Online calibration method and system for errors of power transformer

Abstract

The invention relates to the technical field of error calibration, in particular to an on-line calibration method and system for errors of a power transformer, and the method comprises the following steps: triggering data collection through setting a voltage and current change rate threshold value, and synchronously obtaining high-frequency waveform data of a primary side and a secondary side of the transformer; and taking the trigger point as a center to intercept a waveform segment of a transient state time period. According to the method, the data acquisition process is triggered by setting the threshold condition of the voltage and current change rate, the abnormal response moment in the operation of the current transformer can be accurately locked, invalid or interference data acquisition is avoided, the data validity is improved, the multi-level wavelet packet decomposition is carried out on the transient waveform, and the accuracy of the data acquisition is improved. The energy ratio and the phase difference between the primary side signal and the secondary side signal are clearly depicted under each frequency band, and the high-resolution identification of the local response transaction is effectively realized by constructing a frequency band response difference sequence and reconstructing the frequency band response difference sequence into a dynamic error feature vector.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of error calibration, in particular to an online calibration method and system for power transformer error. BACKGROUND

[0002] With the deepening of power system reform, the separation of power plants and power grids as the focus of system reform can effectively promote the diversification of the power market. The accuracy, reliability, fairness and justice of electric energy metering devices are related to economic benefits and social benefits, and have become a common concern of power plants and power grids. The current transformer is a device used to assist in measuring current data in the power system. According to the demand, it is connected to measuring instruments, relay protection devices or automatic devices, etc. to ensure the safety of the collected data. The current transformer is an important auxiliary current measuring device in the substation, so the error of the current transformer has an important influence on the measurement of the electric energy metering device.

[0003] In the error control process of the current transformer, the prior art mainly relies on periodic manual verification or static comparison analysis, which is difficult to capture the dynamic error change of the device under complex working conditions, and at the same time, the transient disturbance and steady state process are not modeled separately, resulting in mixed processing of abnormal mutation signals and steady state responses, causing the fuzzification of error evaluation dimensions. In the frequency domain analysis, the expression ability of the cycle frequency dimension to the periodic disturbance and the non-linear characteristics of the core is ignored, so that the error identification means cannot effectively distinguish the composite distortion caused by the inherent characteristics of the device and external interference. Therefore, improvement is needed. SUMMARY

[0004] The purpose of the present application is to solve the shortcomings in the prior art and to provide an online calibration method and system for power transformer error.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme, an online calibration method for power transformer error, comprising the following steps: By setting the voltage and current change rate threshold to trigger data acquisition, the high-frequency waveform data of the primary side and the secondary side of the transformer are synchronously acquired, and the waveform fragments of the transient period are intercepted with the trigger point as the center to form a transient event waveform set. Then, continuous steady-state waveform fragments are selected from the area outside the transient event, and multi-dimensional feature source data is established. The transient waveform fragments in the multi-dimensional feature source data are called, the primary side and secondary side waveforms are multi-level wavelet packet decomposed, the energy ratio and phase difference of the primary side and secondary side coefficients under each frequency band are calculated, the frequency band response difference sequence is constructed, and the dynamic error feature vector is generated by reconstructing and arranging all frequency band frequency band response difference sequences. Call the steady-state waveform segment in the multi-dimensional feature source data, extract the energy distribution information on the frequency axis and the cyclic frequency axis, construct the cyclic spectrum energy distribution graph, separate the main cyclic frequency spectrum line energy and the non-main cyclic frequency spectrum line energy introduced by core saturation according to the cyclic spectrum energy distribution graph, and calculate the cyclic distortion index; Fuse the dynamic error feature vector and the cyclic distortion index, convert the fusion result into the compensation amount of amplitude and phase according to the preset error mapping rule, establish a comprehensive error correction matrix, generate a correction parameter according to the comprehensive error correction matrix, and write the correction parameter into the metering chip register of the electric energy meter to obtain an online calibration coefficient set.

[0006] Preferably, the multi-dimensional feature source data acquisition step is: According to the voltage change rate threshold and the current change rate threshold, compare the primary side voltage change rate, the primary side current change rate, the secondary side voltage change rate, and the secondary side current change rate with the threshold, determine the time point satisfying the threshold as the trigger point, and take the primary side high-frequency waveform segment and the secondary side high-frequency waveform segment as the center to form a transient event waveform set; According to the transient event waveform set, determine the boundary of the trigger point time range, exclude all positions in the time range, select the interval according to continuity in the remaining time range, take the primary side high-frequency waveform segment and the secondary side high-frequency waveform segment from the same position to form a steady-state waveform segment; According to the steady-state waveform segment and the transient event waveform set, align the time stamps of the primary side high-frequency waveform segment and the secondary side high-frequency waveform segment, splice the primary side high-frequency waveform segment, the secondary side high-frequency waveform segment, and the steady-state waveform segment, add trigger point identification and interval label, unify structure and field naming, and generate multi-dimensional feature source data.

[0007] Preferably, the frequency band response difference sequence acquisition step is: Call the transient waveform segment in the multi-dimensional feature source data, and sequentially traverse each level according to the multi-level frequency band index. The primary side waveform and the secondary side waveform are subjected to band-pass filtering and down-sampling at each level, the time axis of the primary side waveform and the secondary side waveform is kept aligned, and the coefficient value sequence of each frequency band is recorded to form a wavelet packet coefficient sequence; According to the wavelet packet coefficient sequence, the energy metric of the primary side wavelet packet coefficient sequence and the secondary side wavelet packet coefficient sequence is respectively counted for each frequency band, and the phase metric is extracted. The energy metric is divided by the same frequency band, the phase metric is subtracted, and the corresponding time identifier and frequency band index are added to form a frequency band response difference sequence.

[0008] Preferably, the dynamic error feature vector acquisition step is: According to the frequency band response difference sequence, rearranging all the energy ratio and phase difference values of the frequency bands in a fixed splicing order from low frequency to high frequency according to the multi-level frequency band index and keeping the consistent order of the time sequence, and filling in the placeholder of the missing frequency band position, a dynamic error feature vector is generated.

[0009] Preferably, the step of obtaining the cyclic spectrum energy distribution map is: The steady-state waveform segment in the multi-dimensional feature source data is called, the secondary side signal time sequence is extracted and the DC bias is removed, the time sequence is segmented according to the fixed length and the fixed overlap rate, and the time stamp is recorded, the cyclic autocorrelation sequence is calculated segment by segment and accumulated along the time delay axis, the complex exponential kernel integration is performed along the time delay axis to obtain the frequency axis complex value distribution and establish the joint index of the frequency axis and the cyclic frequency axis, and the cyclic spectrum density function is obtained; According to the cyclic spectrum density function, the maximum peak position of the energy distribution is searched on the cyclic frequency axis, the index corresponding to the maximum peak position is recorded as the main cyclic frequency index, and the main peak bandwidth is counted. The maximum peak position of the energy distribution is searched on the frequency axis, the maximum peak position is recorded as the main frequency index, and the half-height width of the main peak is counted. Then, according to the conditions that the cyclic frequency index is an integer multiple of the main cyclic frequency and the frequency index is located in the neighborhood range of the half-height width of the main frequency, the main cyclic frequency spectrum line energy set is marked, and the region where the cyclic frequency index is not an integer multiple of the main cyclic frequency and the frequency index is far away from the neighborhood range of the half-height width of the main frequency is selected. The upper quantile of the energy distribution in the region is counted as a noise power threshold, and the region is marked as a non-main cyclic frequency spectrum line energy set, and a cyclic spectrum energy distribution map is formed.

[0010] Preferably, the step of obtaining the cyclic distortion index is: according to the cyclic spectrum energy distribution map, the cyclic distortion index is calculated.

[0011] Preferably, the step of obtaining the comprehensive error correction matrix is: The dynamic error feature vector and the cyclic distortion index are called, the dimension range is unified, and the one-to-one correspondence is established with the time stamp and the frequency band index. The value bits are spliced in a fixed field order, and the abnormal jump is removed by threshold and the missing position is linearly interpolated to obtain an error fusion representation sequence; According to the error fusion representation sequence, the condition threshold and the conversion item of the error mapping rule are read, the mapping items are matched one by one according to the frequency band index and the distortion level, the value bits are split into amplitude channel and phase channel, the conversion and saturation clipping are performed respectively, the channel identifier and the time identifier are generated, the compensation amount of the amplitude and the phase is obtained, and the comprehensive error correction matrix is obtained.

[0012] Preferably, the step of obtaining the online calibration coefficient set is: According to the compensation amount of the amplitude and the phase in the comprehensive error correction matrix, the comprehensive error correction matrix is filled in the primary side and the secondary side and the channel sequence, and is expanded into the correction parameter according to the register address mapping table, and is written into the metering chip register item by item, and the checksum and the status bit are compared and verified, and the version number and the time stamp are recorded, so that the online calibration coefficient set is obtained.

[0013] The application also provides an online calibration system, comprising: A data acquisition module triggers data acquisition by setting a voltage and a current change rate threshold, synchronously acquires high-frequency waveform data of the primary side and the secondary side of the mutual inductor, and acquires a waveform segment of a transient period by taking the trigger point as the center, forms a transient event waveform set, and selects a continuous steady-state waveform segment from an area outside the transient event to establish multi-dimensional feature source data. A dynamic feature extraction module is configured to call the transient waveform segment in the multi-dimensional feature source data, perform multi-level wavelet packet decomposition on the primary side and the secondary side waveforms, calculate the energy ratio and the phase difference of the primary side and the secondary side coefficients in each frequency band, constitute a frequency band response difference sequence, reconstruct and arrange the frequency band response difference sequence of all frequency bands, and generate a dynamic error feature vector. A steady-state feature extraction module is configured to call the steady-state waveform segment in the multi-dimensional feature source data, extract energy distribution information on the frequency axis and the cyclic frequency axis, construct a cyclic spectrum energy distribution graph, separate the main cyclic frequency spectrum line energy and the non-main cyclic frequency spectrum line energy introduced by core saturation according to the cyclic spectrum energy distribution graph, and calculate a cyclic distortion index. An error correction module is configured to fuse the dynamic error feature vector and the cyclic distortion index, convert the fusion result into the compensation amount of the amplitude and the phase according to a preset error mapping rule, establish a comprehensive error correction matrix, generate a correction parameter according to the comprehensive error correction matrix, and write the correction parameter into the metering chip register of the electric energy meter to obtain an online calibration coefficient set.

[0014] Compared with the prior art, the application has the advantages and positive effects that: The application can accurately lock the abnormal response moment in the operation of the current transformer, avoid the collection of invalid or interference data, improve the effectiveness of the data, perform multi-level wavelet packet decomposition on the transient waveform, clearly depict the energy ratio and phase difference between the primary side and the secondary side signals in each frequency band, effectively realize high-resolution identification of local response anomalies by constructing a frequency band response difference sequence and reconstructing it into a dynamic error feature vector, and enhance the characterization accuracy of the nonlinear distortion state of the transformer by extracting energy distribution through joint analysis of the frequency axis and the cycle frequency axis in the steady-state waveform, distinguishing the main cycle frequency energy and the non-main cycle frequency energy, fusing the dynamic error feature vector and the cycle distortion index, combining the error mapping rule, converting it into a specific amplitude and phase compensation amount, and further constructing a comprehensive error correction matrix to realize fine dynamic correction of the transformer error, and through the cooperative modeling of the signal features in the transient and steady states and the extraction of nonlinear errors through the frequency and cycle frequency dual-index, the error correction chain is closed through the construction and writing of the structured compensation amount, and the specificity of error identification is enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The application is illustrated by the following steps. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical scheme and advantages of the application more clear, the application is further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application and do not limit the application.

[0017] Please refer to Figure 1 The application provides a technical scheme, an online calibration method for power transformer error, including the following steps: Data collection is triggered by setting voltage and current change rate threshold, high-frequency waveform data of the primary side and the secondary side of the transformer are synchronously acquired, the transient period waveform segment is intercepted with the trigger point as the center to form a transient event waveform set, and continuous steady-state waveform segments are selected from the area outside the transient event to establish multi-dimensional feature source data; The transient waveform segment in the multi-dimensional feature source data is called to perform multi-level wavelet packet decomposition on the primary side and the secondary side waveforms, the energy ratio and phase difference of the primary side and the secondary side coefficients in each frequency band are calculated, a frequency band response difference sequence is constructed, all frequency band response difference sequences are reconstructed and arranged to generate a dynamic error feature vector; The steady-state waveform segment in the multi-dimensional characteristic source data is called, the energy distribution information on the frequency axis and the cyclic frequency axis is extracted, a cyclic spectrum energy distribution diagram is constructed, the main cyclic frequency spectrum line energy and the non-main cyclic frequency spectrum line energy introduced by core saturation are separated according to the cyclic spectrum energy distribution diagram, and a cyclic distortion index is calculated; The dynamic error feature vector and the cyclic distortion index are fused, the fusion result is converted into the compensation amount of amplitude and phase according to a preset error mapping rule, a comprehensive error correction matrix is established, a correction parameter is generated according to the comprehensive error correction matrix, and the correction parameter is written into the metering chip register of the electric energy meter to obtain an online calibration coefficient set.

[0018] The acquisition steps of the multi-dimensional characteristic source data are as follows: According to the voltage change rate threshold and the current change rate threshold, the primary side voltage change rate, the primary side current change rate, the secondary side voltage change rate and the secondary side current change rate are compared with the threshold, and the time point meeting the threshold is determined as the trigger point. The primary side high-frequency waveform segment and the secondary side high-frequency waveform segment are intercepted with the trigger point as the center, and the transient event waveform set is formed; According to the transient event waveform set, the boundary of the trigger point time range is determined, all positions in the time range are excluded, and the interval is selected according to continuity in the remaining time range. The primary side high-frequency waveform segment and the secondary side high-frequency waveform segment are intercepted from the same position to form the steady-state waveform segment; According to the steady-state waveform segment and the transient event waveform set, the time stamps of the primary side high-frequency waveform segment and the secondary side high-frequency waveform segment are aligned, the primary side high-frequency waveform segment and the secondary side high-frequency waveform segment and the steady-state waveform segment are spliced, the trigger point identifier and the interval label are attached, the structure and field naming are unified, and the multi-dimensional characteristic source data is generated.

[0019] Specifically, according to the preset voltage change rate threshold and the current change rate threshold, first, the digitized high-frequency sampling sequences of the primary side voltage, the primary side current, the secondary side voltage and the secondary side current continuously collected are calculated in real time for change rate, which is realized by subtracting the values of adjacent two sampling points and dividing by the sampling time interval, for example, under the condition that the sampling frequency is 1 MHz, the calculation of the primary side voltage change rate is to subtract the voltage value of the previous sampling point from the voltage value of the current sampling point, and then divide by the time interval of 1 microsecond, the same calculation is performed on all four signal streams, and four synchronous change rate time sequences are obtained, next, the trigger threshold is set, the setting of the threshold is based on statistical analysis of the data collected in a reference period (for example, 10 minutes) under the normal operation state of the power grid, specifically, the average value and the standard deviation of the absolute values of the four change rate sequences in the reference period are calculated, and the voltage change rate threshold is set to the sum of 1.8 times the average value of the corresponding voltage change rate sequence and 2.5 times the standard deviation, and the current change rate threshold is set to the sum of 2.0 times the average value of the corresponding current change rate sequence and 3.0 times the standard deviation, taking the primary side voltage as an example, if the average value of its change rate is 200 V / s and the standard deviation is 50 V / s, then its trigger threshold is set to 1.8 * 200 + 2.5 * 50 = 360 + 125 = 485 V / s, then, the absolute values of the four change rate sequences calculated in real time are continuously compared with the respective thresholds, when the value of any one change rate first exceeds the set threshold, the time point is determined as the trigger point, once the trigger point is determined, the data of 100 milliseconds forward and 300 milliseconds backward is immediately intercepted with the timestamp of the trigger point as the center, forming a waveform segment with a total duration of 400 milliseconds, this operation is applied to the waveform data of the primary side voltage, the primary side current, the secondary side voltage and the secondary side current synchronously, thereby obtaining four high-frequency waveform segments that are completely aligned in time, and the four segments together constitute a complete record of a transient event, all transient event record sets generated by different trigger points form a transient event waveform set.

[0020] According to the transient event waveform set obtained in the foregoing step, first, traverse each transient event record in the set, extract the start timestamp and end timestamp of each record, since each transient event intercepts a 400 ms waveform centered on the trigger point, its time range is a clear closed interval, for example, if a trigger point occurs at time T, then its corresponding time range is [T - 100 ms, T + 300 ms], merge the time ranges of all transient events, if two or more time ranges overlap, merge them into a larger continuous time range, finally obtain one or more non-intersecting time periods marked as "transient", then, in the total time span of the entire data acquisition (for example, 24 hours), exclude all time periods marked as "transient", the remaining time region is the steady-state candidate region, then, in the steady-state candidate region, select a continuous interval long enough to intercept a steady-state waveform, set a minimum continuous time length standard, for example, 800 ms, to ensure that the selected section is indeed in a stable running state, scan all steady-state candidate regions, find all continuous sub-intervals with a length greater than or equal to 800 ms, from these sub-intervals that meet the conditions, select a fixed number of steady-state samples in an equal-interval sampling manner, for example, select 50, the length of each sample is fixed at 500 ms, select a 500 ms steady-state interval, then, from the same time position corresponding to the interval, simultaneously intercept high-frequency waveform data of the primary side voltage, the primary side current, the secondary side voltage, and the secondary side current, ensure that the four waveform segments are accurately aligned in time, after this operation, combine all selected steady-state samples to form a steady-state waveform segment.

[0021] According to the acquired steady-state waveform segment and transient event waveform set, first, the timestamp alignment check is performed, for each waveform segment in the set (whether transient or steady-state), check whether the start timestamp of the four waveform data sequences contained by it on the primary side and the secondary side is completely consistent, and whether the data point quantity is equal, ensure that there is no misalignment caused by sampling asynchronization or data loss, confirm that the alignment is correct, sort all waveform segments from the transient event waveform set and all waveform segments from the steady-state waveform segment according to the respective start timestamps, and splice them into a single, time-sequentially arranged complete data stream, in the splicing process, a new descriptive field is added for each data point, specifically, a field named "interval label" is created, for all data points derived from the transient event waveform set, the value of this field is set to 1, for all data points derived from the steady-state waveform segment, the value of this field is set to 0, at the same time, another field named "trigger point identification" is created, for data in the transient event, the field records the trigger point precise timestamp of the corresponding event, while for data in the steady-state segment, the value of the field is empty or set to a predefined invalid value, such as 0, then, the entire spliced data stream is subjected to structure unification processing, standardizing the naming of all data fields, such as naming "timestamp", "primary side voltage", "primary side current", "secondary side voltage", "secondary side current", "interval label" and "trigger point identification", and ensuring that the data types and units of all numerical fields remain consistent, such as voltage unit in volts (V), current unit in amperes (A), and timestamp unit in microseconds (μs), after the above alignment, splicing, labeling and structure unification operations, a structured, information-rich multi-dimensional feature source data is finally generated.

[0022] The acquisition step of the frequency band response difference sequence is: The transient waveform segment in the multi-dimensional feature source data is called, and the primary side waveform and the secondary side waveform are subjected to band-pass filtering and down-sampling at each level according to the multi-level frequency band index, the time axis of the primary side waveform and the secondary side waveform is kept aligned and the coefficient value sequence of each frequency band is recorded, forming a wavelet packet coefficient sequence; According to the wavelet packet coefficient sequence, the energy metric of the primary side wavelet packet coefficient sequence and the secondary side wavelet packet coefficient sequence is respectively calculated for each frequency band, and the phase metric is extracted, the ratio of the energy metric and the difference of the phase metric are calculated in the same frequency band, and the corresponding time identifier and frequency band index are attached, forming the frequency band response difference sequence.

[0023] Specifically, the waveform segment marked as a transient event in the multi-dimensional characteristic source data is called, specifically, for each transient event, the time series data of the primary side waveform and the secondary side waveform corresponding to the transient event is extracted, the "db4" (Daubechies 4) wavelet is selected as the base function, and the decomposition layer is set to 5 layers. This selection is because the "db4" wavelet achieves a good balance between time-frequency localization characteristics and computational complexity, and 5 layers of decomposition can divide the signal frequency into 32 uniform sub-bands, which is sufficient to cover various high-frequency transient components that may occur in the transformer. The decomposition process adopts a step-by-step traversal method, starting from the 0th layer (original signal). In the first layer, the primary side waveform and the secondary side waveform are respectively convolved with the low-pass filter and the high-pass filter corresponding to the "db4" wavelet to obtain the approximation component and the detail component. Then, the two components are 2 times down-sampled, that is, every other point is taken as a point, to generate the wavelet packet coefficient sequence of the two frequency bands in the first layer. In the second layer, the coefficient sequence of the two frequency bands generated in the first layer is repeatedly filtered and down-sampled, and each frequency band is divided into two sub-bands, generating a total of 4 frequency band coefficients. This process is iterated until the 5th layer, and finally 32 frequency band wavelet packet coefficient sequences of the primary side waveform and the secondary side waveform are generated. In the entire decomposition process, the same decomposition tree structure and filter are applied to the primary side and the secondary side waveforms, and the starting point of the down-sampling is consistent, so that the wavelet packet coefficient sequences of the primary side and the secondary side are strictly aligned in time in each frequency band. Finally, all the frequency band coefficients obtained after decomposition of each transient event are recorded in a structured manner. Each frequency band record contains the frequency band index (composed of the layer number and the position index of the layer, for example, (5, k), where k is from 0 to 31), and the coefficient value sequence of the corresponding primary side and secondary side. After collecting all the frequency band records, the wavelet packet coefficient sequence is formed.

[0024] According to the wavelet packet coefficient sequence generated in the previous step, each frequency band of each transient event is independently analyzed. First, the energy measure is calculated. For the frequency band with index k, the primary side wavelet packet coefficient sequence is denoted as , and the secondary side wavelet packet coefficient sequence is denoted as , where n is the coefficient point index. Then, the primary side energy measure and the secondary side energy measure are calculated by summing the squares of the respective coefficient sequences, i.e. and . Next, the phase measure is extracted by calculating the normalized cross-correlation function between the primary side and secondary side coefficient sequences to determine the time delay. The formula for calculating the cross-correlation function is , where is the time delay point number, and the delay point that makes reach the maximum value is found This delay represents the time difference of the two signals in this frequency band, then convert this time delay to phase difference, the formula is where is the center frequency of the kth frequency band, is the time sampling period corresponding to the wavelet packet coefficient of this layer, its value is the original signal sampling period multiplied by 2 raised to the power of the number of decomposition layers, after the completion of energy and phase measurement, in the same frequency band k, the ratio of energy measurement is calculated and directly use the calculated phase difference value Finally, the calculation results of each frequency band, including energy ratio and phase difference, are associated with the starting time identifier of the transient event and the index of the frequency band (for example, (5, k)), forming a record. The records generated by a transient event in all 32 frequency bands are collected to form a frequency band response difference sequence.

[0025] The acquisition steps of the dynamic error feature vector are: According to the frequency band response difference sequence, rearrange the frequency band indexes from low frequency to high frequency and keep the consistent order of time sequence, concatenate and fill in the placeholder identifier of the missing frequency band position according to the fixed splicing order, and generate the dynamic error feature vector.

[0026] Specifically, according to the obtained frequency band response difference sequence, first, reorder the frequency band indexes. The frequency band indexes generated by the original wavelet packet decomposition algorithm (usually in Gray code order) do not have a linear correspondence with the high and low frequencies, so the 32 frequency bands need to be renumbered from the lowest frequency to the highest frequency according to the center frequency actually corresponding to each frequency band index, for example, the frequency band with the lowest center frequency is marked as No. 1, the next lowest is No. 2, and so on, up to No. 32. During the rearrangement, the energy ratio and phase difference value of each frequency band also move, but they must be associated with the original time identifier, that is, all operations are completed within the data of the same transient event to ensure the consistency of the time sequence. Then, according to the new order of the frequency bands from No. 1 to No. 32, concatenate and splice the energy ratio and phase difference value of each frequency band to form a numerical sequence with a length of 64. The fixed splicing order is: the energy ratio of No. 1 frequency band, the phase difference value of No. 1 frequency band, the energy ratio of No. 2 frequency band, the phase difference value of No. 2 frequency band, …, the energy ratio of No. 32 frequency band, and the phase difference value of No. 32 frequency band. Before splicing, an effectiveness check is performed. Set an energy threshold, which is set to 0.001% of the total energy of the primary side of the transient event, for example, if the total energy of the primary side of a transient event is 10000 units, the energy threshold is 0.1 unit. For any frequency band, if its primary side energy measurement Below this threshold, it is considered that the signal of the frequency band is too weak, and the calculated energy ratio and phase difference value is unreliable, at this time, the splicing position corresponding to the frequency band is filled with a preset placeholder (for example -999.0), and after splicing and placeholder filling, the obtained numerical sequence with a length of 64 is the dynamic error feature vector of the transient event.

[0027] The acquisition step of the cyclic spectrum energy distribution diagram is: Call the steady-state waveform segment in the multi-dimensional feature source data, extract the secondary side signal time sequence and remove the DC bias, segment and record the time stamp according to the fixed length and fixed overlap rate, calculate the cyclic autocorrelation sequence segment by segment and accumulate along the time delay axis, perform complex exponential kernel integration along the time delay axis to obtain the frequency axis complex value distribution and establish the joint index of the frequency axis and the cyclic frequency axis, and obtain the cyclic spectrum density function; According to the cyclic spectrum density function, search for the maximum peak position of the energy distribution on the cyclic frequency axis, record the index corresponding to the maximum peak position as the main cyclic frequency index, and count the bandwidth where the main peak is located, search for the maximum peak position of the energy distribution on the frequency axis, record the maximum peak position as the main frequency index, and count the half-height width of the main peak, then according to the conditions that the cyclic frequency index is an integer multiple of the main cyclic frequency and the frequency index is located in the neighborhood range of the half-height width of the main frequency, mark the main cyclic frequency spectrum line energy set, at the same time, select the region where the cyclic frequency index is not an integer multiple of the main cyclic frequency and the frequency index is far away from the neighborhood range of the half-height width of the main frequency, count the upper quantile number of the energy distribution in the region as the noise power threshold, and mark the region as the non-main cyclic frequency spectrum line energy set, and form the cyclic spectrum energy distribution diagram.

[0028] Specifically, all waveform segments marked as steady-state in the multi-dimensional feature source data are called, and time series data of secondary side signals are extracted from each segment. For example, a 2-second length and 20 kHz sampling frequency steady-state waveform of secondary side current is selected to obtain a time series containing 40,000 sampling points. First, the average value of the sequence is calculated, and each data point in the sequence is subtracted by the average value to complete the removal of DC bias. Next, the fixed length of the segment is set to 2048 sampling points, and the fixed overlap rate is set to 50%, i.e. overlapping 1024 sampling points. The entire 40,000-point time series is segmented according to the rule, the first segment is from the first point to the 2048th point, the second segment is from the 1025th point to the 3072nd point, and so on. A total of 38 data segments are obtained, and the timestamp of the starting point in the original sequence is recorded for each segment. Then, the cyclic autocorrelation sequence of each of the 38 data segments is calculated, which is a two-dimensional function dependent on time delay and cycle frequency. After the calculation, the 38 cyclic autocorrelation sequences are averaged at the corresponding time delay and cycle frequency points to obtain a smoother and more stable estimation result. Subsequently, the average cyclic autocorrelation sequence is subjected to a discrete Fourier transform along the time delay axis, i.e. a complex exponential kernel integration operation. This transform converts the time delay domain function into a frequency domain complex function, and the result is a two-dimensional complex matrix, with the two dimensions being the frequency axis and the cycle frequency axis. Finally, a joint index is established for the two-dimensional matrix, with one index p corresponding to the regular frequency and the other index q corresponding to the cycle frequency. The two-dimensional complex matrix with joint index is the cyclic spectral density function to be solved.

[0029] According to the calculated cyclic spectral density function, which is a two-dimensional complex energy matrix with frequency index p and cycle frequency index q as coordinates, first, projection is performed on the cycle frequency axis, i.e. the energy values (i.e. the square of the modulus of the complex value) of all frequency index p corresponding to each cycle frequency index q are added to obtain a one-dimensional energy distribution curve about cycle frequency q. The peak point with the maximum energy value on the curve is searched, and the cycle frequency index corresponding to the peak point is recorded as the main cycle frequency index For a 50 Hz power system, the frequency theoretically corresponds to 100 Hz. Then, the bandwidth of the main peak is determined. The specific method is to find the position of half the peak energy and record the two cycle frequency indexes corresponding to the two sides of the main peak. The range between the two indexes is the bandwidth of the main peak. Next, projection is performed on the frequency axis in the same way, i.e. the energy values of all cycle frequency indexes q corresponding to each frequency index p are added to obtain an energy distribution curve about frequency p. The position of the maximum peak is searched, and the corresponding frequency index is recorded as the main frequency index , theoretically corresponds to 50Hz, and calculate the half-height width of this main peak, that is, the width at half the peak energy, and then start to mark the area of ​​the cyclic spectrum energy distribution diagram. The first condition is that the cyclic frequency index q is approximately equal to the main cyclic frequency index Integer multiples of bandwidth), and the frequency index p is located at the main frequency index The half-height width of the neighborhood (for example, in (within an area extending 1.5 times the half-height width on each side) all the The index pair is marked as the main cyclic frequency spectrum line energy set A. At the same time, the area that does not meet the above conditions is selected as the noise and distortion area. The specific selection criteria are that the cyclic frequency index q is not the main cyclic frequency index An integer multiple of , and the frequency index p is far away from the main frequency index Neighborhood (e.g., distance More than 5 times the half-height width), in all the areas selected in this way, count the energy values ​​of all points and calculate the 99th percentile of these energy values, and set this value as the noise power threshold At the same time, all The index pairs are marked as the non-main cyclic frequency spectrum line energy set B, and the cyclic spectrum density function after the above marking process forms a cyclic spectrum energy distribution diagram.

[0030] The steps to obtain the cyclic distortion index are: According to the cyclic spectrum energy distribution diagram, the cyclic distortion index is calculated using the following formula: ; in, is the cyclic distortion index, is the complex value of the cyclic spectral density function at frequency index p and cyclic frequency index q, is the index pair set corresponding to the main cycle frequency spectrum line energy set, is the index pair set corresponding to the non-main cycle frequency spectrum line energy set, , is the frequency index corresponding to the main frequency, is the loop frequency index corresponding to the main loop frequency, is the frequency attenuation factor, is the noise power threshold, is a non-negative truncation operator, defined as max , p is the frequency index, and q is the cycle frequency index.

[0031] Specifically, the formula: The above formula is to calculate the cycle distortion index which can reflect the signal distortion degree caused by the non-linear effect of the transformer core saturation by quantifying the ratio of the weighted energy of the non-primary cycle frequency component to the total energy of the primary cycle frequency component, by The item subtracts the background noise power before calculating the distortion energy and eliminates the energy contribution below the noise level by the non-negative truncation operator, which improves the detection sensitivity and noise immunity of the index to weak distortion, and introduces a special weight function The function consists of two parts, the logarithmic term makes the high-order cycle frequency components far from the fundamental cycle frequency obtain higher weights, because these high-order components are usually related to more serious non-linear distortion, and the exponential term makes the distortion components close to the main frequency in frequency obtain higher weights, so that the in-band distortion has a greater impact on the measurement accuracy. Through this weighting method, the cycle distortion index obtained not only reflects the total energy of the distortion, but also reflects the relative importance of the distortion components at different frequency positions.

[0032] The acquisition step of the parameter represents the complex value of the cycle spectrum density function at the frequency index p and the cycle frequency index q, which is obtained by a series of signal processing on the steady waveform segment in the multi-dimensional characteristic source data, including extracting the secondary side signal, removing the DC bias, segmenting, calculating and averaging the cycle autocorrelation sequence, and finally performing Fourier transform to obtain a two-dimensional complex matrix, which is the basic data source for the calculation of the formula, and its value directly reflects the energy and phase information of the signal at a certain frequency and cycle frequency. For example, in one calculation, the index pair is located in set A, and the corresponding complex value is , and the index pair in set B is .

[0033] The acquisition step of the parameter represents the index pair corresponding to the main cycle frequency spectrum line energy set, which is first determined by the energy projection method. The main frequency index and the main cycle frequency index are determined, and the half-width and bandwidth of the spectral peak are calculated. Then, according to the preset condition that the cycle frequency index is an integer multiple of and the frequency index is located in the neighborhood of , all index pairs that meet the condition are selected to form set A, which represents the main periodic component in the signal and is the basis for calculating the total energy of the signal. For example, by calculation, we get , , the set A is determined to contain the index pairs of a series of points.

[0034] The parameter acquisition step is that the parameter represents the index pairs of the set of non-primary cycle frequency spectral line energy, the identification process is synchronous with the determination of the set A, and the range of the set is defined by selecting the integer multiple of the cycle frequency index which is not the integer multiple of the primary cycle frequency index , and the frequency index is far away from the primary frequency index neighborhood, and the set represents the non-periodic or weak periodic components generated by non-linear distortion, noise and other factors, and is the main object for measuring the degree of signal distortion, for example, in an analysis, the set B is determined to contain the index pairs and and the like.

[0035] The parameter acquisition step is that the parameter is a weight function, and the calculation formula is , which is calculated according to the index of the current point and the values of the predetermined primary frequency index , the primary cycle frequency index and the frequency attenuation factor , the function does not need to collect data from the outside, but is dynamically calculated according to the characteristics of the cycle spectrum itself, and is used to adjust the contribution of distortion components at different positions in the final index calculation.

[0036] The parameter acquisition step is that the parameter is the frequency index corresponding to the primary frequency, which is determined by searching the position of the maximum peak point on the frequency energy distribution curve obtained by summing the energy of the cycle spectrum density function along the cycle frequency axis, which identifies the position of the fundamental frequency with the most concentrated energy in the signal, and for the 50Hz power system in China, if the frequency resolution is 1Hz / index, the value of should be 50.

[0037] The parameter acquisition step is that the parameter is the cycle frequency index corresponding to the primary cycle frequency, which is determined in a similar manner to , by searching the position of the maximum peak point on the cycle frequency energy distribution curve obtained by summing the energy of the cycle spectrum density function along the frequency axis, which usually corresponds to twice the fundamental frequency, i.e. 100Hz, and if the cycle frequency resolution is 2Hz / index, the value of should be 50.

[0038] The parameter is the frequency attenuation factor, which is used to control the attenuation speed of the Gaussian part in the weight function. The value is set according to the spectrum width of the main frequency peak. A reasonable setting method is to associate the value with the standard deviation or half-height width of the energy distribution of the main frequency peak. After calculating the half-height width (FWHM) of the main frequency peak, A fixed multiple of FWHM can be set, for example, set If the measured half-height width of the main frequency peak is 10 index units, then It can be set to 8.49. Here, a fixed value based on empirical adjustment is set for ease of calculation This value ensures sensitivity to in-band distortion while also considering a certain range of out-of-band distortion.

[0039] The parameter is the noise power threshold. The energy values of all points in the selected non-main cyclic frequency spectrum line energy set B region in the cyclic spectrum energy distribution diagram are counted A data set is formed by these energy values, and the 99th percentile of the data set is calculated as the value of For example, 10,000 energy points are collected in set B region, sorted from small to large, and the value of the 9,900th point is taken. If the value is 0.08, then .

[0040] According to the parameter, the calculation is as follows: The setting example parameters are as follows: Set A contains one point , and its energy .

[0041] Set B contains two points and , and their energies are respectively: .

[0042] .

[0043] Other parameters: , , , .

[0044] Calculate the denominator: ; Calculate the weight of each point in the numerator : For point : ; ; ; For the point : ; ; ; The calculation of the molecular: ; ; ; ; ; The calculation of the cycle distortion index : ; The results show that in this steady waveform analysis, the calculated cycle distortion index is 0.001236, and the evaluation standard is set as follows: if , it indicates that the transformer is working in the linear region, and the distortion level is very low, if , it indicates that there is slight nonlinear distortion, if , it indicates that the distortion is relatively serious, which may be caused by core saturation, etc. The current calculation result 0.001236 is much smaller than 0.005, which shows that the steady-state segment corresponds to the good operation state of the transformer, and the nonlinear distortion is very low.

[0045] The steps for obtaining the comprehensive error correction matrix are: Call the dynamic error feature vector and the cycle distortion index, unify the dimension range and establish a one-to-one correspondence with the time stamp and the frequency band index, splice the value bits in the order of fixed fields, and carry out threshold rejection and linear interpolation of missing positions to obtain the error fusion representation sequence; According to the error fusion representation sequence, read the condition threshold and conversion item of the error mapping rule, match the mapping items one by one according to the frequency band index and the distortion level, and split the value bits into amplitude channel and phase channel, respectively, for conversion and saturation clipping, generate channel identifier and time identifier, obtain the compensation amount of amplitude and phase, and summarize the comprehensive error correction matrix.

[0046] Specifically, the dynamic error feature vector and the cycle distortion index generated by the foregoing steps are called. First, the two different features are dimensionally unified. The energy ratio in the dynamic error feature vector is dimensionless, and the phase difference value is in radians. In order to facilitate fusion, the phase difference value is normalized to by dividing by In the interval, the energy ratio and the cycle distortion index are in the appropriate range, and no additional processing is required. Then, based on the timestamp, the dynamic error feature vector corresponding to each transient event is associated with the cycle distortion index calculated during the nearest steady-state period before and after the transient event. Since a dynamic error feature vector contains information of 32 frequency bands, and the cycle distortion index is a scalar, the value of the cycle distortion index is copied 32 times to establish a one-to-one correspondence with the 32 frequency band data in the dynamic error feature vector, forming a data table containing the timestamp, frequency band index, normalized energy ratio, normalized phase difference, and cycle distortion index. Then, according to the fixed field order of "frequency band index", "normalized energy ratio", "normalized phase difference", and "cycle distortion index", the data under each timestamp is spliced into a longer feature sequence, and the sequence is subjected to abnormal data processing. An abnormal jump rejection threshold is set, which is determined by statistical analysis of the feature sequence calculated from a reference data set (e.g., 24 hours of data collected at the beginning). Specifically, the average value and standard deviation of the difference between each feature field at adjacent timestamps are calculated, and the threshold is set to the average value plus 3.5 times the standard deviation. For example, if the average value of the energy ratio difference is 0.01 and the standard deviation is 0.005, the jump threshold is 0.01 + 3.5 * 0.005 = 0.0275. Traverse the spliced feature sequence, if the absolute value of the difference between a data point and the previous data point exceeds the threshold, mark the point as abnormal and delete it. For the missing positions originally existing in the dynamic error feature vector (marked by placeholder -999.0) and the empty positions generated after the abnormal jump is deleted, linear interpolation is used for filling, i.e., using the average value of the nearest two valid data points before and after the position to fill. After all the above processing is completed, a complete and uniform error fusion representation sequence is obtained.

[0047] According to the error fusion representation sequence generated in the last step, first, a preset error mapping rule is loaded, which is a multi-dimensional lookup table established through offline experiments and simulation. The construction process is as follows: in the laboratory environment, different types and intensities of transient disturbances and different degrees of saturation excitation are applied to the standard transformer, and the corresponding dynamic error feature vector and cycle distortion index are calculated by accurately measuring the primary side and secondary side signals. Compare with the actual amplitude error and phase error measured by the standard instrument, and through regression analysis, decision tree and other machine learning methods, the mapping relationship from "frequency band index", "distortion level" (divided by cycle distortion index) to "amplitude compensation coefficient" and "phase compensation angle" is established, and this relationship is solidified as a rule table. Each row of the table is a conversion entry, including condition thresholds (such as frequency band index range, distortion level) and corresponding conversion parameters. For example, a rule may be: when the frequency band index is between 5-8 and the distortion level is "medium" (cycle distortion index is between 0.02 and 0.05), the amplitude compensation coefficient is 1.0015 and the phase compensation angle is -0.002 radians. Next, the data in the error fusion representation sequence is processed one by one. For each data, first read its frequency band index, and according to the value of its cycle distortion index, determine its distortion level according to the preset distortion level threshold (for example: <0.02 is "low", 0.02-0.05 is "medium", >0.05 is "high"). Then, use the frequency band index and distortion level as the key to search for the matching conversion entry in the error mapping rule table. After finding the matching entry, the normalized energy ratio and normalized phase difference values in the data are split out and sent to the amplitude channel and phase channel respectively for conversion. In the amplitude channel, the energy ratio is multiplied by the found amplitude compensation coefficient, and in the phase channel, the phase difference value is added to the found phase compensation angle. After conversion, saturation clipping processing is performed. Set a reasonable range for the amplitude compensation amount, for example , a reasonable range for the phase compensation amount, for example radian. If the calculated compensation amount exceeds this range, it will be forcibly set to the boundary value of the range. Finally, the calculated amplitude and phase compensation amounts are attached with channel identifiers ("amplitude" or "phase") and original time identifiers. All time points and all frequency band compensation amounts are summarized to form a comprehensive error correction matrix.

[0048] The online calibration coefficient set acquisition step is: According to the amplitude and phase compensation amounts in the comprehensive error correction matrix, fill in the comprehensive error correction matrix according to the primary side and secondary side and channel order, and expand it into correction parameters according to the register address mapping table. Write each item into the measurement chip register and read back to compare the checksum and status bit, and record the version number and time stamp to obtain the online calibration coefficient set.

[0049] Specifically, according to the generated comprehensive error correction matrix, which contains the compensation amounts of amplitude and phase calculated under different frequency bands and different distortion levels, first, the compensation amounts are organized into a structured data block according to a predetermined channel order, which is usually the primary side voltage amplitude compensation, the primary side voltage phase compensation, the primary side current amplitude compensation, the primary side current phase compensation, and the corresponding four compensation amounts of the secondary side. Since the method mainly calibrates the secondary side measurement error, the compensation amounts of the primary side are usually set to zero, and the compensation amounts of the secondary side are directly taken from the corresponding values in the comprehensive error correction matrix. After filling, the register address mapping table of the metering chip is loaded, which defines the addresses of each register in the chip for storing correction coefficients and their functions in detail, for example, address 0x1000 is used to store the amplitude gain correction coefficient of the voltage channel, address 0x1002 is used to store the phase correction coefficient, etc. Then, the filled structured data block is expanded into a series of specific correction parameters according to the requirements of the address mapping table, which may involve data format conversion, such as converting the compensation amounts of floating point type into fixed point numbers or integer format recognizable by the chip. Next, through the communication interface (such as SPI or I2C), these correction parameters are written into the corresponding register addresses of the metering chip one by one. After writing each parameter, a read-back operation is immediately performed to read data from the same address and compare it with the written value. At the same time, the state bit register of the chip is checked to confirm whether the write operation is successful and the error flag is set. In addition, the checksum (such as CRC16) of all written parameters is calculated and compared with the checksum automatically calculated by the chip according to the written data. Only when the read-back data is consistent, the state bit is normal, and the checksum matches, the parameter is confirmed to be successfully written. After all parameters are successfully written, the version number (a self-increasing serial number) and the time stamp accurate to seconds of the current calibration operation are recorded in a specific information register. The set of correction parameters that have been written and verified successfully finally forms the online calibration coefficient set.

[0050] The above is only a preferred embodiment of the present application, and does not limit the present application in other forms. Any skilled person in the art can modify or change the above disclosed technical content to equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made according to the technical essence of the present application to the above embodiments without departing from the technical solution content of the present application still falls within the protection scope of the present application.

Claims

1. A method of on-line calibration of power transformer error, characterized by, The following steps are involved: By setting the voltage and current change rate thresholds to trigger data acquisition, high-frequency waveform data from the primary and secondary sides of the transformer are synchronously acquired. Waveform segments of the transient period are intercepted with the trigger point as the center to form a transient event waveform set. Continuous steady-state waveform segments are then selected from the area outside the transient event to establish multi-dimensional feature source data. Recalling transient waveform segments in the multidimensional feature source data, performing multi-level wavelet packet decomposition on the primary and secondary waveforms, calculating the energy ratio and phase difference of the primary and secondary coefficients in each frequency band to form a frequency band response difference sequence, reconstructing and arranging the frequency band response difference sequences for all frequency bands, and generating a dynamic error feature vector; Recalling steady-state waveform segments from the multidimensional feature source data, extracting energy distribution information on the frequency axis and the cyclic frequency axis, constructing a cyclic spectrum energy distribution diagram, separating the main cyclic frequency spectrum line energy and the non-main cyclic frequency spectrum line energy introduced by core saturation based on the cyclic spectrum energy distribution diagram, and calculating a cyclic distortion index; The dynamic error eigenvector and the cyclic distortion index are fused, and according to a preset error mapping rule, the fusion result is converted into compensation amounts of amplitude and phase, a comprehensive error correction matrix is ​​established, correction parameters are generated according to the comprehensive error correction matrix, and the correction parameters are written into the metering chip register of the electric energy meter to obtain an online calibration coefficient set.

2. The method of on-line calibration of power transformer error according to claim 1, characterized in that, The steps for obtaining the multi-dimensional feature source data are: According to the voltage change rate threshold and the current change rate threshold, the primary side voltage change rate, the primary side current change rate, the secondary side voltage change rate, and the secondary side current change rate are compared with the threshold, and the moment when the threshold is met is determined as the trigger point. The primary side high-frequency waveform segment and the secondary side high-frequency waveform segment are intercepted with the trigger point as the center to form a transient event waveform set; Determine the boundaries of the trigger point time range based on the transient event waveform set, exclude all positions within the time range, select intervals based on continuity in the remaining time range, and intercept the primary side high-frequency waveform segment and the secondary side high-frequency waveform segment from the same position to form a steady-state waveform segment; According to the steady-state waveform segment and the transient event waveform set, the timestamps of the primary-side high-frequency waveform segment and the secondary-side high-frequency waveform segment are aligned, the primary-side high-frequency waveform segment, the secondary-side high-frequency waveform segment and the steady-state waveform segment are spliced, trigger point identifiers and interval labels are added, the structure and field naming are unified, and multi-dimensional feature source data is generated.

3. The method of on-line calibration of power transformer error according to claim 1, wherein, The steps for obtaining the frequency band response difference sequence are: Calling transient waveform segments in the multidimensional feature source data, traversing level by level according to the multi-level frequency band index, performing bandpass filtering and downsampling on the primary side waveform and the secondary side waveform at each level, keeping the time axis of the primary side waveform and the secondary side waveform aligned and recording the coefficient value sequence of each frequency band to form a wavelet packet coefficient sequence; According to the wavelet packet coefficient sequence, the energy measurements of the primary side wavelet packet coefficient sequence and the secondary side wavelet packet coefficient sequence are counted for each frequency band and the phase measurement is extracted. The ratio of the energy measurement and the difference of the phase measurement are calculated within the same frequency band, and the corresponding time mark and frequency band index are attached to form a frequency band response difference sequence.

4. The method of on-line calibration of power transformer error of claim 1, wherein, The steps for obtaining the dynamic error characteristic vector are: According to the frequency band response difference sequence, the multi-level frequency band indexes are rearranged from low frequency to high frequency while maintaining the consistent order of the time series. The energy ratios and phase difference values ​​of all frequency bands are concatenated in a fixed splicing order and the placeholders of the missing frequency band positions are filled in to generate a dynamic error feature vector.

5. The method of on-line calibration of power transformer error as claimed in claim 1, wherein, The steps for obtaining the cyclic spectrum energy distribution diagram are: Recalling steady-state waveform segments from the multidimensional feature source data, extracting a secondary-side signal time series and removing a DC bias, segmenting the data into segments with a fixed length and a fixed overlap ratio and recording timestamps, calculating a cyclic autocorrelation sequence segment by segment and accumulating the segments along a time delay axis, performing a complex exponential kernel integration along the time delay axis to obtain a complex-valued distribution along the frequency axis, and establishing a joint index between the frequency axis and the cyclic frequency axis to obtain a cyclic spectral density function; According to the cyclic spectral density function, the maximum peak position of the energy distribution is searched on the cyclic frequency axis, the index corresponding to the maximum peak position is recorded as the main cyclic frequency index, and the bandwidth of the main peak is counted; the maximum peak position of the energy distribution is searched on the frequency axis, the maximum peak position is recorded as the main frequency index, and the half-height width of the main peak is counted; then, according to the condition that the cyclic frequency index is an integer multiple of the main cyclic frequency and the frequency index is located in the neighborhood range of the half-height width of the main frequency, the main cyclic frequency spectrum line energy set is marked; at the same time, the area where the cyclic frequency index is not an integer multiple of the main cyclic frequency and the frequency index is far away from the neighborhood range of the half-height width of the main frequency is selected, the upper quantile of the energy distribution in the area is counted as the noise power threshold, and the area is marked as the non-main cyclic frequency spectrum line energy set to form a cyclic spectrum energy distribution diagram.

6. The method of on-line calibration of power transformer error of claim 1, wherein, The step of obtaining the cyclic distortion index is: calculating the cyclic distortion index according to the cyclic spectrum energy distribution diagram.

7. The method of online calibration of power transformer error as claimed in claim 1 wherein, The steps for obtaining the comprehensive error correction matrix are: The dynamic error feature vector and cyclic distortion index are called, the dimension range is unified, and a one-to-one correspondence is established between the timestamp and the frequency band index. The numerical bits are spliced ​​in a fixed field order, and abnormal jumps are thresholded and linearly interpolated for missing positions to obtain an error fusion representation sequence. According to the error fusion characterization sequence, the conditional thresholds and conversion entries of the error mapping rules are read, the mapping entries are matched one by one according to the frequency band index and the distortion level, and the numerical bits are split into amplitude channels and phase channels, which are converted and saturated respectively. The channel identifier and time identifier are generated, and the compensation amounts of amplitude and phase are obtained, which are summarized into a comprehensive error correction matrix.

8. The method of online calibration of power transformer error as claimed in claim 1 wherein, The steps for obtaining the online calibration coefficient set are: According to the compensation amounts of amplitude and phase in the comprehensive error correction matrix, the comprehensive error correction matrix is ​​filled in the order of primary side, secondary side and channel, and expanded into correction parameters according to the register address mapping table. The correction parameters are written into the metering chip register item by item, and the checksum and status bits are read back and compared, and the version number and timestamp are recorded to obtain the online calibration coefficient set.

9. The on-line calibration system of the on-line calibration method of errors of power transformers according to any of claims 1-8, characterized in that, include: The data acquisition module triggers data acquisition by setting voltage and current change rate threshold, synchronously acquires high-frequency waveform data of the primary side and the secondary side of the mutual inductor, intercepts waveform fragments of transient period with the trigger point as the center to form a transient event waveform set, selects continuous steady-state waveform fragments from the area outside the transient event, and establishes multi-dimensional characteristic source data; The dynamic feature extraction module is configured to call the transient waveform fragments in the multi-dimensional characteristic source data, perform multi-level wavelet packet decomposition on the primary side and the secondary side waveforms, calculate the energy ratio and phase difference of the primary side and the secondary side coefficients in each frequency band to form a frequency band response difference sequence, reconstruct and arrange the frequency band response difference sequences of all frequency bands, and generate a dynamic error feature vector; The steady-state feature extraction module is configured to call the steady-state waveform fragments in the multi-dimensional characteristic source data, extract energy distribution information on the frequency axis and the cyclic frequency axis, construct a cyclic spectrum energy distribution diagram, separate main cyclic frequency spectrum line energy and non-main cyclic frequency spectrum line energy introduced by core saturation according to the cyclic spectrum energy distribution diagram, and calculate a cyclic distortion index; The error correction module is configured to fuse the dynamic error feature vector and the cyclic distortion index, convert the fusion result into compensation amounts of amplitude and phase according to a preset error mapping rule, establish a comprehensive error correction matrix, generate a correction parameter according to the comprehensive error correction matrix, and write the correction parameter into a metering chip register of the electric energy meter to obtain an online calibration coefficient set.

Images