Arc direction discrimination method and system based on double-path characteristic current ratio
By calculating the ratio of the dual characteristic currents in an arc event and combining a sliding time window and a forgetting ratio weighted update method, the problem of transient fluctuation interference in arc direction discrimination is solved, and more accurate and efficient arc direction identification is achieved.
Patent Information
- Application Number
- CN202611074327.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-20
- Publication Date
- 2026-11-17
AI Technical Summary
Existing techniques for determining arc direction based on the ratio of dual currents result in a high misjudgment rate when dealing with transient fluctuations during the arc initiation and extinction phases, and cannot effectively isolate interference from transient current fluctuations in the arc.
By calculating the ratio of the peak changes of the time-synchronized positive current characteristic and the associated current characteristic in the same arc event, and combining the volatility within the sliding time window to determine the stable combustion stage, an initial ratio benchmark is generated through moving average processing. The dynamic ratio benchmark is then updated using a forgetting ratio weighting. Finally, the arc direction is determined based on the polarity of the offset.
It effectively isolates the interference of transient current fluctuations during arc initiation and extinction, reduces the probability of misjudgment in arc direction detection, and improves the stability of detection and overall computational efficiency.
Smart Images

Figure CN123112596A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of arc detection technology, and in particular to an arc direction discrimination method and system based on the ratio of dual characteristic currents. Background Technology
[0002] Arc detection is a core technology in the field of electrical fault protection. By collecting line current signals, it can identify arc faults and distinguish the direction of arc generation, accurately locate fault points such as line short circuits and insulation damage, and provide a basis for judgment for the safety protection of power distribution circuits. The dual-channel synchronous current acquisition scheme is widely used in arc direction discrimination scenarios because it can simultaneously acquire the current change characteristics of two channels. The ratio of the two current channels is the core characteristic parameter for distinguishing the positive and negative directions of the arc.
[0003] Existing techniques for determining arc direction based on the ratio of dual current sources typically employ a fixed reference or a simple dynamic reference combined with threshold comparison. However, the arc combustion process exhibits distinct phases: the current fluctuates dramatically during the arc initiation and extinction phases, while the ratio remains relatively stable during the steady combustion phase. Current methods process all data from arc initiation to extinction uniformly during reference construction and direction determination. This directs transient fluctuations during initiation and extinction phases into the reference calculation and direction determination, leading to frequent polarity jumps in the offset and a persistently high arc direction misjudgment rate. Therefore, suppressing the misjudgment of direction caused by transient arc current fluctuations has become a pressing technical problem. Summary of the Invention
[0004] In view of the above problems, the purpose of this invention is to provide an arc direction discrimination method and system based on the ratio of dual characteristic currents, so as to solve the problem of misjudgment of arc direction caused by transient current fluctuations.
[0005] The arc direction determination method based on the ratio of dual characteristic currents provided by this invention includes: Step 1: Take the ratio of the peak change of the time-synchronized positive current characteristic to the associated current characteristic in the same arc event as the dual-path characteristic current ratio. Step 2: Calculate the volatility of the dual-path characteristic current ratio sequence within the sliding time window. When the volatility of multiple consecutive time windows is less than the preset stability threshold, it is determined that the electric arc has entered the stable combustion stage. Step 3: Extract the ratio sequence of the stable combustion stage and perform moving average processing to obtain the initial ratio benchmark sequence, and use the last value of the initial ratio benchmark sequence as the initial value of the dynamic ratio benchmark. Step 4: After the stable combustion stage ends, the dynamic ratio benchmark of the previous moment and the dual characteristic current ratio of the current moment are weighted and summed according to a preset forgetting ratio, and the summation result is used as the ratio change benchmark of the current moment to continuously update the dynamic ratio benchmark. Step 5: Take the difference between the ratio of the dual characteristic currents at the current moment and the ratio change reference at the corresponding moment as the offset. When the offset maintains the same polarity in a continuous time period, determine the arc direction based on the polarity.
[0006] This invention also provides an arc direction discrimination system based on the ratio of dual characteristic currents, the system comprising: The characteristic ratio calculation module is used to take the ratio of the peak change of the time-synchronized positive current characteristic and the associated current characteristic in the same arc event as the dual-path characteristic current ratio. The steady-state discrimination module is used to calculate the volatility of the dual-path characteristic current ratio sequence within a sliding time window. When the volatility of multiple consecutive time windows is less than a preset stability threshold, the arc is determined to have entered a stable combustion stage. The baseline initialization module is used to extract the ratio sequence of the stable combustion stage, perform moving average processing to obtain an initial ratio baseline sequence, and use the last value of the initial ratio baseline sequence as the initial value of the dynamic ratio baseline. The benchmark update module is used to continuously update the dynamic ratio benchmark after the stable combustion stage ends, by weighting and summing the dynamic ratio benchmark of the previous moment and the dual characteristic current ratio of the current moment according to a preset forgetting ratio, and using the summing result as the ratio change benchmark of the current moment. The direction determination module is used to take the difference between the ratio of the dual characteristic currents at the current time and the ratio change reference at the corresponding time as the offset. When the offset maintains the same polarity in a continuous time period, the direction of the electric arc is determined according to the polarity.
[0007] As can be seen from the above technical solution, the arc direction discrimination method and system based on the ratio of dual characteristic currents provided by the present invention accurately identifies the stable combustion stage of the arc by calculating the fluctuation rate of the ratio of dual characteristic currents through a sliding time window. Based on the stable ratio of this stage, dynamic benchmark initialization is completed. Subsequently, the benchmark is adaptively updated by combining the forgetting ratio weighting and adding amplitude limiting. Then, the arc direction is determined by continuously unifying the polarity of the offset. This can effectively isolate the interference caused by transient current fluctuations during arc initiation and extinguishing, reduce the polarity jump of the offset, and significantly reduce the probability of misjudgment of arc direction. At the same time, the entire set of hierarchical judgment logic simplifies invalid calculations and significantly improves the overall calculation efficiency and recognition stability of arc direction discrimination. Attached Figure Description
[0008] Other objects and results of the invention will become more apparent and readily understood by referring to the following description taken in conjunction with the accompanying drawings, and with a more complete understanding of the invention. In the drawings: Figure 1 This is a flowchart illustrating the arc direction determination method based on the ratio of dual characteristic currents according to an embodiment of the present invention. Figure 2 This is a functional block diagram of the arc direction discrimination system based on the ratio of dual characteristic currents according to an embodiment of the present invention; Figure 3 This is a distribution chart of the overall discrimination accuracy data for four types of samples in an embodiment of the present invention; Figure 4 This is a graph showing the variation of the steady threshold accuracy in an embodiment of the present invention. Figure 5 This is a graph showing the change in the accuracy of the forgetting rate in an embodiment of the present invention. Detailed Implementation
[0009] Existing arc direction discrimination schemes directly compare the ratio of dual characteristic currents using a fixed reference. They do not distinguish the stable burning stage of the arc, lack steady-state judgment logic based on the sliding window volatility, and the dynamic reference does not incorporate forgetting ratio weighted updates and lacks amplitude limiting protection. The polarity offset judgment lacks continuous threshold constraints and is easily affected by transient current fluctuations during arc initiation and extinction, causing frequent polarity jumps and a high misjudgment rate. This invention aims to solve the technical problem of arc direction identification misjudgment caused by such transient current fluctuations.
[0010] To address the aforementioned problems, this invention provides an arc direction discrimination method and system based on the ratio of dual characteristic currents. The specific embodiments of this invention will be described in detail below with reference to the accompanying drawings.
[0011] To illustrate the arc direction discrimination method and system based on the ratio of dual characteristic currents provided by this invention Figure 1 An exemplary illustration is provided for the arc direction discrimination method based on the ratio of dual characteristic currents in this invention. Figure 2 An exemplary illustration is provided for the arc direction discrimination system based on the ratio of dual characteristic currents in an embodiment of the present invention.
[0012] The following description of exemplary embodiments is merely illustrative and is in no way intended to limit the invention or its application or use. Technologies and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such technologies and equipment should be considered part of the specification.
[0013] Reference Figure 1The diagram shown is a flowchart illustrating the arc direction determination method based on the ratio of dual-path characteristic currents according to an embodiment of the present invention. In this embodiment, the arc direction determination method based on the ratio of dual-path characteristic currents includes: Step 1: Take the ratio of the peak change of the time-synchronized positive current characteristic to the associated current characteristic in the same arc event as the dual-path characteristic current ratio. In practice, the positive current signal and the associated current signal in the power distribution circuit are first synchronously acquired and preprocessed. The peak value changes of the two signals are extracted and the ratio is calculated to obtain the dual characteristic current ratio sequence as the basic characteristic quantity for subsequent steady-state determination and direction discrimination.
[0014] In some embodiments, this application preferably uses a combination of time-series synchronous sampling and extreme point screening to extract the peak value variation.
[0015] The forward current channel and the associated current channel are sampled synchronously at the same sampling frequency, with the sampling frequency ranging from 10kHz to 100kHz. The preferred sampling frequency in this application is 25.6kHz.
[0016] The engineering basis for selecting this range is that: when the sampling point density is below 10kHz, it is insufficient to accurately capture the peak details within the half-wave of the arc current, which can easily cause deviations in the calculation of peak value changes; when the sampling data volume is above 100kHz, it increases linearly, which greatly increases the computational load and storage overhead of the embedded processor, which is not conducive to the real-time requirements of the field protection device.
[0017] 25.6kHz is one of the standard sampling rates for power system fault recording. It can ensure that each power frequency half-wave contains 256 sampling points to accurately locate the peak position, and it is compatible with the hardware sampling circuit of existing relay protection devices without the need for additional sampling chip upgrades.
[0018] By selecting the above sampling frequency, an optimal balance can be achieved between peak recognition accuracy and system computational load.
[0019] After sampling, interference extreme points are removed from the positive half-wave sequence and the associated half-wave sequence respectively. The preset tolerance threshold ranges from 15% to 40% of the half-wave average amplitude. In this application, the preferred tolerance threshold is 25% of the half-wave average amplitude.
[0020] The physical basis for selecting this parameter is that the normal peak fluctuation within half-wave of the arc current usually does not exceed 15% of the average amplitude. A threshold below 15% will misjudge the normal waveform distortion as an interference point and remove the effective peak value, resulting in a distortion in the calculation of the peak value change. A threshold above 40% cannot effectively filter out instantaneous interference such as sampling spikes and electromagnetic pulses. The noise data mixed in at the extreme points directly affects the stability of the ratio.
[0021] A 25% threshold setting can cover the amplitude range of most on-site electromagnetic interference while preserving the true peak fluctuation characteristics during the arc combustion process.
[0022] By eliminating the aforementioned interference extreme points, the interference of sampling noise on peak extraction can be effectively suppressed, thereby improving the calculation stability of the dual-path characteristic current ratio.
[0023] After the extreme point screening is completed, the amplitude difference between adjacent effective peak points in each half-wave is taken as the peak value change of that half-wave, and the corresponding positive peak value change and associated peak value change are obtained. Then, the ratio of the positive peak value change and the associated peak value change under the same half-wave is taken as the dual characteristic current ratio.
[0024] By calculating the ratio of the two characteristic currents, the absolute amplitude changes of the two currents can be converted into relative ratio characteristics, eliminating the influence of common-mode factors such as power supply voltage fluctuations and line impedance changes on arc direction discrimination, and improving the anti-interference capability of characteristic quantities.
[0025] In one feasible implementation, the peak value variation is calculated using the peak-to-trough difference method. Specifically, the maximum and minimum values within each half-wave are extracted, and the difference between their amplitudes is taken as the peak value variation of that half-wave.
[0026] This scheme is suitable for low-voltage power distribution scenarios with small arc current waveform distortion and obvious single-peak characteristics within half-wave. Its advantages are simple algorithm logic and minimal computation, requiring only one traversal of half-wave data, making it suitable for implementation with resource-constrained low-end microcontrollers. Its relative disadvantage is that when a multi-peak structure appears within half-wave, the difference between a single peak and valley cannot accurately reflect the overall characteristics of current change, and the accuracy of ratio calculation decreases slightly.
[0027] In another feasible implementation, the root mean square increment method is used to calculate the peak change. Specifically, the difference between the current half-wave current RMS value and the previous half-wave RMS value is used as an equivalent representation of the peak change.
[0028] This solution is suitable for medium-voltage power distribution scenarios with unstable arc combustion and severe waveform distortion. Its advantage lies in the fact that the statistical characteristics of the effective value can smooth out local spike interference and the ratio results are more statistically stable. Its relative disadvantage is that the calculation of the effective value involves square and square root operations, and the processor operation time is about 3 to 4 times that of the peak method, which has a certain delay for protection scenarios with high real-time requirements.
[0029] In another feasible implementation, the peak value change is calculated using a multi-peak weighted method. Specifically, the top three peak values within a half-wave are extracted, and the weighted peak value difference between adjacent half-waves is calculated by summing the peak values according to their amplitude weights.
[0030] This scheme is suitable for special operating conditions where the fault arc waveform is complex and there are multiple local peaks, such as nonlinear load circuits such as the output side of the frequency converter and the load of the switching power supply. Its advantage is that it can comprehensively use the information of multiple peaks within half a wave, which can more comprehensively characterize the changing characteristics of the arc current and has the highest accuracy in direction determination. Its relative disadvantage is that the peak sorting and weighted calculation increase the algorithm complexity and require higher processor computing power.
[0031] Step 2: Calculate the volatility of the dual-path characteristic current ratio sequence within the sliding time window. When the volatility of multiple consecutive time windows is less than the preset stability threshold, it is determined that the electric arc has entered the stable combustion stage. In practice, after obtaining the dual characteristic current ratio sequence, the volatility of the ratio sequence is calculated point by point using a sliding time window. By comparing the volatility of the continuous multi-window with the stability threshold, it is determined whether the arc has entered the stable combustion stage, thus realizing the automatic division between the transient and steady states of the arc.
[0032] In some embodiments, this application preferably uses a fixed-length sliding window combined with a composite volatility index to determine steady state.
[0033] The window width of the sliding time window ranges from 5 and a half waves to 20 and a half waves, and the preferred window width in this application is 10 and a half waves.
[0034] The engineering basis for this parameter selection is as follows: when the window width is less than 5 half-waves, the number of data samples in the window is insufficient, the volatility calculation results are greatly affected by a single outlier, and the stability judgment is prone to false triggering; when the window width is greater than 20 half-waves, the steady-state judgment time lag is too long, and it is impossible to capture the moment when the arc enters the stable combustion stage in time, resulting in the compression of effective data in the baseline initialization stage.
[0035] The 10 half-waves correspond to a time span of approximately 100 milliseconds under the power frequency cycle, which not only includes sufficient statistical samples to ensure the reliability of volatility calculation, but also controls the steady-state determination delay to the order of hundreds of milliseconds, meeting the speed requirements of arc protection.
[0036] By selecting the sliding window width as described above, both the accuracy of steady-state determination and the response speed can be balanced.
[0037] Volatility is calculated using a normalized index that combines the average deviation magnitude and the extreme value span. The average deviation magnitude represents the average degree of deviation of the ratio data around the median ratio, while the extreme value span represents the maximum amplitude distribution range of the ratio subsequence within the window.
[0038] The preset stability threshold ranges from 0.03 to 0.15, and the preferred stability threshold in this application is 0.08.
[0039] The physical basis for selecting this threshold is that after the arc enters the stable combustion stage, the relative fluctuation of the ratio of the dual characteristic currents is usually maintained within 5%. A threshold below 0.03 is too stringent. In actual field, even steady-state arcs have slight fluctuations, making it difficult to trigger steady-state judgment. A threshold above 0.15 will misjudge the violent fluctuations in the arc initiation stage as steady state, resulting in the introduction of a large amount of transient data into the reference initialization and excessive deviation of the dynamic reference initial value.
[0040] The threshold setting of 0.08 provides a certain margin for on-site interference, which can adapt to the differences in steady-state fluctuations under different load conditions, while effectively blocking large fluctuations in arc-start transient data.
[0041] By setting the aforementioned stable threshold, the transient combustion and steady-state combustion stages of the electric arc can be accurately distinguished, providing a clean steady-state data range for subsequent baseline initialization.
[0042] The threshold for consecutive valid counts ranges from 3 to 8 times, and this application preferably sets the consecutive valid count to 5 times.
[0043] The engineering basis for selecting this parameter is that: when the number of consecutive tests is less than 3, the anti-interference capability is insufficient, and the low fluctuation rate caused by a single noise may trigger a false judgment; when the number of consecutive tests is more than 8, the steady-state judgment is severely lagging, and the effective data length of the compression stable combustion stage is reduced.
[0044] The time length corresponding to 5 consecutive effective windows is approximately 50 milliseconds, which can filter out most of the false steady-state signals caused by transient interference, while ensuring that the response speed of steady-state determination meets the protection requirements.
[0045] By setting the number of consecutive valid counts as described above, we can effectively suppress steady-state misjudgments caused by random disturbances and improve the reliability of stage division.
[0046] After the volatility calculation is completed, the volatility of each sliding time window is compared with the stability threshold. If it is less than the threshold, it is marked as a valid stability window. At the same time, a stability window counter is maintained. When a valid stability window is encountered, the count value is accumulated. When an invalid stability window is encountered, the count value is cleared. When the count value reaches the threshold for consecutive valid times, it is determined that the arc has entered the stable combustion stage and the start time is recorded. After entering the steady state, monitoring continues. When an invalid stability window occurs and causes the count value to be cleared, it is determined that the stable combustion stage has ended.
[0047] Through the aforementioned sliding window volatility continuous determination mechanism, the time interval of stable arc combustion can be automatically identified, and transient fluctuation data of arc initiation and extinction can be accurately isolated, providing high-quality steady-state ratio samples for dynamic benchmark initialization.
[0048] In one feasible implementation, standard deviation volatility is used instead of composite volatility. Specifically, the standard deviation of the ratio data within the sliding window is directly calculated and divided by the mean to obtain the coefficient of variation, which is then used as the volatility.
[0049] This scheme is suitable for conventional electric arc scenarios where the ratio data follows a normal distribution. Its advantage lies in the fact that the standard deviation is a classic statistic, the algorithm is mature, and the computational efficiency is relatively high. Its relative disadvantage is that for ratio sequences with outliers, the standard deviation is greatly affected by extreme values, the volatility index is easily distorted, and the robustness of steady-state determination is slightly lower than that of the composite volatility scheme.
[0050] In another feasible implementation, the range method volatility is used to determine steady state. Specifically, the volatility is determined by dividing the difference between the maximum and minimum values within the sliding window by the median value, ignoring the average deviation magnitude term.
[0051] This solution is suitable for low-end protection devices with extremely limited computing resources. Its advantage is that it only needs to record the extreme values within the window, without storing all the data within the window, reducing memory usage by about 60% and increasing the computing speed by about 40%. Its relative disadvantage is that it loses the intermediate layer information of data distribution, which is insufficient for distinguishing sequences with similar fluctuation amplitudes but different distribution patterns, and the precision of steady-state determination is reduced.
[0052] In another feasible implementation, an adaptive threshold method is used for steady-state determination. Specifically, the steady-state threshold is dynamically adjusted according to the overall amplitude level of the arc current. The larger the current amplitude, the more relaxed the steady-state threshold is, and the smaller the current amplitude, the more relaxed the steady-state threshold is.
[0053] This solution is suitable for wide-range arc detection scenarios, such as full-range power distribution protection devices covering 10 amps to 630 amps. Its advantage is that it can eliminate the influence of the ratio amplitude itself on the relative amount of fluctuation, and the steady-state judgment sensitivity remains consistent under different current levels. Its relative disadvantage is that the threshold adaptive logic increases the algorithm complexity, requires additional maintenance of current amplitude statistics, and the threshold mapping curve requires a large amount of experimental data for calibration.
[0054] Step 3: Extract the ratio sequence of the stable combustion stage and perform moving average processing to obtain the initial ratio benchmark sequence, and use the last value of the initial ratio benchmark sequence as the initial value of the dynamic ratio benchmark. In practice, after determining that the electric arc has entered the stable combustion stage, the dual characteristic current ratio sequence within this stage is extracted, and an initial ratio reference sequence is generated through moving average processing. The last value of the sequence is taken as the initial value of the dynamic ratio reference to complete the reference initialization.
[0055] In some embodiments, this application preferably uses the overlapping sliding window moving average method to generate the initial benchmark sequence.
[0056] The window width of the moving average window ranges from 3 to 12 ratio points, and the preferred window width in this application is 6 ratio points.
[0057] The physical basis for this parameter selection is as follows: when the window width is less than 3 ratio points, the smoothing effect is weak, the initial reference sequence still contains a lot of high-frequency fluctuations, and the stability of the initial reference value is insufficient; when the window width is greater than 12 ratio points, the smoothing is excessive, the initial reference cannot follow the slow change trend in the steady state stage, resulting in a systematic deviation between the initial reference value and the actual ratio at the end of the steady state.
[0058] The time span corresponding to the 6 ratio points is approximately 60 milliseconds, which can effectively filter out high-frequency random fluctuations in the steady-state ratio while retaining the slow variation trend in the steady-state stage, so that the initial benchmark value can accurately reflect the ratio level when the arc has just left the steady state.
[0059] By selecting the moving average window width as described above, an optimal balance can be achieved between baseline smoothness and trend following.
[0060] The sliding step size of the moving average ranges from one ratio point to half the window width, and the preferred step size in this application is two ratio points.
[0061] The engineering basis for this parameter selection is as follows: when the step size is 1, the initial reference sequence data density is the highest, but the overlap between adjacent windows reaches 83%, resulting in a large amount of redundant calculation and low computational efficiency; when the step size is greater than half the window width, the adjacent windows do not overlap, the time resolution of the reference sequence is too low, and it is impossible to accurately locate the ratio level at the end of the steady state.
[0062] A step size of 2 balances computational efficiency and time resolution, with the computational load being approximately 50% of that of point-by-point sliding. Meanwhile, the time interval of the reference sequence is kept within 20 milliseconds, which is sufficient to meet the accuracy requirements of the initial reference value.
[0063] By setting the sliding step size as described above, the amount of initialization calculation can be reduced by half while ensuring the accuracy of the reference.
[0064] After the moving average processing is completed, an initial ratio benchmark sequence is obtained in chronological order. The single-point benchmark value at the last position in the chronological order of this sequence is used as the initial value of the dynamic ratio benchmark.
[0065] The engineering basis for selecting the final value instead of the mean or initial value is that the arc state still has a slow evolution trend during the stable combustion stage. The final ratio is closest to the actual ratio level at the end of the steady state. Using the final value as the initial reference can ensure the continuity of the reference at the moment of switching between steady state and transient state, and avoid false offset caused by reference jump.
[0066] By employing the aforementioned initial value selection strategy, a smooth transition of the reference during the steady-state to transient transition phase can be achieved, reducing the reference shock at the moment of switching.
[0067] In one feasible implementation, a simple arithmetic mean method is used to calculate the initial baseline value. Specifically, the average of all ratio data during the entire stable combustion phase is directly calculated and used as the initial value of the dynamic ratio baseline, without generating a complete initial baseline sequence.
[0068] This scheme is suitable for scenarios with a long stable combustion phase and minimal fluctuations in the ratio within the steady state. Its advantage lies in the minimal computational load, requiring only one accumulation and division operation. However, its relative disadvantage is that it ignores the slow changing trend of the ratio within the steady state phase. When there is a systematic deviation between the initial and final steady-state ratios, the initial reference value error is large, which may lead to misjudgment in the initial direction determination.
[0069] In another feasible implementation, an exponentially weighted moving average method is used to generate the initial benchmark. Specifically, the ratio sequence in the steady state phase is subjected to exponentially decaying weights in reverse chronological order, and the initial benchmark value is obtained by weighted averaging.
[0070] This scheme is suitable for scenarios with a long stable combustion phase and a clear monotonic trend in the ratio, such as the slow development of a metallic arc. Its advantage is that the data weight is higher closer to the end of the steady state, and the baseline initial value is more accurate in representing the end state. Its relative disadvantage is that the weighting coefficient needs to be adaptively adjusted according to the steady state length, the algorithm implementation is relatively complex, and the rapid decay of the weights may lead to insufficient smoothing effect.
[0071] In another feasible implementation, the segmented averaging method is used to generate the initial reference sequence. Specifically, the stable combustion stage is divided into several segments, and the average value within each segment is used as the reference value for that segment, forming a stepped initial reference sequence. The average value of the last segment is then used as the initial value of the dynamic reference.
[0072] This scheme is suitable for special working conditions where the ratio changes segmentally during the steady-state phase, such as when the load side switches during the arc burning process. Its advantage is that segmented averaging can adapt to step-like ratio changes and avoids the distortion of the reference caused by averaging across step intervals. Its relative disadvantage is that the reference has step jumps at the segment boundaries, and the transition smoothness is not as good as continuous averaging with a sliding window.
[0073] Step 4: After the stable combustion stage ends, the dynamic ratio benchmark of the previous moment and the dual characteristic current ratio of the current moment are weighted and summed according to a preset forgetting ratio, and the summation result is used as the ratio change benchmark of the current moment to continuously update the dynamic ratio benchmark. In practice, after the stable combustion phase ends, the deviation of the dual characteristic current ratio is continuously monitored. A forgetting proportional weighting mechanism is used to update the dynamic ratio benchmark one by one, and an amplitude limiting constraint is introduced to prevent the benchmark from diverging, so as to achieve adaptive tracking of the benchmark.
[0074] In some embodiments, this application preferably employs a baseline update strategy that combines unidirectional continuous offset triggering with first-order inertial filtering.
[0075] The forgetting ratio ranges from 0.7 to 0.98, and the preferred forgetting ratio in this application is 0.9.
[0076] The physical basis for selecting this parameter is as follows: when the forgetting ratio is less than 0.7, the reference tracking speed is too fast, which is prone to frequent deviations due to instantaneous noise fluctuations, resulting in loss of reference stability. In some cases, the reference tracking direction offset may drift in the same direction, causing the offset to be canceled out. When the forgetting ratio is greater than 0.98, the reference tracking speed is too slow, which lags in responding to the slow ratio changes during the arc development process. The offset accumulation speed is too fast, which can easily lead to misjudgment.
[0077] A forgetting ratio of 0.9 corresponds to a time constant of approximately 10 sampling points, which ensures the benchmark's ability to track slowly changing trends while effectively suppressing instantaneous spike interference, achieving an optimal balance between the benchmark's stability and tracking performance.
[0078] By selecting the forgetting ratio as described above, the dynamic benchmark can achieve the best trade-off between smoothing noise and tracking trends.
[0079] The value of the direction continuation threshold ranges from 3 to 10 times, and the preferred direction continuation threshold in this application is 5 times.
[0080] The engineering basis for selecting this parameter is as follows: when the threshold is less than 3 times, the reference update is triggered too frequently, and a single reverse noise may interrupt the update rhythm, resulting in a decrease in reference tracking efficiency; when the threshold is greater than 10 times, the reference update response is severely lagging, and the reference update is triggered only after the offset has accumulated to a large extent, which is not conducive to the early triggering of direction determination.
[0081] Five consecutive offsets in the same direction take approximately 50 milliseconds, which effectively filters out noise interference from random polarity jumps, ensuring that the benchmark is updated only when a clear unidirectional trend is formed, while also ensuring the timeliness of benchmark updates.
[0082] By continuing the threshold setting in the aforementioned direction, selective updates of the benchmark can be achieved, invalid jitter updates can be avoided, and the overall stability of the benchmark can be improved.
[0083] Amplitude limiting protection is introduced during the benchmark update process. The upper limit threshold ranges from 120% to 200% of the initial benchmark value, and the preferred upper limit threshold in this application is 160% of the initial benchmark value. The lower limit threshold ranges from 20% to 80% of the initial benchmark value, and the preferred lower limit threshold in this application is 40% of the initial benchmark value.
[0084] The physical basis for selecting the limiting range is that during the normal development of an electric arc, the change in the ratio of the two characteristic currents usually does not exceed ±50% of the initial value. The upper limit is set at 160% and the lower limit at 40% to reserve sufficient margin for normal changes. If it exceeds this range, it is likely to be a sampling abnormality or an extreme fault condition. At this time, continuing to update the reference will lead to the reference going out of control. The limiting can constrain the reference within a reasonable range.
[0085] When the polarity of the baseline weight component and the ratio weight component are the same, they are added together; when their polarities are opposite, they are subtracted. If the result exceeds the limit range, the corresponding boundary value is output.
[0086] The above amplitude limiting mechanism can prevent the reference from drifting without limit under extreme conditions and ensure the robustness of the discrimination system.
[0087] The specific process of benchmark update is as follows: At the effective update time, the forgetting ratio is multiplied by the dynamic ratio benchmark of the previous time as the first scaling factor to obtain the benchmark weight component. The forgetting ratio is multiplied by the dual characteristic current ratio of the current time as the second scaling factor to obtain the ratio weight component. The two components are calculated according to the polarity rule and subjected to amplitude limiting to obtain the ratio change benchmark of the current time, which replaces the dynamic ratio benchmark of the previous time. At non-effective update times, the benchmark value of the previous time remains unchanged.
[0088] The aforementioned forgetting proportional weighted benchmark update mechanism enables the dynamic benchmark to slowly track the ratio baseline drift during the arc development process, while effectively suppressing the interference of instantaneous fluctuations and ensuring that the reference benchmark for offset calculation is always at a reasonable level.
[0089] In one feasible implementation, the baseline is updated at equal intervals. Specifically, the baseline is updated once every fixed number of sampling points without making a directional continuity judgment. Each update is weighted according to the forgetting ratio.
[0090] This scheme is suitable for scenarios where the arc ratio changes in a stable manner and there are few polarity jumps. Its advantage lies in its simplest logic, the absence of a direction counter maintenance requirement, and the minimal computational overhead. Its relative disadvantage is that when encountering instantaneous reverse spikes, the reference will be pulled in the opposite direction, causing short-term distortion of the reference. Its anti-interference capability is weaker than that of condition-triggered updates.
[0091] Another feasible implementation method is to adopt a variable forgetting ratio update strategy. Specifically, the forgetting ratio is dynamically adjusted according to the size of the offset. The larger the offset, the smaller the forgetting ratio and the faster the baseline tracking. The smaller the offset, the larger the forgetting ratio and the more stable the baseline.
[0092] This scheme is suitable for complex working conditions with a large range of arc development speed variations. Its advantage lies in its ability to adaptively adjust the reference response speed, quickly follow up when there is a large offset, and remain stable when there is a small fluctuation. Its overall performance is better than that of a fixed forgetting ratio. Its relative disadvantage is that the mapping function between the forgetting ratio and the offset requires a lot of experimental calibration, and the parameter tuning is difficult.
[0093] In another feasible implementation, a dual-benchmark parallel update mechanism is adopted. Specifically, two sets of benchmark values, a fast benchmark and a slow benchmark, are maintained simultaneously. The fast benchmark uses a lower forgetting ratio to achieve fast tracking, while the slow benchmark uses a higher forgetting ratio to maintain high stability. The actual offset is calculated by a weighted combination of the two.
[0094] This solution is suitable for high-end protection devices that require high sensitivity and stability in discrimination. Its advantage lies in balancing fast response and long-term stability, and it can adapt to both rapidly developing and slowly evolving electric arcs. Its relative disadvantage is that the memory usage and computational load are doubled, and the requirements for hardware resources are higher.
[0095] Step 5: Take the difference between the ratio of the dual characteristic currents at the current moment and the ratio change reference at the corresponding moment as the offset. When the offset maintains the same polarity in a continuous time period, determine the arc direction based on the polarity.
[0096] In practice, based on the continuous updating of the dynamic ratio benchmark, the offset of the current dual-path characteristic current ratio relative to the benchmark is calculated in real time. The offset trend is identified by maintaining continuous polarity, and the arc direction is finally output based on the polarity.
[0097] In some embodiments, this application preferably employs a polarity continuous counting mechanism for direction determination.
[0098] The value of the continuous polarity retention threshold ranges from 4 to 12 times, and this application preferably sets the continuous polarity retention threshold to 7 times.
[0099] The engineering basis for selecting this parameter is as follows: when the threshold is less than 4 times, the direction discrimination is too sensitive, and the brief polarity shift caused by instantaneous noise is prone to triggering misjudgment, and the arc direction output jumps frequently; when the threshold is greater than 12 times, the direction discrimination delay is too long, which cannot meet the speed requirement of arc protection, and the extended fault duration may cause more serious equipment damage.
[0100] The time for seven consecutive pulses of the same polarity is approximately 70 milliseconds, which can effectively filter out most interference pulses with random polarity reversals, while ensuring that direction determination is completed within 100 milliseconds, meeting the time requirements for arc fault protection in low-voltage power distribution systems.
[0101] By setting the continuous polarity retention threshold as described above, an optimal balance can be achieved between direction discrimination sensitivity and anti-interference capability.
[0102] The offset polarity identification rule is as follows: an offset greater than zero is marked as positive polarity, corresponding to the positive arc direction; an offset less than zero is marked as negative polarity, corresponding to the reverse arc direction; when the offset is equal to zero, the polarity mark of the previous moment is maintained, and no count change is triggered.
[0103] The polarity count update rule is as follows: compare the polarity marker at the current time with that at the previous time. If they match, the polarity count value is incremented by one. If they do not match, the count value is cleared to zero and a new polarity marker is recorded.
[0104] When the polarity count reaches the continuous polarity maintenance threshold, the offset is determined to form a stable unidirectional trend. The current polarity mark is read and the arc determination signal in the corresponding direction is output.
[0105] The above-mentioned continuous polarity determination mechanism ensures that the direction determination result is only output when the offset forms a clear and continuous unidirectional trend, effectively avoiding misjudgment of direction and frequent reversals caused by instantaneous fluctuations.
[0106] In one feasible implementation, the direction is determined by a weighted counting method based on the offset amplitude. Specifically, instead of incrementing by one each time the polarity is consistent, the count is weighted and accumulated based on the magnitude of the offset. The larger the offset, the larger the increment of the count value. Once the threshold is reached, the determination is triggered.
[0107] This scheme is suitable for scenarios where it is necessary to distinguish the severity of electric arcs. Its advantage is that large-amplitude offsets can trigger direction determination more quickly, enabling rapid response to severe faults. Small-amplitude offsets require more consecutive confirmations, thus balancing speed and reliability. The relative disadvantage is that the calibration of the weighting coefficients requires a large amount of experimental data, and the amplitude normalization process increases the number of calculation steps.
[0108] In another feasible implementation, a dual-threshold hierarchical judgment strategy is adopted. Specifically, two levels of consecutive counts are set: a warning threshold and an action threshold. When the warning threshold is reached, a warning signal is output, and when the action threshold is reached, a formal direction judgment signal is output.
[0109] This solution is applicable to power distribution systems with a hierarchical protection architecture. Its advantage lies in providing early warning information on arc development, allowing maintenance personnel time for emergency response, while maintaining high reliability in formal action determination. Its relative disadvantage is that it adds a level of state output, making the system state machine logic more complex, and the coordination of the two levels of thresholds needs to be precisely matched with the protection level difference.
[0110] In another feasible implementation, a sliding window polarity ratio determination method is adopted. Specifically, the ratio of positive to negative polarity within a fixed-length sliding window is statistically analyzed, and when the ratio of a certain polarity exceeds a preset threshold, it is determined to be the corresponding direction.
[0111] This scheme is suitable for complex arc scenarios with intermittent polarity reversals, such as intermittent arcs caused by poor contact. Its advantage is that it does not require strict continuity, and can be judged as long as the overall trend is dominant, and it has a higher tolerance for intermittent reversals. Its relative disadvantage is that the time window is fixed and the judgment delay is fixed as the window length, which cannot be triggered in advance when the trend is clear, like the continuous counting method, and the average response speed is slightly slower.
[0112] To fully verify the comprehensive performance of the arc direction discrimination method based on the ratio of dual characteristic currents described in this application, a standardized arc fault test platform was built, and systematic comparative experiments were carried out from five dimensions: discrimination accuracy, response speed, anti-interference ability, computational overhead, and stability. The experimental verification contents are as follows.
[0113] 1. Experimental Platform and Testing Conditions The main circuit of the experiment uses a programmable AC power supply as the power supply terminal. The power supply model is Chroma61703, with a rated output voltage range of 0 to 300 volts AC, a rated output current of 0 to 63 amps, a voltage regulation accuracy of 0.1 volts, and a current stability better than 0.2%. It can simulate various grid voltage fluctuation conditions.
[0114] The arc generating device adopts a standard copper-tungsten electrode arc generator with an electrode diameter of 6 mm and an electrode tip angle of 30 degrees. The electrode gap is adjusted by a precision stepper motor with an adjustment range of 0 to 5 mm and an adjustment accuracy of 0.01 mm, which can accurately control the arc ignition time and arc intensity.
[0115] The current acquisition unit uses two identical flexible current probes, model FLUKEi1000s, with a measurement range of 0 to 1000 amperes AC, a bandwidth of 1 kHz to 100 kHz, an amplitude error of less than 1%, and a phase error of less than 1 degree.
[0116] Two probes are connected in series on the power input side and the load output side of the arc fault point, respectively, corresponding to the positive current channel and the associated current channel. During installation, keep the probes in the same direction and mark their positions to ensure that the polarity of the two currents is defined uniformly.
[0117] The data acquisition card is model NIUSB-6363, which has 16 synchronous analog input channels, a sampling resolution of 16 bits, and a maximum single-channel sampling rate of 2 MHz. In this experiment, the two synchronous sampling rates are set to 25.6 kHz, and the synchronization deviation between channels is less than 50 nanoseconds.
[0118] The host computer is an industrial control computer with an Intel Core i7-10700 processor, a clock speed of 2.9 GHz, 16 gigabytes of memory, a 512 gigabyte solid-state drive, and runs a Windows 10 64-bit operating system.
[0119] The algorithm verification program is written in Python 3.9, numerical calculations use the NumPy library, and data visualization uses the Matplotlib library. All comparison algorithms run on the same hardware platform and software environment to ensure the fairness of the performance comparison.
[0120] The experimental environment temperature was controlled at 25±2 degrees Celsius, the relative humidity at 45%±5%, the grounding resistance of the experimental platform was less than 0.5 ohms, and there were no high-power equipment operating nearby to eliminate the influence of environmental electromagnetic interference on the experimental results.
[0121] Before the experiment, the system was calibrated: First, a standard 50 Hz, 20 amp sinusoidal current was applied to calibrate the amplitude error of the two current probes, so that the amplitude deviation between the two channels was less than 0.5%; second, the phase difference was calibrated by connecting the two probes in series in the same current loop, and the sampling delay was adjusted so that the phase deviation between the two channels was less than 0.1 degrees; finally, a step signal with a known amplitude was injected using a standard signal generator to verify the synchronization of the data acquisition card and ensure that the sampling time deviation between the two channels was less than 1 microsecond.
[0122] After calibration, the calibration parameters are saved, and all experimental data are corrected by the calibration parameters before algorithm processing.
[0123] 2. Test Sample Design The experiment was designed with four major categories of test samples, each of which had several subgroups, totaling 300 valid electric arc test samples. The sample size met the requirements for statistical significance testing.
[0124] The first category is the baseline arc sample, which includes two subgroups: forward arc and reverse arc. Each subgroup consists of 60 experiments, for a total of 120 subgroups.
[0125] The fault point of the forward arc sub-group is set on the load side, and the current flows from the power supply end through the arc to the load, simulating an arc fault caused by insulation failure at the load end; the fault point of the reverse arc sub-group is set on the power supply side, and the current flows from the load end through the arc to the power supply side, simulating an arc fault caused by loose wiring at the power supply end.
[0126] Each subgroup is divided into three levels according to the current level: 10 amps low current level, 30 amps medium current level, and 50 amps high current level. Each level is repeated 20 times to cover the common current operating range of low voltage power distribution systems.
[0127] The second category consists of 60 load characteristic samples, used to verify the impact of different load types on the discrimination effect.
[0128] It has three subgroups: the resistive load subgroup uses resistance wire load with a power factor of 0.98 or higher; the inductive load subgroup uses reactors in series with resistors with a power factor of about 0.5; and the nonlinear load subgroup uses a diode rectifier bridge with capacitor filtering with a total harmonic distortion of about 35%.
[0129] Each subgroup underwent 20 experiments, with the arc current uniformly set to 30 amperes, to compare the discrimination accuracy and stability under different load waveforms.
[0130] The third category consists of 60 electromagnetic interference samples, used to simulate the discrimination performance under complex electromagnetic environments in the field.
[0131] The system consists of three interference subgroups: a random noise interference subgroup that superimposes Gaussian white noise with an amplitude of 10% of the fundamental frequency into the current sampling signal; a power frequency harmonic interference subgroup that superimposes the 3rd, 5th, and 7th harmonics, with a total harmonic content of 15%; and a pulse interference subgroup that injects a narrow pulse interference with an amplitude of 30% of the fundamental frequency and a width of 10 microseconds every 100 milliseconds to simulate the electrical fast transient pulse generated by the contactor opening and closing.
[0132] Each subgroup conducted 20 experiments, with a baseline arc current of 30 amperes, and half of the arcs were forward arcs and half were reverse arcs.
[0133] The fourth category consists of 60 extreme condition samples, used to verify the algorithm's boundary performance.
[0134] It has three subgroups: the weak arc subgroup has an arc current of only 5 amperes and a weak arc characteristic; the strong arc subgroup has an arc current of up to 60 amperes, close to the rated upper limit of the device; the intermittent arc subgroup has electrodes that vibrate periodically, and the arc is intermittent, extinguishing and reigniting once every 100 milliseconds.
[0135] Each subgroup has 20 experiments, half in the forward direction and half in the reverse direction, to assess the algorithm's adaptability under extreme conditions.
[0136] The comparison algorithm selected three mainstream schemes as control groups: Scheme 1 is the traditional fixed benchmark ratio method, which directly uses the fixed ratio calibrated at the factory as the benchmark. The ratio is judged as positive if it exceeds the positive threshold and as negative if it is below the negative threshold. There is no steady-state division and dynamic update; Scheme 2 is the moving average benchmark method, which uses the mean of a fixed-length sliding window as the dynamic benchmark. There is no steady-state stage distinction and the benchmark is updated point by point; Scheme 3 is the method proposed in this application, which includes a complete process of steady-state stage identification, forgetting ratio weighted benchmark update, and continuous polarity determination.
[0137] All three algorithms use the same raw sampled data for offline processing to ensure that the input conditions are completely consistent.
[0138] 3. Experimental Procedure The first step is to build the experimental circuit and complete the system calibration.
[0139] Connect the main circuit in series in the order of power supply-probe 1-arc generator-probe 2-load. Confirm that the polarity of the two current probes is consistent. Connect the data acquisition card to the host computer and run the calibration program to complete the calibration of amplitude, phase and synchronization. Record the calibration coefficients.
[0140] The second step is to set the experimental operating parameters.
[0141] According to the requirements of the current test sample group, adjust the AC power supply output current to the set value, set the initial electrode gap of the arc generator, select the load type and interference injection mode, configure the parameters of three comparison algorithms in the host computer software, set the sampling duration to 5 seconds, and the pre-trigger sampling length to 1 second.
[0142] The third step is to trigger the arc and collect data.
[0143] Start the data acquisition card to enter the pre-trigger waiting state, slowly adjust the electrode gap to reduce until the arc is ignited, and after the arc burns stably for about 2 seconds, gradually increase the electrode gap to extinguish the arc. Acquire the complete waveform of the entire process of arc ignition-burning-arc extinguishing, automatically save the original current data file and label it with experimental conditions.
[0144] The fourth step is offline algorithm processing and metric statistics.
[0145] For each set of raw data, three comparison algorithms are input for processing, and the following indicators are automatically recorded: whether the direction determination result is correct, the time of the first output of the direction determination, the number of polarity reversals during the determination process, the time consumed by the algorithm for a single set of data, whether the steady-state stage identification is successful, and the number of baseline updates.
[0146] The metrics are automatically written into the results statistics table after each experiment is completed.
[0147] The fifth step is to repeat the experiment and summarize the data.
[0148] Complete all 300 sets of experiments according to the sample design. Before each set of experiments, check the electrode burn-off condition. Replace the electrode when the burn-off depth exceeds 0.5 mm to ensure the consistency of the conditions for arc generation.
[0149] After all experiments are completed, the average, maximum, minimum and standard deviation of each indicator are calculated according to the sample groups, and a summary data table is generated.
[0150] Step 6: Conduct supplementary verification experiments.
[0151] Typical working conditions were selected for repeatability verification. Ten experiments were randomly selected from each group for repeated testing to verify the reproducibility of the experimental results. At the same time, parameter sensitivity tests were conducted. The three core parameters, namely the stability threshold, forgetting ratio, and continuous polarity threshold, were adjusted respectively, and the accuracy change curves were observed to verify the rationality and robustness range of the parameter selection.
[0152] As can be seen from the above embodiments, the arc direction discrimination method based on the ratio of dual characteristic currents provided by the present invention accurately identifies the stable combustion stage of the arc by calculating the fluctuation rate of the ratio of dual characteristic currents through a sliding time window. It completes dynamic benchmark initialization based on the stable ratio of this stage, and then achieves benchmark adaptive update by combining forgetting ratio weighting and adding amplitude limiting. Finally, it completes the arc direction determination based on the continuous and unified polarity of the offset. It can effectively isolate the interference caused by transient current fluctuations during arc initiation and arc extinction, reduce the polarity jump of the offset, and significantly reduce the probability of arc direction discrimination error. At the same time, the entire set of hierarchical judgment logic simplifies invalid calculations and significantly improves the overall calculation efficiency and recognition stability of arc direction discrimination.
[0153] like Figure 2 The diagram shown is a functional block diagram of an arc direction discrimination system 100 based on the ratio of dual characteristic currents provided in an embodiment of the present invention, including a characteristic ratio calculation block 101, a steady-state discrimination module 102, a reference initialization module 103, a reference update module 104, and a direction determination module 105.
[0154] In this embodiment, the functions of each module are as follows: The feature ratio calculation module 101 is used to take the ratio of the peak change of the time-synchronized positive current feature and the associated current feature in the same arc event as the dual-path feature current ratio. The steady-state discrimination module 102 is used to calculate the volatility of the dual-path characteristic current ratio sequence within a sliding time window. When the volatility of multiple consecutive time windows is less than a preset stability threshold, the arc is determined to have entered a stable combustion stage. The benchmark initialization module 103 is used to extract the ratio sequence of the stable combustion stage, perform moving average processing to obtain an initial ratio benchmark sequence, and use the last value of the initial ratio benchmark sequence as the initial value of the dynamic ratio benchmark. The reference update module 104 is used to, after the stable combustion stage ends, sum the dynamic ratio reference of the previous moment and the dual characteristic current ratio of the current moment according to a preset forgetting ratio, and use the summation result as the ratio change reference of the current moment, so as to continuously update the dynamic ratio reference. The direction determination module 105 is used to take the difference between the ratio of the dual characteristic currents at the current time and the ratio change reference at the corresponding time as the offset. When the offset maintains the same polarity in a continuous time period, the direction of the electric arc is determined according to the polarity.
[0155] As can be seen from the above embodiments, the arc direction discrimination system based on the ratio of dual characteristic currents provided by the present invention accurately identifies the stable combustion stage of the arc by calculating the fluctuation rate of the ratio of dual characteristic currents through a sliding time window. It completes dynamic benchmark initialization based on the stable ratio of this stage, and then achieves benchmark adaptive update by combining forgetting ratio weighting and adding amplitude limiting. Finally, it completes the arc direction determination based on the continuous and unified polarity of the offset. It can effectively isolate the interference caused by transient current fluctuations during arc initiation and arc extinction, reduce the polarity jump of the offset, and significantly reduce the probability of arc direction discrimination error. At the same time, the entire set of hierarchical judgment logic simplifies invalid calculations and significantly improves the overall calculation efficiency and recognition stability of arc direction discrimination.
[0156] Figure 3This chart presents the overall accuracy distribution data for the four sample categories, summarizing the accuracy distribution of the four test sample categories (reference arc, load characteristics, electromagnetic interference, and extreme operating conditions) in the patent experiment. It is used to compare the overall performance differences between the proposed solution and two existing methods (fixed reference ratio method and moving average reference method). The horizontal axis represents the four sample categories, and the vertical axis represents the accuracy rate (percentage). For each sample category, box plots or bar charts are used to display the accuracy distribution of the three algorithms. Data shows that the accuracy of the proposed solution in the reference arc sample (120 groups in total, forward and reverse) exceeds 98%, significantly higher than the fixed reference method (approximately 82%) and the moving average reference method (approximately 89%). This indicates that steady-state phase identification and dynamic reference initialization can accurately establish a reference benchmark, avoiding transient interference during arc initiation and extinction. In the load characteristic sample (resistive, inductive, and nonlinear loads), the accuracy of the proposed solution remains around 96%, while the accuracy of the comparative solutions drops to 75% and 81% respectively due to increased ratio fluctuations caused by waveform distortion. This confirms that the forgetting ratio weighted benchmark is more effective. The invention demonstrates superior adaptability to various waveform distortions. In electromagnetic interference samples (random noise, harmonics, and pulse interference), the accuracy rate is approximately 94%, thanks to the continuous polarity determination mechanism effectively filtering out instantaneous polarity jumps. In contrast, the accuracy rate of the fixed reference method is only 68% due to interference, and the moving average reference method is approximately 73%, indicating a significant anti-interference advantage. In extreme operating condition samples (weak arc, strong arc, and intermittent arc), the accuracy rate is approximately 91%, still higher than the 60% and 70% of the comparative schemes, indicating that the amplitude limiting protection and polarity counting mechanism can maintain reliable discrimination under boundary conditions. Overall, the average accuracy rate of the four types of samples is approximately 94.8%, an improvement of 18.5 percentage points compared to the fixed reference method (approximately 76.3%) and 11.5 percentage points compared to the moving average reference method (approximately 83.3%), fully demonstrating the superiority and stability of the invention under diverse operating conditions.
[0157] Figure 4The graph showing the change in accuracy of the stationary threshold illustrates the influence of a preset stationary threshold (ranging from 0.03 to 0.15) on the accuracy of arc direction discrimination in the sliding time window volatility determination, used to verify the rationality of the patented preferred threshold of 0.08. The horizontal axis represents the stationary threshold value, and the vertical axis represents the overall discrimination accuracy (%) of all test samples at the corresponding threshold. The curve shows that as the threshold gradually increases from 0.03 to 0.08, the accuracy shows a monotonically increasing trend. At the initial value of 0.03, the accuracy is only about 78%, because the threshold is too strict and a large number of steady-state arcs cannot be effectively identified. The baseline initialization stage fails to capture the stationary ratio sufficiently, the initial value of the dynamic baseline has a large deviation, and the direction offset calculation is distorted. When the threshold is raised to 0.08, the accuracy reaches a peak of about 94.8%. At this point, the recognition sensitivity and false rejection rate in the steady-state stage are in optimal balance. It can effectively exclude transient fluctuations in arc initiation and fully capture the stable combustion range. When the threshold continues to increase to 0.15, the accuracy gradually drops back to about 86%, because the threshold is too lenient and some violent fluctuations in the arc initiation stage are misjudged as steady state. The baseline initialization introduces a large amount of non-stationary data, the dynamic baseline deviates from the true steady-state level, and the polarity of the subsequent direction discrimination frequently jumps. Figure 4 The sub-curves under different load types were also marked, showing that the optimal threshold shifts slightly (about 0.09~0.10) under nonlinear loads and extreme conditions, but 0.08 is still the globally optimal compromise value, verifying the engineering universality of the patented parameter selection.
[0158] Figure 5 This graph shows the change in accuracy due to the forgetting rate, reflecting the impact of the forgetting rate (ranging from 0.7 to 0.98) on the overall discrimination accuracy during the dynamic ratio benchmark update process. It aims to verify the effectiveness of the preferred patent value of 0.9. The horizontal axis represents the forgetting rate coefficient, and the vertical axis represents the overall discrimination accuracy (%) at the corresponding rate. The curve exhibits a convex trend of first rising and then falling: when the forgetting ratio is low (e.g., 0.7), the accuracy is about 81% because the benchmark update is too sensitive, closely following instantaneous ratio fluctuations or even noise, causing the benchmark itself to drift in the same direction, the offset is partially canceled out, and the directional polarity is difficult to form a continuous trend; as the forgetting ratio increases to 0.9, the accuracy rises to a peak of about 94.8%, at which point the first-order inertial filter time constant is moderate, the benchmark can both smoothly suppress high-frequency random noise and slowly track the baseline drift during the arc development process, and the reliability of the offset maintaining continuous same polarity is the highest; when the forgetting ratio further increases to above 0.95, the accuracy drops to about 88% because the benchmark update response is lagging and cannot follow the slow unidirectional change of the ratio in the later stage of arc combustion, the offset accumulates too quickly, causing the direction determination to be triggered prematurely, especially in nonlinear load and intermittent arc scenarios, the misjudgment rate increases. Figure 5The figure also shows individual curves for different current levels marked with dashed lines. It shows that the optimal forgetting ratio is slightly lower (about 0.88) under high current (50A) conditions and slightly higher (about 0.93) under low current (10A) conditions. However, 0.9 maintains an accuracy of over 92% under all conditions, which verifies that a fixed forgetting ratio of 0.9 has strong adaptability to operating conditions and engineering simplicity.
[0159] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.
[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims
1. An arc direction determination method based on the ratio of dual characteristic currents, characterized in that, The method includes: Step 1: Take the ratio of the peak change of the time-synchronized positive current characteristic to the associated current characteristic in the same arc event as the dual-path characteristic current ratio. Step 2: Calculate the volatility of the dual-path characteristic current ratio sequence within the sliding time window. When the volatility of multiple consecutive time windows is less than the preset stability threshold, it is determined that the electric arc has entered the stable combustion stage. Step 3: Extract the ratio sequence of the stable combustion stage and perform moving average processing to obtain the initial ratio benchmark sequence, and use the last value of the initial ratio benchmark sequence as the initial value of the dynamic ratio benchmark. Step 4: After the stable combustion stage ends, the dynamic ratio benchmark of the previous moment and the dual characteristic current ratio of the current moment are weighted and summed according to a preset forgetting ratio, and the summation result is used as the ratio change benchmark of the current moment to continuously update the dynamic ratio benchmark. Step 5: Take the difference between the ratio of the dual characteristic currents at the current moment and the ratio change reference at the corresponding moment as the offset. When the offset maintains the same polarity in a continuous time period, determine the arc direction based on the polarity.
2. The arc direction determination method based on the ratio of dual characteristic currents as described in claim 1, characterized in that, The process of using the ratio of the peak change of the time-synchronized positive current characteristic to the associated current characteristic in the same arc event as the dual-path characteristic current ratio is as follows: The forward current signal and the associated current signal are sampled synchronously in time to obtain the forward half-wave sequence and the associated half-wave sequence under the same time base; Disturbance extreme points in the positive half-wave sequence and the associated half-wave sequence that have amplitude jumps exceeding a preset tolerance threshold are removed respectively, and the remaining extreme points are taken as valid peak points; The amplitude difference between adjacent effective peak points within each half-wave is taken as the peak value change of the half-wave, and the positive peak value change and the associated peak value change are obtained accordingly. The ratio of the positive peak change to the associated peak change under the same half-wave is used as the dual-path characteristic current ratio.
3. The arc direction determination method based on the ratio of dual characteristic currents as described in claim 1, characterized in that, The process for calculating the volatility of the dual-path characteristic current ratio sequence within the sliding time window is as follows: By sliding a fixed-length window across the ratio sequence point by point, a subsequence of ratios at each sliding position is obtained; Arrange all ratio data in the ratio subsequence in ascending order of magnitude, take the ratio at the middle position as the median ratio, and take the ratio at the first and last positions as the minimum ratio and the maximum ratio, respectively. The average deviation is obtained by summing the absolute deviations of the ratio data and the median ratio and dividing by the total number of the ratio data. The difference between the maximum ratio and the minimum ratio is taken as the extreme value span. Calculate the volatility within the sliding time window, wherein the volatility is calculated using the following formula: ; in, This represents the volatility; a smaller value indicates a more stable subsequence. The average deviation magnitude represents the average degree of deviation of the ratio data from the median ratio. The extreme value span represents the range of maximum amplitude distribution of the ratio subsequence within the window. The median ratio is represented as an order of magnitude adjustment factor in the normalization benchmark. The volatility is compared with a preset stability threshold, and the sliding time window is marked as a valid stability window when the volatility is less than the preset stability threshold.
4. The arc direction determination method based on the ratio of dual characteristic currents as described in claim 3, characterized in that, When the fluctuation rate of multiple consecutive sliding time windows is less than the preset stability threshold, the electric arc is determined to have entered the stable combustion stage. The process is as follows: After each sliding time window is marked, if the current sliding time window is marked as a valid smooth window, the smooth window counter is incremented by one count value; otherwise, the smooth window counter is cleared to zero. The current count value of the stable window counter is read in real time. When the current count value reaches the threshold of consecutive valid times, it is determined that the electric arc has entered the stable combustion stage, and the current time is recorded as the start time of the stable combustion stage. After determining that the stable combustion stage has been entered, the sliding time window continues to be marked and updated. When an ineffective stable window occurs, causing the count value to be cleared to zero, the stable combustion stage is determined to end.
5. The arc direction determination method based on the ratio of dual characteristic currents as described in claim 1, characterized in that, The process of extracting the ratio sequence of the stable combustion stage and performing moving average processing to obtain an initial ratio benchmark sequence, and using the last value of the initial ratio benchmark sequence as the initial value of the dynamic ratio benchmark, is as follows: Starting from the beginning of the stable combustion stage, the dual-path characteristic current ratio sequence is successively extracted at fixed step sizes to obtain multiple local window sequences with time overlap. The ratio data of the local window sequence are accumulated sequentially in time order to obtain the cumulative total value of the local window sequence; Divide the accumulated total value by the number of data in the local window sequence to obtain the single-point benchmark value corresponding to the local window sequence, and arrange the single-point benchmark values corresponding to all local window sequences in chronological order to form the initial ratio benchmark sequence. The single-point reference value that is the last in the time sequence of the initial ratio reference sequence is used as the initial value of the dynamic ratio reference.
6. The arc direction determination method based on the ratio of dual characteristic currents as described in claim 1, characterized in that, After the stable combustion phase is completed, the process is as follows: After the stable combustion phase ends, the deviation of the current dual-path characteristic current ratio from the previous dynamic ratio reference is continuously monitored, and a deviation direction sequence is generated. The number of consecutive occurrences of the same direction in the deviation direction sequence is counted. When the number reaches a preset direction continuation threshold, it is determined that the ratio of the dual characteristic currents forms a unidirectional continuous offset trend, and the current moment is determined as the effective update moment. The dynamic ratio benchmark is updated at the sampling point at the effective update time, and the dynamic ratio benchmark at the previous time is maintained unchanged at the sampling point that is not determined to be the effective update time.
7. The arc direction determination method based on the ratio of dual characteristic currents as described in claim 6, characterized in that, After the stable combustion phase ends, the dynamic ratio benchmark from the previous moment and the dual-path characteristic current ratio from the current moment are weighted and summed according to a preset forgetting ratio, and the sum is used as the ratio change benchmark for the current moment to continuously update the dynamic ratio benchmark. The process is as follows: At the effective update time, the preset forgetting ratio is used as the first scaling factor and multiplied by the dynamic ratio benchmark of the previous time to obtain the benchmark weight component. The difference between the forgetting ratio and the value one is used as the second scaling factor, which is multiplied by the ratio of the dual-path characteristic current at the current time to obtain the ratio weight component. The benchmark weight component and the ratio weight component are superimposed to obtain the ratio change benchmark at the current moment, and the ratio change benchmark replaces the dynamic ratio benchmark at the previous moment to complete the continuous update of the dynamic ratio benchmark.
8. The arc direction determination method based on the ratio of dual characteristic currents as described in claim 7, characterized in that, The process of superimposing the benchmark weight component with the ratio weight component to obtain the benchmark for the ratio change at the current moment is as follows: Check the polarity of the baseline weight component and the ratio weight component. If the polarities are the same, perform an addition operation; if the polarities are opposite, perform a subtraction operation to obtain the initial calculation result. The initial calculation result is compared with the predetermined upper limit threshold and lower limit threshold of amplitude. If the initial calculation result is greater than the upper limit threshold of amplitude, the upper limit threshold of amplitude is output as the amplitude limiting result. If the initial calculation result is less than the lower amplitude threshold, then the lower amplitude threshold is output as the amplitude limiting result; If the initial calculation result is between the lower amplitude threshold and the upper amplitude threshold, the initial calculation result is directly output as the amplitude limiting result, and the amplitude limiting result is used as the ratio change benchmark at the current moment.
9. The arc direction determination method based on the ratio of dual characteristic currents as described in claim 1, characterized in that, The difference between the ratio of the dual characteristic currents at the current moment and the ratio change reference at the corresponding moment is used as the offset. When the offset maintains the same polarity within a continuous time period, the arc direction is determined based on the polarity. The process is as follows: The difference between the current ratio of the dual characteristic currents and the ratio change reference is used as the offset, and the polarity of the offset is identified. If the offset is greater than zero, it is marked as positive polarity; if the offset is less than zero, it is marked as negative polarity. Compare the polarity markers at the current time with those at the previous time. If the polarity markers match, the polarity continuity count is incremented. If the polarity markers do not match, the polarity continuity count is cleared and a new polarity marker is recorded. When the polarity count value reaches the continuous polarity maintenance threshold, it is determined that the offset maintains the same polarity within a continuous time period. The polarity mark at the current moment is read. If the polarity mark is positive, a positive arc direction determination signal is output. If the polarity mark is negative, a reverse arc direction determination signal is output.
10. An arc direction determination system based on the ratio of dual characteristic currents, characterized in that, The system is used to implement the arc direction determination method based on the ratio of dual characteristic currents as described in any one of claims 1-9, the system comprising: The characteristic ratio calculation module is used to take the ratio of the peak change of the time-synchronized positive current characteristic and the associated current characteristic in the same arc event as the dual-path characteristic current ratio. The steady-state discrimination module is used to calculate the volatility of the dual-path characteristic current ratio sequence within a sliding time window. When the volatility of multiple consecutive time windows is less than a preset stability threshold, the arc is determined to have entered a stable combustion stage. The baseline initialization module is used to extract the ratio sequence of the stable combustion stage, perform moving average processing to obtain an initial ratio baseline sequence, and use the last value of the initial ratio baseline sequence as the initial value of the dynamic ratio baseline. The benchmark update module is used to continuously update the dynamic ratio benchmark after the stable combustion stage ends, by weighting and summing the dynamic ratio benchmark of the previous moment and the dual characteristic current ratio of the current moment according to a preset forgetting ratio, and using the summing result as the ratio change benchmark of the current moment. The direction determination module is used to take the difference between the ratio of the dual characteristic currents at the current time and the ratio change reference at the corresponding time as the offset. When the offset maintains the same polarity in a continuous time period, the direction of the electric arc is determined according to the polarity.
