error

[0068]

[0069] In the formula, θ r [k] is the measured value of the master-slave clock error. is the total measurement error, including the uncertainty of the time stamp and the jitter caused by the asymmetric link, and its mean is 0 and the variance is where

[0070]

[0071] Similarly, the measurement equation of the master-slave clock frequency deviation is

[0072]

[0073] According to the relationship between the clock error and the clock frequency offset

[0074]

[0075] The variance of the measurement noise of the master-slave clock frequency deviation is

[0076]

[0077] Equations (11) and (13) together constitute the transmission error model of the master-slave clock.

[0078] Step B. Build a mutual inductor verification platform to obtain the initial error of the master-slave clock.

[0079] The mutual inductor verification platform is built as shown in Figure 3 .

[0080] The circuit structure for obtaining the master-slave clock error is shown in Figure 4 .

[0081] As can be seen from the figure, the master-slave clock will capture its own GPS second pulse signal, and both sides will communicate through the serial port. The slave clock flips the pin after capturing the GPS signal to notify the master clock, and the master clock starts timing. When the master clock captures its own GPS signal, it ends timing, thereby obtaining the initial error of the master-slave clock. Finally, it is printed to the computer through the serial port.

[0082] The experimental results are shown in Figure 5 . The initial deviation of the master-slave clock is about 820ns.

[0083] Step C. Correct the master-slave clock error using Kalman filtering and RBF-PID.

[0084] Step C1: Use Kalman filtering to filter out the process noise in the error model and obtain effective waveform information.

[0085] Substitute equations (5) and (6) as the state transition equation of the system, and equations (11) and (13) as the measurement equation of the system into the state space model of Kalman filtering

[0086]

[0087] Perform Kalman filtering operation, and the prediction equation is ​

[0088]

[0089] wherein, is the prior estimate value at the current time, is the optimal estimate value at the last time, P t - is the covariance matrix corresponding to the prior estimate value at the current time.

[0090] The update part equation is

[0091]

[0092] wherein, K t is the Kalman gain coefficient at the current time, P t is the covariance matrix corresponding to the optimal estimate value at the current time.

[0093] The whole Kalman filtering process is as follows: firstly, according to the prediction part equation, the estimate value at the current time is calculated from the optimal estimate value of the state at the last time, which is called the prior estimate, and the covariance matrix corresponding to the prior estimate is calculated. Then the update part equation is brought back to obtain the latest Kalman gain coefficient, and the observation value at the current time and the prior estimate value are weighted and compared according to the difference of the Kalman gain coefficient to obtain the optimal estimate at the current time. Thus, the optimal estimate of data acquisition is obtained through repeated cycles, and the filtering effect is achieved.

[0094] Step C2: Based on step C1, the RBF-PID algorithm is used to correct the master-slave error to obtain a more accurate transformer calibration result.

[0095] After Kalman filtering, the master-slave clock still has error, and the RBF-PID algorithm is used for error correction. The structure block diagram of the RBF-PID algorithm is as Figure 6 shown.

[0096] In the figure, r(k) is the expected output error of the master-slave clock, which is generally set to 0. The controller output u(k) and the actual output error y out (k) of the master-slave clock together constitute the input of the RBF neural network. After training and learning of the neural network, the required Jacobian information is outputted, which is used for parameter setting of the PID. The specific algorithm principle is as follows.

[0097] In order to avoid the cumulative error generated by the model, the incremental PID is adopted, and the controller output expression is

[0098]

[0099] The activation function adopted by the RBF neural network is

[0100]

[0101] In the formula, c ij Let b be the center vector of the j-th node in the network. j The node width of the Gaussian function, where x is the input vector of the network.

[0102] The output of the RBF neural network at time k is

[0103]

[0104] In the formula, w i This refers to network weights.

[0105] The performance metric of an RBF neural network is defined as follows:

[0106] J = 0.5(y out (k)-y m (k)) 2 (twenty two)

[0107] Gradient descent is used to adjust the network weights, node centers, and Gaussian function node widths. During parameter adjustment, if the learning factor is too large, it may cause instability in the network weight changes. Therefore, a momentum factor needs to be added during the correction process to improve stability. The specific iterative process is as follows:

[0108]

[0109] In the formula, η is the learning factor and α is the momentum factor.

[0110] The Jacbian information of a system is defined as the sensitivity of the controlled object's output to the control input.

[0111]

[0112] Based on the Jacobian information identification described above, the PID parameters are tuned. The output error of the system at time k is defined as...

[0113] e(k)=r(k)-y out (k) (25)

[0114] The tuning index of RBF-PID is set to

[0115] E(k) = 0.5e(k) 2 (26)

[0116] make

[0117]

[0118] Using the gradient descent method to tune PID parameters, we have

[0119]

[0120] Based on the above algorithm principle, the initial values ​​of each PID parameter and the learning factor and momentum factor of the neural network are selected. Through iterative learning, Jacobian information is obtained and the PID parameters are adjusted to finally achieve the required tuning index.

[0121] To enable those skilled in the art to better understand the present invention and its advantages over the prior art, the applicant provides further explanation in conjunction with specific embodiments.

[0122] Based on the current transformer clock error model and error correction algorithm constructed above, and the initial master-slave clock error obtained through experiments, a simulation experiment was conducted. The initial master-slave clock error was set to 820 ns, and the master-slave clock frequency deviation was 10 ns. -6 s, the measurement noise variance is 10 -14 First, Kalman filtering is applied to filter the master-slave clock error, and the result is as follows. Figure 7 As shown.

[0123] As shown in the figure, the Kalman filter algorithm effectively filters out measurement noise while preserving the valid information of the original waveform. Multiple experiments were conducted with different values ​​for the Gaussian noise variance, and the Kalman filter still demonstrated good filtering performance.

[0124] The master-slave clock error of the filtered current transformer is corrected, and the result is as follows: Figure 8 As shown.

[0125] As shown in the figure, the master-slave clock error of the current transformer is significantly reduced after algorithm correction, and the correction accuracy of the RBF-PID algorithm is better than that of the ordinary PID algorithm. The master-slave clock error corrected by the RBF-PID algorithm is approximately 20ns, which translates to a waveform phase error of approximately 1.3″, meeting the requirement of 0.1 accuracy class for current transformer calibration. Furthermore, this algorithm converges in approximately 0.4s, which is relatively fast and meets the requirements for real-time error correction of GPS second pulses in master-slave current transformer monitoring systems in practical engineering applications.

[0126] The reliability of this algorithm was further verified by setting different initial errors of master and slave clocks, different master and slave clock frequency deviations, and noise variances. The results are shown in Table 1.

[0127] Table 1 Results of RBF-PID correction of master-slave clock error under different conditions

[0128]

[0129] From Table 1, it can be seen that the initial error of the master-slave clock has little effect on the algorithm accuracy. The frequency deviation of the master-slave clock has a greater effect on the algorithm accuracy. The greater the frequency deviation of the master-slave clock, the worse the error correction effect, but the error after correction still meets the requirements of the 0.1 accuracy level of the transformer calibration. The size of the measurement noise has a certain degree of influence on the error correction, but the correction result still meets the needs of the engineering practice.

[0130] According to the previously obtained initial error 820ns of the master-slave clock, the algorithm described above is applied to the transformer calibration system to obtain more accurate transformer data acquisition results, as shown in Figure 9 .

[0131] The corrected error of the master-slave clock of the transformer is about 6.974ns, and the phase difference between the standard transformer and the calibrated transformer waveform is about 0.45", which meets the requirements of the 0.1 accuracy level of the transformer calibration.

[0132] The Fourier calculation is performed on the collected transformer data to obtain the amplitude information of the current signal, and the result is shown in Figure 10 .

[0133] As can be seen from the figure, the amplitude error of the calibrated transformer after the master-slave clock error correction decreases from 0.00418% to 0.00358%, and the amplitude accuracy increases by about 15%.

[0134] In the face of different initial deviations of the master-slave clock, different frequency deviations of the master-slave clock, and different measurement noises, this algorithm has good convergence, and the corrected clock error is less than 200ns, and the algorithm convergence time is about 0.4s, which meets the real-time correction requirements of the GPS second pulse. At the same time, this algorithm is feasible and practical. After the master-slave clock error correction, the phase error of the transformer calibration is about 0.45", and the amplitude error is about 0.00358%, which meets the requirements of the 0.1 accuracy level of the transformer calibration. In addition, this algorithm is well combined with the embedded system, and has the advantages of simple operation, low cost, and good application prospect.

Claims

1. A method for correcting master-slave clock error in a transformer calibration, characterized in that, The method comprises the following steps: Step A. Analyze the specific reasons for the generation of the master-slave clock deviation in the mutual inductor verification system, and complete the mathematical modeling of the master-slave clock error according to the error reasons; Step B. Obtain the initial deviation between the mutual inductor master-slave clocks; Step C. Correct the master-slave clock error by using the Kalman filtering and RBF-PID methods, and obtain more accurate mutual inductor verification results.

2. According to the method of claim 1, for step A, comprising: Step A1: Analyzing the specific process of mutual inductor verification, it is determined that the error causes are concentrated in two aspects: theoretical error and transmission error; Step A2: Analyze the temperature drift phenomenon of the GPS crystal oscillator, and establish a theoretical error model of the mutual inductor verification master-slave clock; Step A3: Analyze the transmission process of the collected data in the mutual inductor verification process, and establish a transmission error model of the mutual inductor verification master-slave clock.

3. The method of claim 2, wherein, The step A1 comprises: The mutual inductor edge cloud collaborative verification system is built. Each monitoring system communicates with the cloud server through WiFi, transmits the mutual inductor collected data to the cloud side by using the TCP protocol, and then exports it to the edge computing side for subsequent operation. The selected standard mutual inductor can be used to verify the remaining mutual inductors to be verified. In this process, the reasons for the generation of the master-slave clock error are mainly concentrated in two aspects, one is the error of the GPS crystal oscillator itself, that is, the theoretical error, and the other is the link delay error generated in the process of transmitting the mutual inductor collected data to the cloud by the TCP protocol, that is, the transmission error. In the following, from these two aspects, the specific process of mutual inductor verification will be analyzed, and the master-slave clock error model will be established.

4. The method of claim 2, wherein, The step A2 comprises: Define the clock of the standard mutual inductor monitoring system as the master clock, and the clock of the mutual inductor monitoring system to be verified as the slave clock. Let t be the standard time, the master clock time at this time is m(t), and m(t)=t, the slave clock time is s(t), then the master-slave clock error is θ(t)=s(t)-t (1) The fundamental reason for the generation of the theoretical error of the master-slave clock is clock drift, that is, there is a deviation in the pulse generation rate of different crystal oscillators, so the master-slave clock frequency deviation is defined as In the formula, τ is the time interval, and τ→0. From formula (2), we can get Under stable conditions, the master-slave clock frequency deviation can be considered as a constant, so formula (3) can be discretized. In a fixed time interval ΔT, the slave clock can be represented as In the formula, s[k] is the slave clock time at time k, and s[k+1] is the slave clock time at time k+1. Therefore, the state equation of the discrete master-slave clock error is In the formula, ω θ [k+1] is the clock error noise at time k+1, with mean value 0 and variance And the state equation of the master-slave clock frequency deviation can be represented as In the formula, is the master-slave clock frequency deviation noise at time k+1, with mean value 0 and variance Formula (5) and formula (6) together constitute the theoretical error model of the master-slave clock.

5. The method of claim 2, wherein, The step A3 comprises: For the mutual inductor verification problem, only the synchronization of the standard mutual inductor monitoring system and the mutual inductor monitoring system to be verified, that is, the relative synchronization of the master-slave clock, needs to be considered, and all monitoring systems satisfy the relationship of formula (7) when transmitting to the cloud server, only the transmission link delay and the initial clock error between different monitoring systems and the cloud server are different, so the transmission model between the master-slave clocks can be obtained as In the formula, Δt1 is the time difference between the master clock sending a connection request to the cloud server and the slave clock sending a connection request to the cloud server, Δt2 is the time difference between the cloud server receiving the master clock connection request and receiving the slave clock connection request, Δt3 is the time difference between the cloud server sending a confirmation request to the master clock and sending a confirmation request to the slave clock, and Δt4 is the time difference between the master clock receiving the confirmation request and the slave clock receiving the confirmation request.D sm D is the transmission link delay difference between the master clock to the cloud server and the slave clock to the cloud server, ms D is the transmission link delay difference between the master clock to the cloud server and the slave clock to the cloud server, θ is the master-slave clock error. From formula (8), we can get Discretize formula (9) to get In practice, the transmission link is asymmetric, i.e. D ms ≠ D sm , which can be regarded as independent and identically distributed Gaussian processes, setting their mean value as 0 and variance as At the same time, the time stamp recorded when transmitting and receiving messages in the transmission process will also have a certain degree of error, so Δt1 and Δt4 are regarded as Gaussian processes with mean value 0 and variance , and Δt2 and Δt3 are regarded as Gaussian processes with mean value 0 and variance . Therefore, equation (10) can be further rewritten to obtain the measurement equation of master-slave clock error where θ r [k] is the measured value of the master-slave clock error. is the measurement error sum, including the uncertainty of the timestamps and the jitter due to the link asymmetry, with mean 0 and variance where Similarly, the measurement equation of the master-slave clock frequency deviation is According to the relationship between clock error and clock frequency offset Obtaining a measurement of the master-slave clock frequency offset the variance of which is Equations (11) and (13) together constitute the transmission error model of the master-slave clock.

6. The method of claim 1, wherein step B comprises: Build a mutual inductor verification platform to obtain the initial error of the master-slave clock.

7. The method of claim 1, wherein step C comprises: Step C1: using Kalman filtering to filter out the process noise in the error model to obtain effective waveform information; Step C2: based on step C1, using RBF-PID algorithm to correct the master-slave error to obtain more accurate mutual inductor verification results.

8. The method of claim 7, wherein, The step C1 comprises: Substitute equations (5) and (6) as the state transition equation of the system and equations (11) and (13) as the measurement equation of the system into the state space model of Kalman filtering to filter out the measurement noise in the process.

9. The method of claim 7, wherein, The step C2 comprises: Based on step C1, using RBF-PID algorithm to correct the master-slave error to obtain more accurate mutual inductor verification results.

Images