A method and system for detecting the synchronization of digital outputs of current transformers

Abstract

This invention discloses a method and system for detecting the synchronization of digital outputs of instrument transformers, specifically relating to the field of power engineering testing. It addresses the problem of concealed synchronization deviations in the digital outputs of electronic instrument transformers in smart substations caused by phase-locked loop (PLL) drift. The method uses an independent external time base to anchor the data acquisition starting point and utilizes a unified time frame throughout the entire process. First, the synchronization deviation is continuously mapped as a phase walk trajectory. Then, long-term drift and instantaneous jitter are analyzed on a dual-scale basis using phase shift stacking and pulse scattering rate. Gaussian process regression converts these two heterogeneous features into a single drift tradeoff coefficient, which is used to dynamically revise the stability domain threshold. Upon threshold triggering, a minimum adjustment amount is immediately sent to the PLL and simultaneously verified, enabling clock adjustment and drift monitoring to be nested. The protection device continuously receives stable sampling sequences, significantly reducing the tripping misjudgment rate, extending the inspection cycle, and eliminating the need for additional hardware investment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power engineering detection, more particularly, the present application relates to a kind of mutual inductor digital output synchronism detection method and detection system. BACKGROUND

[0002] In modern intelligent substation, electronic mutual inductor continuously sends sampling value message by means of optical fiber, and protection device determines fault type and action sequence according to the message. The internal mutual inductor maintains time base by phase-locked loop, and encapsulates and outputs cache data into frames in a set interval. Field inspection usually adopts closed-loop calibration device, and the output of mutual inductor and reference channel are connected to the same second pulse. Unified time scale can align two data instantaneously, and the calibration result only shows average delay, so the slight swing in message interval is completely hidden.

[0003] The mutual inductor clock runs independently for a long time, the phase-locked loop continuously accumulates slight drift due to temperature drift and aging, and the sampling cache queue also injects random delay in the message assembly stage. After the superposition of the two factors, the actual sending interval of the message appears irregular dispersion; the closed-loop calibration automatically eliminates such dispersion due to sharing the second pulse, and the detection process always shows ideal synchronization. After running for a period of time, the interval jitter continuously amplifies, the direction-finding threshold of the protection device is quietly pushed away from the safety zone, and finally mis-trip is induced, and the inspection record lacks early warning. The present application aims to directly reveal the real jitter characteristics of the digital output of the mutual inductor through asynchronous capture and statistical analysis, and evaluate the synchronization reliability from the source. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the present application provides a kind of mutual inductor digital output synchronism detection method and detection system, to anchor data acquisition starting point with independent external time base, utilize unified time frame throughout the whole process, first continuously map the synchronization deviation into phase walk trajectory, then analyze long-term drift and instantaneous jitter in double scale by using the amount of phase shift superposition and pulse dispersion rate, convert the two types of heterogeneous characteristics into a single drift trade-off coefficient by Gaussian process regression, and dynamically revise the threshold of stable domain; the minimum adjustment amount is immediately issued to the phase-locked loop and verified synchronously after the threshold is triggered, so that the clock adjustment action and drift monitoring are nested with each other, to solve the problems proposed in the above background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0006] A kind of mutual inductor digital output synchronism detection method, comprising the steps of:

[0007] S1: synchronously intercepting mutual inductor digital message and reference pulse under the constraint of independent external time base, and recording to unified time frame;

[0008] S2: obtaining synchronization deviation sequence by using unified time frame;

[0009] S3: mapping the synchronization deviation sequence into a phase walk trajectory, constructing a phase shift stack by sliding window cumulative offset, combining the Mandelbrot segmentation on message interval and calculating the pulse dispersion rate from the dispersion, and generating the drift weighting coefficient by Gaussian process regression;

[0010] S4: injecting the drift weighting coefficient into the synchronization stability domain to form a deviation weight vector, and dynamically adjusting the threshold value according to the deviation weight vector;

[0011] S5: when the deviation weight vector reaches the threshold value, triggering the phase-locked loop adjustment instruction, and returning to step S1 to complete the collection verification and threshold iteration.

[0012] In a preferred embodiment, step S1 includes deploying an independent external time base as a collection constraint source, accessing the collection device between the output port of the transformer and the reference pulse source, and generating a trigger signal to start the transformer digital message sequence and the reference pulse sequence interception operation. At the beginning of each cycle, the current frame and timestamp are buffered with the current pulse and timestamp to ensure that the interception window covers the complete message interval period. The sequence is recorded in a unified time frame, which is a two-dimensional structure with the independent external time base period as the unit. The first dimension stores the transformer digital message sequence frame content and timestamp pair, and the second dimension stores the reference pulse sequence pulse content and timestamp pair. The two sequences are merged by timestamp alignment.

[0013] In a preferred embodiment, step S2 includes extracting the transformer timestamp sequence and the reference timestamp sequence from the unified time frame, calculating the difference between the transformer timestamp sequence elements and the reference timestamp sequence elements to form a synchronization deviation sequence, initializing a distribution container to accumulate the count of the synchronization deviation sequence, updating the frequency of each deviation interval in the distribution container when a new set of unified time frame data is received, and retaining the synchronization deviation sequence elements within the recent fixed window using a rolling update mechanism. The statistical distribution is defined by a deviation value to frequency mapping function.

[0014] In a preferred embodiment, step S3 includes mapping the phase walk trajectory from the synchronization deviation sequence, converting the synchronization deviation sequence elements to phase values using the statistical distribution weight, forming a phase walk trajectory sequence by accumulating the deviation modulo 2π, applying a sliding window to the phase walk trajectory to calculate the absolute difference value accumulation of adjacent elements to construct a phase shift stack sequence, calculating the adjacent difference value from the transformer timestamp sequence to form a message interval sequence, performing Mandelbrot segmentation on the message interval sequence to divide self-similar sub-sections, calculating the Hurst index of each sub-section and taking the geometric mean as the pulse dispersion rate.

[0015] In a preferred embodiment, the phase shift stack sequence and the pulse dispersion rate are normalized and input into a Gaussian process regression model to generate a drift weighting coefficient by fitting a radial basis function kernel.

[0016] In a preferred embodiment, step S4 comprises defining the synchronization stability domain as a vector space corresponding to the number of statistical distribution intervals of the synchronization deviation sequence, and applying the drift weighting coefficient as an exponential to the initial value of each dimension to form a deviation weight vector.

[0017] In a preferred embodiment, the Tsallis entropy form is applied to the deviation weight vector to calculate the normalized element preset non-additive parameter power sum transformation as an adjustment threshold.

[0018] In a preferred embodiment, step S5 comprises comparing the norm of the deviation weight vector with the dynamically adjusted threshold in real time, determining the touch by Holder norm quantization of the amplitude of the deviation weight vector, generating the minimum adjustment amount based on the maximum element of the deviation weight vector as the phase-locked loop adjustment instruction to the internal phase-locked loop of the transformer, and executing step S1 to obtain a new unified time frame, applying steps S2 to S4 to verify the adjustment effect on the new synchronization deviation sequence, and guiding the iterative dynamic adjustment threshold according to the sign function of the difference between the average value of the new synchronization deviation sequence and the average value of the old synchronization deviation sequence.

[0019] A transformer digital output synchronization detection system comprises a data interception module, a deviation maintenance module, a drift fusion module, a threshold dynamic module and an adjustment cycle module.

[0020] The data interception module synchronously intercepts the transformer digital message and the reference pulse under the constraint of an independent external time base, and records them to a unified time frame.

[0021] The deviation maintenance module obtains a synchronization deviation sequence from the unified time frame.

[0022] The drift fusion module maps the synchronization deviation sequence to a phase walk trajectory, and constructs a phase shift superposition amount by sliding window accumulation offset, combines the pulse dispersion rate obtained by performing Mandel segmentation and calculating the dispersion degree on the message interval, and generates a drift weighting coefficient by Gaussian process regression.

[0023] The threshold dynamic module injects the drift weighting coefficient into the synchronization stability domain to form a deviation weight vector, and dynamically adjusts the threshold value accordingly.

[0024] The adjustment cycle module is used to trigger the phase-locked loop adjustment instruction when the deviation weight vector touches the threshold value, and returns to the data interception module to complete the acquisition verification and threshold iteration.

[0025] The technical effects and advantages of the transformer digital output synchronization detection method and detection system of the present application are as follows:

[0026] The application independently anchors data collection starting point of external time base, uses uniform time frame throughout the whole process, firstly continuously maps synchronization deviation into phase walk trajectory, then extracts phase shift superposition and pulse dispersion rate representing long-term drift and instantaneous jitter from double scales respectively, then fuses the two into single drift trade-off coefficient through Gaussian process regression method, and adjusts stable domain threshold based on the coefficient, once the threshold is activated, sends minimum range step instruction to phase-locked loop at the same time and completes verification work, then the clock adjustment behavior and drift detection behavior form closed loop cycle. Finally, through the whole chain mode of collection, decomposition, prediction and intervention, parallel clock adjustment behavior mechanism is constructed, the effect that digital output of the transformer can keep intra-frame compactness in low load or high load scene is obtained, and the effect that hidden erosion caused by external temperature drift and network congestion can be realized in real time, so that the protection device can continuously receive stable sampling sequence, thereby reducing the effect of misjudgment of tripping, extending the patrol cycle, and without additional hardware investment. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 FIG. 1 is a flowchart of a transformer digital output synchronization detection method according to the present application;

[0028] Figure 2 FIG. 2 is a structural diagram of a transformer digital output synchronization detection system according to the present application. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0030] Embodiment 1: Figure 1 A transformer digital output synchronization detection method according to the present application is given, which comprises:

[0031] S1: Synchronously intercept transformer digital message and reference pulse under independent external time base constraint, and record to uniform time frame.

[0032] S2: Obtain synchronization deviation sequence by using uniform time frame.

[0033] S3: Map the synchronization deviation sequence into phase walk trajectory, and construct phase shift superposition through sliding window cumulative offset, combine with mandelbrot segmentation performed on message interval and calculate dispersion rate to obtain pulse dispersion rate, and generate drift trade-off coefficient through Gaussian process regression.

[0034] S4: inject the drift weighting coefficient into the synchronous stability domain to form a deviation weight vector, according to which the threshold is dynamically adjusted.

[0035] S5: when the deviation weight vector touches the threshold, a phase-locked loop adjustment instruction is triggered, and step S1 is returned to complete the collection verification and threshold iteration.

[0036] The electronic transformer continuously sends sampling value messages to the protection device of the modern intelligent substation, and through this information, the protection device can derive the fault type and action sequence. Due to the temperature drift and aging problems of the phase-locked loop inside the transformer, the interval of the sampling value message sent by the transformer will show random dispersion. In such a case, the traditional closed-loop verification method cannot truly verify the synchronization of the transformer from the source.

[0037] In order to solve the problems of hidden drift and sudden random delay of the digital output of the transformer during long-term use, the method of separating independent external time base to lock the collection starting point is used to replace the rejection effect generated by traditional automatic verification, and the timing relationship between the digital message sequence and the standard pulse sequence in the actual measurement period is directly captured.

[0038] Step S1 is divided into the following sub-steps:

[0039] S1.1: deploy an independent external time base as a collection constraint source, which is derived from a high-precision atomic clock or a satellite time module. The frequency stability of the independent external time base is better than that between the output port of the transformer and the reference pulse source in the transformer. The independent external time base generates a trigger signal, which in turn starts the interception operation of the transformer digital message sequence and the reference pulse sequence. Among them, the transformer digital message sequence refers to the sampling value frame continuously output in the optical fiber link of the electronic transformer; the reference pulse sequence refers to the standard second pulse or sub-second level reference signal. This method uses an external reference that is more accurate than external references to replace the internal clock of the transformer, eliminating the problem of cumulative error in the collection process caused by the internal clock, ensuring that the collection process has an absolute collection time reference, and in the subsequent collection process, it will not be affected by the change of the internal clock of the transformer, thereby ensuring that the constituent units of the time sequence are all under the unified time frame.

[0040] S1.2: Based on the trigger signal of the independent external time base, the synchronization of the transformer digital message sequence and the reference pulse sequence is completed. The acquisition device buffers the current frame of the transformer digital message sequence, the timestamp, and the current pulse of the reference pulse sequence at the starting position of each cycle of the independent external time base, respectively, to ensure that the width of the interception window can cover at least one complete message interval cycle; and the timestamp used is the absolute timing scale of the independent external time base. The trigger signal of the independent external time base is used to capture the original time sequence of the two sequences at the same starting time, to find the real jitter of the transformer digital message sequence, to eliminate the deviation caused by automatic alignment, and to ensure that after the result after downsampling is recorded in the same time frame, each sequence obtained is an unbiased original data set.

[0041] S1.3: The intercepted transformer digital message sequence and the reference pulse sequence are recorded in the same time frame, which is a two-dimensional array composed of the independent external time base cycle as one dimension, the frame content of the transformer digital message in each cycle and the corresponding timestamp as another dimension, and the pulse content and timestamp of the reference pulse sequence as another dimension; the transformer digital message sequence elements and the reference pulse sequence elements occupying the same position in the unified time frame obtained after timestamp alignment share the same independent external time base marker; the above recording process places the two different sequences in the same time sequence coordinate system through merging, which facilitates subsequent direct calculation of the synchronization deviation sequence based on the time sequence, and avoids the existence of faults in analysis due to different data islands. It is a data container containing all information, which can be used to extract the synchronization deviation sequence in the subsequent step S2 to complete the stage judgment of the transformer digital output synchronization source packet in the intelligent substation.

[0042] The sub-steps described in this paper complete the time frame unification, the original time sequence of the digital message sequence and the reference pulse sequence is filled in a one-to-one corresponding unified sequence, and the foundation data of the unmasked synchronization deviation is laid for the subsequent steps. Compared with the implicit elimination of the interval jitter between the messages in the traditional closed-loop verification, it shows the real synchronization of the transformer digital output in the intelligent substation.

[0043] Step S2 uses the unified time frame to calculate the synchronization deviation sequence and maintain the statistical distribution to quantify the time sequence difference of the message and the reference pulse, and prepare data for subsequent drift analysis. The reason is that the unified time frame has obtained the original time sequence, so the most real jitter can be obtained directly based on the unified deviation sequence. Real-time maintenance of the statistical distribution is to monitor the change of the deviation at all times, which avoids the neglect of dynamic changes caused by the use of static average value. For substation inspection, it can discover possible synchronization faults in advance.

[0044] According to the unified time frame established in the S1 step, the time sequence mapping relationship between the transformer digital message sequence number and the reference pulse sequence number is recorded, on the basis of which the synchronization deviation sequence is extracted to represent the subtle swing of the message interval, and a statistical distribution method is used for dynamic maintenance to obtain the variation range of the deviation value, and is applied to the quantitative evaluation of the reliability of the digital output of the transformer in the smart substation.

[0045] The step S2 is divided into the following sub-steps:

[0046] S2.1: For the transformer digital message sequence and the reference pulse sequence time sequence data that have been fused in the unified time frame, the correspondence relationship of the respective time sequence data caused by separate processing can be avoided. The time stamp of the transformer digital message sequence and the time stamp of the reference pulse sequence are extracted from the unified time frame. The time stamp of each frame content of the transformer digital message sequence is obtained by traversing the time stamp of each frame content of the transformer digital message sequence in the first dimension of the unified time frame, and the time stamp of each pulse content of the reference pulse sequence is obtained by traversing the time stamp of each pulse content of the reference pulse sequence in the second dimension of the unified time frame. The extraction ensures that the transformer time stamp sequence and the reference time stamp sequence correspond to the position information of the same time point on the unified time frame in a one-to-one correspondence, so as to preserve the original time sequence relationship, accurately reflect the time offset of the transformer digital message sequence relative to the reference pulse sequence, and facilitate the subsequent deviation calculation operation.

[0047] S2.2: The transformer time stamp sequence and the reference time stamp sequence are separated from the unified time frame to form a one-to-one mapping, so that the difference between the time stamp sequences is quantitatively calculated without considering the influence of the average delay, and the synchronization deviation sequence is calculated according to the transformer time stamp sequence and the reference time stamp sequence. In the unified time frame, the difference value between the elements in the transformer time stamp sequence and the corresponding elements in the reference time stamp sequence is used as the element value of the synchronization deviation sequence, and each element in the synchronization deviation sequence represents the time sequence offset of the transformer digital message sequence at the time point relative to the reference pulse sequence. The specific calculation steps of the synchronization deviation sequence are as follows: the first item of the transformer time stamp sequence is subtracted from the first item of the reference time stamp sequence to obtain the first difference value of the synchronization deviation sequence, the second item of the transformer time stamp sequence is subtracted from the second item of the reference time stamp sequence to obtain the second difference value of the synchronization deviation sequence, and so on until all the difference values at the time points are calculated. The direct difference value is used for instantaneous offset quantization, which avoids the influence of cumulative error and can grasp the irregular discrete condition of the message interval, directly reflecting the characteristics of the jitter.

[0048] S2.3: The synchronization deviation sequence has been generated, and the continuous record after timing offset is beneficial to dynamic tracking rather than static observation, and adapts to the needs of long-time drift accumulation. The synchronization deviation sequence distribution is maintained in real time. An initial distribution container, such as an empirical distribution function, is accumulated for the entire synchronization deviation sequence. After obtaining a new set of unified time frame data, the frequency corresponding to each deviation interval in the distribution container is updated. Only the synchronization deviation sequence elements within a certain fixed window are maintained during the maintenance process, which can directly reflect the current jitter situation. The statistical distribution is defined as a function of mapping deviation values to frequencies, which facilitates the subsequent step S3 of calling the distribution to evaluate the phase walk trajectory, thereby realizing continuous quantification of deviation evolution and suppressing the adverse effects of implicit losses on the protection device.

[0049] In step S2, the synchronization deviation sequence and the statistical distribution are periodically obtained according to the unified time window and are maintained in real time to measure the dynamic situation of the timing offset of the digital message sequence of the transformer, providing basis data for subsequent drift weighting. There is no problem of missing interval jitter in the traditional way, so it can be continuously monitored in the smart substation, realizing continuous monitoring of synchronization reliability.

[0050] According to the synchronization deviation sequence and the statistical distribution of the deviation sequence obtained in step S2, step S3 maps the synchronization deviation sequence to the phase walk trajectory and establishes the phase shift stacking amount. At the same time, the message interval sequence is processed to obtain the pulse dispersion rate, which is then fused into a drift weighting coefficient using Gaussian process regression to analyze the long-term drift and instantaneous jitter. The reason for this scheme is that the quantized synchronization deviation sequence can be mapped to the phase trajectory to obtain accumulated offset, the Mandelbrot segmentation can reflect the fractal properties of the message interval sequence, and the Gaussian process regression can fuse heterogeneous features into a single coefficient, so it can be applied to the multi-scale feature quantization of the drift evaluation of the transformer digital output in the smart substation.

[0051] Step S3 is divided into the following sub-steps:

[0052] S3.1: The synchronization deviation sequence represents a continuous time offset record, and a linear model cannot accurately represent the long-period drift cycle. The synchronization deviation sequence is mapped to a phase walk trajectory, the timestamp sequence of some selected mutual inductors in the synchronization deviation sequence is selected, the statistical distribution deviation value of each sequence is taken as the weight, the deviation value of each sequence is assigned to the weight at the same time, the deviation value is converted into phase value in sequence, and the phase value at each time point of the phase walk trajectory is obtained by taking modulus 2π after accumulating the deviation value. The specific method is: from the first deviation value of the synchronization deviation sequence, the accumulated value after the value is accumulated is taken modulo 2π to obtain the first phase value on the phase walk trajectory, and then the next deviation value is added and taken modulo 2π to obtain the second phase value. The values are the same, and after all the values in the synchronization deviation sequence are accumulated once and then taken modulo 2π, the corresponding phase value is obtained to form a complete phase walk trajectory sequence. By accumulating and taking modulus, the deviation is mapped to the periodic phase space, thereby avoiding the overflow of linear phase module accumulation, and the long-term drift cycle characteristics of the deviation can be more accurately reflected. The implicit cumulative effect of the mutual inductor digital output is more clearly described.

[0053] S3.2: The phase walk trajectory is a periodic sequence, which is convenient for analyzing and stacking the offset with a local window without considering the problem that it is canceled due to transient influence under global average statistics. The cumulative offset in each sliding window is calculated to obtain the phase shift stacking amount: the absolute value difference between the front and rear two adjacent values in the sliding window is taken to obtain the phase shift stacking amount sequence. Taking the sliding window width of ten phase values as an example, the absolute difference between the first and tenth phase values of the phase walk trajectory is taken as the first phase shift stacking amount value. Then the sliding window is shifted by one position (from the second phase value to the eleventh phase value), and the cumulative absolute difference between the phase values contained in the window is calculated to obtain the next phase shift stacking amount value. …… Such a shift cycle is performed until the cumulative absolute difference between all adjacent phase values contained in the phase walk trajectory is obtained to form a phase shift stacking amount sequence. By accumulating the absolute difference, the local offset stacking is quantified, the amplification effect of transient jitter is reflected, the local dynamics of message interval jitter are reflected, and the sensitivity to short-term fluctuations of the mutual inductor digital output is increased.

[0054] S3.3: The message interval sequence is the adjacent difference value of the transformer timestamp sequence, and the message is mixed with random delay information, which is convenient for using fractal analysis to expose irregular discrete self-similar structure instead of simple linearity measurement, using fractal dimension instead of its unit dimension, and using the Mandelbrot segmentation method to obtain the pulse dispersion rate after calculating the dispersion of the message interval. The time difference of adjacent time stamps obtained from the transformer timestamp sequence is the message interval sequence, and then the message interval sequence is iteratively divided into multiple self-similar sub-sections using the Mandelbrot segmentation method of Mandelbrot. Determine whether the sub-sections meet the boundary conditions of the Mandelbrot set (the point set boundary on the complex plane where the iteration function does not diverge), apply the boundary conditions to judge the fractal self-similarity of the sub-sections, calculate the dispersion of each sub-section according to the local Hurst index calculated for each sub-section, and take the geometric mean of the Hurst index of each sub-section to obtain the pulse dispersion rate. For the message interval sequence, it is divided into two parts in turn to test the self-similarity of the message interval sequence. When the message interval sequence does not meet the self-similarity condition, the message interval sequence is divided into four parts in turn and the sequence self-similarity is tested one by one until each self-similar sub-section meets the boundary condition of the Mandelbrot set. Then the Hurst index of each sub-section is calculated by the rescaled range analysis method. The Hurst index is estimated by the slope of the fluctuation range and the time span on the logarithmic scale, and the Hurst index of the first sub-section is multiplied by the Hurst index of the second sub-section, until the Hurst index of the last sub-section is multiplied to obtain a product value. Then take the number of sub-sections as the base to take the root of the product value to obtain the pulse dispersion rate. The random dispersion of the interval sequence is measured by the geometric mean integration of the fractal dispersion, and the quantized interval sequence can better capture the scattering characteristics of instantaneous jitter, and has higher estimation accuracy for the random delay of the transformer digital output.

[0055] S3.4: The phase shift superimposed sequence and the pulse dispersion rate have quantified the double-scale characteristics, which can more effectively perform heterogeneous data fusion under the characterization of nonlinear relationship rather than simple linear combination ignoring complex dependence. In order to obtain the drift weighting coefficient, we use Gaussian process regression to obtain the drift weighting coefficient. Specifically, the phase shift superimposed sequence and the pulse dispersion rate are used as input features, which are preprocessed and normalized, and then input into the Gaussian process regression model. The radial basis function kernel is used to fit the nonlinear mapping of the characteristics to the output, and thus the drift weighting coefficient is obtained. In the calculation process, the covariance matrix inversion and mean prediction are involved, which aims to obtain the joint drift weight under the condition of linear combination of characteristics. The specific implementation is as follows: first, subtract the minimum value in the phase shift superimposed sequence from each value and divide it by the difference between the maximum and minimum values to obtain a value normalized to the 0-1 interval. Then, the pulse dispersion rate is also normalized in the same way to obtain a value normalized to the 0-1 interval. In this way, the input feature vector is formed. Then, Gaussian process regression is used to calculate the covariance matrix between input features, where the radial basis function, i.e. Gaussian kernel, is used to measure the similarity between feature points. The inverse of the covariance matrix and the mixture of observation noise are obtained to obtain the posterior distribution, and the posterior mean is used to obtain a unique single drift weighting coefficient. Finally, the nonlinear mapping is used to fuse the features, ensuring that the drift weighting coefficient can fully grasp the joint effect, thereby better fusing the information of long-term drift and instantaneous jitter, and further enhancing the effect of predicting the overall synchronization reliability of the digital output of the mutual inductor.

[0056] In step S3, the method of mapping the synchronization deviation sequence into the phase walk trajectory and establishing the corresponding phase shift superimposed quantity, and then performing multifractal analysis based on the message interval sequence to obtain the pulse dispersion rate, and then performing Gaussian process regression fusion to obtain the drift weighting coefficient, realizes the multiscale quantitative analysis of the digital output synchronization detection of the mutual inductor, and at the same time, the periodic mapping of the phase walk trajectory is used to consider the long-term cumulative drift cycle mode of the phase-locked loop, and the masking effect of linear deviation on the implicit periodic fluctuation is offset, so that the clock offset evolution of the output message of the electronic mutual inductor in the smart substation environment can be accurately distinguished. The sliding window accumulation mode of the phase shift superimposed quantity can quantitatively express the local amplification effect of the instantaneous jitter, ensure that the detection process is most sensitive to the uncertain random delay caused by the sampling buffer queuing, and ensure the time stability index evaluation of the fault judgment sequence of the protection device. Whether the protection device can correctly and reliably act is directly related to the delay of the measured message, and the present technology reduces the drift factors to reduce the false alarm or miss alarm caused by the super-long message.

[0057] In the message interval processing, the message interval sequence is obtained by using the Mandelbrot division method (an iterative binary division based on fractal self-similarity), and the Hurst index is calculated as a dispersion index to obtain the pulse dispersion rate. The fractal structure and long-range correlation of the interval sequence can be correctly reflected. This method is better than the traditional statistical variance analysis in distinguishing the persistent trend caused by temperature drift or aging from the irregular random jitter, and can be used for irregular dispersion detection under the independent clock of the transformer, thereby suppressing the error trip caused by the implicit shift of the direction-finding threshold due to the jitter of the measurement message interval caused by this factor.

[0058] The Gaussian process regression as an integration module maps the different characteristics of the phase shift stacking amount and the pulse scattering rate to the same drift weighting coefficient, and completes the mapping calculation between the values of the above two parameters through the nonlinear kernel function mapping. The drift weighting coefficient obtained by the mapping can obtain the drift trend prediction under low samples through the uncertainty modeling under the Bayesian framework, without specific parameter distribution. This is especially beneficial to the specific optimization of the transformer digital output, realizes the real-time convergence and fusion of its long-term drift and instantaneous jitter information, and real-time correction of the domain threshold of synchronization stability, so that the closed-loop nested operation of the collection-analysis-intervention is formed, the patrol efficiency is improved, and the influence of network congestion or external disturbance of the transformer digital output is eliminated. In summary, the whole process can use two-scale feature extraction and prediction fusion to detect the synchronization characteristics of the transformer optical fiber transmission message more robustly, to a certain extent, to avoid the detection conclusion of the single frame data abnormality caused by the closed-loop test, and to improve the advantage of the continuity of the transformer protection information frame received by the protection device.

[0059] The drift weighting coefficient is injected into the synchronization stability domain, a deviation weight vector is constructed by the drift weighting coefficient, and the threshold is adjusted according to the vector to realize the nonlinear adjustment of the synchronization deviation. The reason for using this method is that the drift weighting coefficient has integrated the two-scale characteristics, and through the non-additive injection, a large-scale weight vector corresponding to any complex distribution can be generated. In addition, the dynamic threshold adjustment is also based on the correction boundary for generalized entropy calculation, so the linear method is not easy to ignore the non-Gaussian drift, and the jitter can be optimized in real time during the actual operation of the substation to ensure the accuracy of the time sequence of the protection device.

[0060] The drift weighting coefficient is obtained by the Gaussian process regression output obtained in step S3 and injected into the synchronization stability domain, and then the threshold is dynamically adjusted according to the weight vector to obtain a nonlinear adaptive intelligent substation transformer digital output synchronization intervention method.

[0061] Step S4 is divided into the following substeps:

[0062] S4.1: The drift tradeoff coefficient quantifies the heterogeneous jitter characteristics, enabling the nonlinear injection vector space to extract scale invariance, while nonlinear methods ignore complex correlations; the drift tradeoff coefficient is injected into the synchronous stable domain to construct the deviation weight vector.

[0063] The synchronization stability region is the number of dimensions of the vector space corresponding to the statistical distribution interval of a synchronization deviation sequence. The bias weight vector is obtained by using the drift tradeoff coefficient as the exponent of the initial values ​​of each dimension. Each element of the bias weight vector is the power-law drift tradeoff coefficient corresponding to the initial value of the dimension of the synchronization stability region. This process involves using the first initial value of the synchronization stability region as the base and the drift tradeoff coefficient as the exponent to perform a power operation to obtain the first element of the bias weight vector; using the second initial value of the synchronization stability region as the base and the drift tradeoff coefficient as the exponent to perform a power operation to obtain the second element of the bias weight vector; and so on, processing all initial values ​​of dimensions to obtain all bias weight vectors. Then, power-law injection is used to achieve nonlinear amplification, which better captures the scale invariance of the drift and thus better captures the non-Gaussian distribution characteristics of the digital output of the transformer, obtaining a more accurate weight representation of the synchronization deviation. For example, the calculation formula can be: [Formula omitted for brevity]. ,in The first element of the bias weight vector represents the first element. One element, The first stable region represents the synchronous stable region. Initial values ​​for each dimension (preset as interval normalized values) This represents the drift tradeoff coefficient.

[0064] S4.2: The deviation weight vector has formed a nonlinear structure and is suitable for generalized entropy transformation to measure dispersion, in order to adapt to the distribution emphasis at the tail, while non-standard entropy cannot take into account non-additive effects; dynamic threshold adjustment is performed based on the deviation weight vector. The specific approach is as follows: The deviation weight vector is processed using the Tsallis entropy form, i.e., the generalized entropy form, to describe the partial information uncertainty of the non-additive system. The transformation after raising the normalized vector elements to the power of the preset non-additive parameter is used as the new threshold. The actual operation steps are as follows: First, the sum of all elements of the deviation weight vector is calculated as the denominator for normalization. Then, each element value in the deviation weight vector is divided by its sum to obtain its normalized value, and the power of each element is raised to the power of the preset non-additive parameter. Finally, all powers are summed to obtain an intermediate sum. This intermediate sum is then subtracted from one, and the result is divided by one and the preset non-additive parameter is subtracted to obtain the new threshold. Simultaneously, the generalized entropy transformation quantifies the non-additive dispersion to better handle the dynamic boundary changes caused by complex drifts, improving the sensitivity to the jitter changes in the digital output of the transformer and enhancing the robustness of the intervention mechanism. For example, the Tsallis entropy expression can be used to transform each element of the deviation weight vector, and the power of each normalized component can be raised to the power of the sum to obtain the new threshold. For example, the calculation formula could be: applying the Tsallis entropy form to the bias weight vector to calculate the normalized vector elements. The result of summing the powers of a factor and then transforming it serves as the new threshold; the calculation formula is as follows: ,in This indicates the adjusted threshold. The first element of the bias weight vector represents the first element. One element, This indicates a preset non-additive parameter (with a value greater than 1 to emphasize the tail distribution). This represents the total number of vector dimensions. This represents the sum of all elements in the bias weight vector.

[0065] In step S4, the deviation weight vector is generated as a drift tradeoff coefficient. Based on this, a corresponding threshold is generated to nonlinearly adjust the jitter of the digital output of the current transformer. On this basis, the optimized boundary setting for triggering intervention is realized, thereby alleviating the problem that linear methods cannot be applied to smart substations.

[0066] When the deviation weight vector triggers the phase-locked loop adjustment instruction, the closed-loop nested process of collection verification and threshold iteration is returned to S1 again. The reason for choosing this scheme is that the deviation weight vector represents the current time drift state, and its value change reaching the threshold will directly issue the minimum adjustment amount to the phase-locked loop, and at the same time, new data points will be collected to update the threshold. In this way, the use of static threshold can be avoided, and the problem of not responding in time due to timeout state can be avoided. In the operation of the substation, the purpose of correcting the clock drift at any time is achieved, the inspection frequency is reduced, and the loss caused by the hidden loss is reduced.

[0067] According to the deviation weight vector formed in S4 and the dynamic adjustment threshold, it is judged whether the adjustment instruction is issued and whether it is looped back to S1 to continue collection verification and threshold iteration, so as to provide a closed-loop intervention means for the synchronization of the digital output of the mutual inductor in the intelligent substation.

[0068] Step S5 is divided into the following sub-steps:

[0069] S5.1: The deviation weight vector has a non-linear weighted item, and the Holder norm is used to highlight the extreme value deviation, so it is more sensitive to the average deviation than the Euclidean norm; and whether the deviation weight vector reaches the threshold can be tracked and monitored.

[0070] The deviation weight vector norm and the dynamic adjustment threshold are compared in real time. If the norm is greater than the threshold, it is determined to be touched. The process of calculating the norm is to take the first element of the deviation weight vector to the power of the prepared hold index, take the second element of the deviation weight vector to the power of the prepared hold index, and so on until the last element of the deviation weight vector is taken to the power of the prepared hold index. The powers are added, and the sum is taken as the prepared hold index power to obtain the deviation weight vector norm. The prepared hold index is greater than 1, so the end value is more important. The entire vector amplitude is quantified by the prepared hold norm, and the non-Euclidean distance is used to find the position of the value fluctuation reaching the early warning value, so as to accurately locate the starting position of the mutual inductor digital output jitter threshold exceeding, and ensure the accuracy of the trigger response.

[0071] S5.2: Based on the touch judgment, the minimum increment is obtained based on the maximum element, and the fine adjustment is realized instead of the mean adjustment ignoring the local peak value. At this time, the trigger level is too low, and the phase-locked loop adjustment instruction is triggered.

[0072] According to the minimum adjustment amount as an adjustment instruction into the phase-locked loop inside the transformer, and the maximum element of the deviation weight vector is used to calculate the basis for the minimum adjustment amount, that is, the calculation method is: find the maximum value in the deviation weight vector, then multiply the maximum value by a set scaling factor, the value is between (0, 1), and the scaling factor can ensure that the adjustment amplitude is as small as possible, and the final adjustment work is completed in the form of scaling with the maximum value. In order to quantify the order of magnitude of the time offset, so as to ensure that the offset of the time base jitter in the adjustment amount and the measurement signal is in the order of magnitude correspondence, so as to suppress the instantaneous amplification of the interval jitter of the digital output of the transformer, so as to make the protection device more accurate and stable in judging the time sequence.

[0073] S5.3: The adjustment instruction is issued to the substation monitoring system, and the iterative threshold is gradually modified according to the sign generated thereby to adapt to the change instead of ignoring the directionality of the deviation size, and after the adjustment instruction is completed, it returns to the S1 step to verify the newly collected data and continue to iterate the threshold.

[0074] The step S1 is circularly executed to obtain a new unified time frame, and when the deviation is reduced in the verification of the adjustment effect of the new synchronization deviation sequence by using the foregoing method, the iterative threshold is adjusted upwards, and otherwise, the iterative threshold is adjusted downwards, and the specific calculation formula is as follows: first, the average value of the new synchronization deviation sequence is calculated, the average value of the old synchronization deviation sequence is subtracted to obtain an average difference, and then the value of the sign function is obtained, the sign function: if the average difference is greater than 0, +1 is output, if the average difference is less than 0, -1 is output, and if the average difference is equal to 0, 0 is output, the sign function is multiplied by a preset iterative step length as an intermediate factor, the iterative step length is a very small positive value given in advance, and the intermediate factor value is accumulated in one to obtain the intermediate factor, finally, the current dynamic adjustment threshold is multiplied by the intermediate factor to obtain the threshold after iteration, the boundary value is adaptively iterated and optimized by using the sign for guidance, and finally, the digital output synchronization calibration of the transformer is realized, and the mis-trip is reduced.

[0075] After the foregoing operation, when the deviation weight vector trigger level reaches the threshold value, the phase-locked loop is started to adjust, and sampling verification and threshold iteration are performed, the digital output jitter of the transformer is corrected in a closed loop, the continuous synchronization in the smart substation is maintained, and the mis-trip caused by drift amplification is prevented.

[0076] Embodiment 2: Figure 2 A kind of transformer digital output synchronization detection system of the present application is given, comprising: data intercepting module, deviation maintenance module, drift fusion module, threshold dynamic module and adjustment cycle module;

[0077] The data intercepting module synchronously intercepts transformer digital message and reference pulse under independent external time base constraint, and records to unified time frame;

[0078] The deviation maintenance module obtains synchronization deviation sequence using unified time frame;

[0079] The drift fusion module maps the synchronization deviation sequence into a phase walk trajectory, and constructs a phase shift stack through a sliding window cumulative offset, combines the Mandelbrot segmentation performed on the message interval and the pulse dispersion rate calculated by the dispersion, and generates a drift weighting coefficient through Gaussian process regression;

[0080] The threshold dynamic module injects the drift weighting coefficient into the synchronization stable domain to form a deviation weight vector, and dynamically adjusts the threshold according to the deviation weight vector;

[0081] The adjustment cycle module is used to trigger the phase-locked loop adjustment instruction when the deviation weight vector reaches the threshold, and returns to the data interception module to complete the collection verification and threshold iteration.

[0082] The above formulas are all dimensionless numerical calculations, the formulas are obtained by software simulation of a large amount of collected data to obtain a formula of the nearest real situation, and the preset parameters in the formula are set by the person skilled in the art according to the actual situation.

[0083] It should be noted that the system of the present application can be deployed in the device itself to realize embedded application, or can be run on PC or other terminal with user interface, so as to meet various hardware environments and use requirements.

[0084] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present application. Therefore, the above figures and descriptions are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

[0085] It should be noted that in this paper, if there are relationship terms such as first and second, they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the term "includes" "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitation, the element defined by the sentence "includes a" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0086] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A mutual inductor digital output synchronism detection method, characterized in that, The method comprises the steps of: S1: synchronously intercepting the digital message of the mutual inductor and the reference pulse under the constraint of an independent external time base, and recording to a unified time frame; S2: obtaining a synchronization deviation sequence using the unified time frame; S3: mapping the synchronization deviation sequence into a phase walk trajectory, constructing a phase shift stack through sliding window cumulative offset, combining the pulse dispersion rate obtained by performing Mandelbrot segmentation on the message interval and calculating the dispersion, and generating a drift weighting coefficient through Gaussian process regression; S4: injecting the drift weighting coefficient into a synchronization stability domain to form a deviation weight vector, and dynamically adjusting the threshold value according to the deviation weight vector; S5: when the deviation weight vector reaches the threshold value, triggering a phase-locked loop adjustment instruction, and returning to step S1 to complete collection verification and threshold iteration. Step S1 includes deploying an independent external time base as a collection constraint source, connecting the collection device between the output port of the mutual inductor and the reference pulse source, and generating a trigger signal to start the mutual inductor digital message sequence and the reference pulse sequence interception operation. At the beginning of each cycle, the current frame and timestamp and the current pulse and timestamp are buffered to ensure that the interception window covers the complete message interval period. The sequences are recorded to a unified time frame, which is a two-dimensional structure with the independent external time base period as the unit. The first dimension stores the mutual inductor digital message sequence frame content and timestamp pair, and the second dimension stores the reference pulse sequence pulse content and timestamp pair. The two sequences are merged through timestamp alignment.

2. The mutual inductor digital output synchronization detection method according to claim 1, wherein: Step S2 includes extracting the mutual inductor timestamp sequence and the reference timestamp sequence from the unified time frame, calculating the difference between the elements of the mutual inductor timestamp sequence and the elements of the reference timestamp sequence to form a synchronization deviation sequence, initializing a distribution container to accumulate the count of the synchronization deviation sequence, updating the frequency of each deviation interval in the distribution container every time a new set of unified time frame data is received, retaining the synchronization deviation sequence elements within the recent fixed window using a rolling update mechanism, and defining the statistical distribution using a deviation value to frequency mapping function.

3. The mutual inductor digital output synchronization detection method according to claim 2, wherein: Step S3 includes projecting a phase walk trajectory from the synchronization deviation sequence, converting the synchronization deviation sequence elements into phase values using the statistical distribution weight, forming a phase walk trajectory sequence by taking the cumulative deviation modulo 2π, applying a sliding window to the phase walk trajectory to calculate the absolute difference between adjacent elements and accumulate to construct a phase shift stack sequence, performing Mandelbrot segmentation on the message interval sequence to divide self-similar sub-sections, calculating the Hurst index of each sub-section and taking the geometric mean as the pulse dispersion rate.

4. The mutual inductor digital output synchronization detection method according to claim 3, wherein: the phase shift stack sequence and the pulse dispersion rate are normalized and input into a Gaussian process regression model to generate a drift weighting coefficient through a radial basis function kernel fitting.

5. The mutual inductor digital output synchronization detection method according to claim 4, wherein: ​ Step S4 includes defining the synchronization stability domain as a vector space corresponding to the number of statistical distribution intervals of the synchronization deviation sequence, and applying the drift weighting coefficient as an exponential to the initial value of each dimension to form a deviation weight vector.

6. The mutual inductor digital output synchronization detection method according to claim 5, characterized in that: The Tsallis entropy form is applied to the deviation weight vector to calculate the sum of the preset non-additive parameter powers of the normalized elements, and the transformed result is used as the adjustment threshold.

7. The mutual inductor digital output synchronization detection method according to claim 6, characterized in that: Step S5 includes comparing the norm of the deviation weight vector with the dynamically adjusted threshold in real time, determining the amplitude of the deviation weight vector through Holder norm quantization, generating the minimum adjustment amount based on the maximum element of the deviation weight vector as a phase-locked loop adjustment instruction, and issuing the adjustment instruction to the internal phase-locked loop of the mutual inductor, and cyclically executing Step S1 to obtain a new unified time frame, applying Steps S2 to S4 to verify the adjustment effect on a new synchronization deviation sequence, and guiding the iterative dynamic adjustment threshold according to the sign function of the difference between the average value of the new synchronization deviation sequence and the average value of the old synchronization deviation sequence.

8. A mutual inductor digital output synchronism detection system for implementing the mutual inductor digital output synchronism detection method of any one of claims 1-6, characterized in that, The mutual inductor digital output synchronization detection method comprises a data interception module, a deviation maintenance module, a drift fusion module, a threshold dynamic module, and an adjustment cycle module. The data interception module synchronously intercepts the mutual inductor digital message and the reference pulse under the constraint of an independent external time base, and records them to a unified time frame. The deviation maintenance module obtains a synchronization deviation sequence from the unified time frame. The drift fusion module maps the synchronization deviation sequence to a phase walk trajectory, constructs a phase shift superposition amount through sliding window cumulative offset, combines the pulse dispersion rate obtained by performing Mandel segmentation and calculating the dispersion degree on the message interval, and generates a drift weighting coefficient through Gaussian process regression. The threshold dynamic module injects the drift weighting coefficient into the synchronization stability domain to form a deviation weight vector, and dynamically adjusts the threshold value according to the deviation weight vector. The adjustment cycle module is used to trigger a phase-locked loop adjustment instruction when the deviation weight vector touches the threshold value, and returns to the data interception module to complete the collection verification and threshold iteration.

Images