A Multi-parameter Fusion Analysis Protection Method and System Based on Load Characteristic Identification

Abstract

This invention belongs to the technical field of emergency protection circuit devices, specifically relating to a multi-parameter fusion analysis protection method and system based on load characteristic identification. The method includes: dynamically shortening the analysis window based on the current change rate after responding to an overcurrent; performing energy envelope morphology matching and fundamental phase verification in parallel within the window; morphology matching confirms normal operation or marks abnormal morphology by comparing a self-normalized sequence with a template; phase verification uses first-order trend prediction residuals to determine phase abnormalities if they continuously exceed limits; combining the results of both, if normal operation is confirmed or the similarity exceeds the enhancement threshold, the circuit breaker does not trip; otherwise, it trips; and if the current still exceeds the limit within the maximum allowable disconnection time, it forces a trip. This invention reduces the impact of power level through self-normalized envelope morphology matching, sensitively detects hidden faults using phase trend prediction, dynamically adjusts the analysis window, and fuses multiple parameters to improve the reliability of protection.

Description

Technical Field

[0001] This invention relates to the field of emergency protection circuit devices. More specifically, this invention relates to a multi-parameter fusion analysis protection method and system based on load characteristic identification. Background Technology

[0002] In low-voltage power distribution systems containing a large number of nonlinear loads such as LED lighting and variable frequency air conditioners, the high harmonic current generated when the load starts can easily cause traditional overcurrent protection devices based on current amplitude or effective value to frequently malfunction.

[0003] To improve selectivity, some existing solutions identify faults by extracting single transient features, such as matching methods based on transient energy envelope morphology or detection methods based on fundamental phase difference.

[0004] However, transient energy envelope shape matching heavily relies on the completeness of the pre-set template library. When the template coverage is insufficient or the load characteristics drift, the matching degree decreases, which may cause protection delay or even failure to operate. At the same time, directly performing shape matching does not eliminate the influence of load power level differences on envelope amplitude, thus limiting the accuracy of identification.

[0005] Detection based on fundamental phase difference is sensitive to transient phase shifts caused by normal power grid operations such as transformer inrush current, which can easily lead to false tripping. Furthermore, conventional phase comparison lacks a mechanism for predicting phase change trends and confirming continuous damage, making it difficult to eliminate random interference.

[0006] In addition, the current increment of low-amplitude fault arcs is small, and its energy injection process is highly similar in shape to that of a resistive load under normal operation. The above-mentioned single characteristic method is difficult to distinguish effectively, and there is a high risk of missed detection.

[0007] Existing discrimination mechanisms mostly use fixed time windows, which cannot dynamically adjust the decision duration according to the rise rate of the fault current. In the case of severe short circuits, the fault may not be cleared in time, while in the case of slow change, the fault may be misjudged due to insufficient observation.

[0008] Therefore, existing protection methods struggle to balance speed and selectivity under complex load conditions. How to provide a protection scheme that can integrate multi-dimensional transient characteristics, adaptively adjust the analysis window, and achieve robust decision-making has become a pressing technical problem in this field. Summary of the Invention

[0009] To address the aforementioned technical problems, the present invention provides solutions in the following aspects.

[0010] In a first aspect, the present invention provides a multi-parameter fusion analysis protection method based on load characteristic identification, employing the following technical solution: The multi-parameter fusion analysis protection method based on load characteristic identification includes: in response to overcurrent, dynamically shortening the analysis window based on the current change rate; wherein, the current change rate... Analysis window duration and number of sampling points in the window Determined according to the following formula:

[0011]

[0012]

[0013] In the formula, For the first time after overcurrent triggering The instantaneous current value at each current sampling point The sampling time interval, The number of sampling points used to calculate the rate of change of current. The preset maximum reference value for the rate of change of current. This is the preset base window length;

[0014] Energy envelope morphology matching and fundamental phase verification are performed in parallel within the window;

[0015] Morphological matching is performed by accumulating instantaneous power values ​​from the trigger point. During the windowing process, the Pearson correlation is directly calculated between the transient energy accumulation sequence prefix obtained up to the current sampling point and the corresponding equal-length prefix of the template. This direct Pearson correlation calculation is algebraically equivalent to dividing each element in the transient energy accumulation sequence prefix by the sum of the cumulative energy at the end of the same window and a preset positive constant before performing the Pearson correlation calculation. After the window ends and the cumulative energy at the end of the window is obtained, the complete transient energy accumulation sequence is self-normalized. If the morphological matching degree reaches the first threshold, normal load input is confirmed; if it remains below the second threshold, morphological abnormality is marked. Phase verification performs first-order trend prediction on the fundamental phase difference; if the prediction residual continuously exceeds the limit, phase abnormality is judged.

[0016] If, at the end of the window, it is confirmed that normal load is being supplied and there is no phase abnormality, or if the similarity is higher than the enhanced threshold which is stricter than the first threshold, then the circuit breaker will not trip.

[0017] Otherwise, if there is a phase anomaly and the morphological matching degree is not higher than the enhancement threshold, or if there is a morphological anomaly, the circuit breaker will trip.

[0018] It will force a trip when the preset maximum allowable disconnection time is reached and the current still exceeds the limit.

[0019] Furthermore, the step of dynamically shortening the analysis window based on the current change rate includes: obtaining the absolute value of the current change rate within a preset time after triggering; obtaining a window shortening coefficient based on the absolute value of the current change rate through a preset linear mapping function; and calculating the product of the window shortening coefficient and a preset basic window length as the duration of the analysis window.

[0020] Furthermore, the self-normalization includes: starting from the overcurrent triggering moment, multiplying the product of the instantaneous voltage value and the instantaneous current value at each sampling point by the sampling interval and accumulating the results to obtain a transient energy accumulation sequence; during the analysis window, for the prefix formed by the currently obtained transient energy accumulation sequence, directly performing Pearson correlation calculation with the corresponding equal-length prefix of the preset load template; after the analysis window ends and the accumulated energy at the end of the window is obtained, dividing each element in the complete transient energy accumulation sequence by the sum of the accumulated energy at the end of the window and a preset positive constant to obtain a normalized energy envelope pattern sequence.

[0021] Furthermore, the first threshold and the second threshold are preset parameters.

[0022] Furthermore, the similarity comparison with the template includes: calculating the correlation between the energy envelope pattern sequence obtained by the instantaneous value of the accumulated power and the self-normalization process and the multiple preset load templates one by one to obtain the correlation coefficient corresponding to each template; and taking the maximum value among all correlation coefficients as the similarity.

[0023] Furthermore, the first-order trend prediction includes: acquiring the measured phase difference between the fundamental voltage and fundamental current in the current cycle; calculating the predicted phase difference value for the next cycle based on the measured phase difference, a preset trend inertia coefficient, and the statistical average of historical steady-state phase differences; calculating the prediction residual based on the predicted phase difference value and the measured phase difference for the next cycle; and determining a phase anomaly if the number of times the prediction residual exceeds a preset phase disruption threshold reaches a preset confirmation number.

[0024] Furthermore, the historical steady-state phase difference statistical mean is obtained in the following way: when the prediction residuals of multiple consecutive cycles do not exceed the limit and the current is in a stable range, the measured phase differences of the corresponding cycles are arithmetically averaged and updated to the historical steady-state phase difference statistical mean.

[0025] Furthermore, the application of the enhancement threshold includes: when the fundamental phase verification determines a phase abnormality, if the similarity obtained by the energy envelope morphology matching is higher than the enhancement threshold, then the circuit breaker will not trip; if the similarity is lower than or equal to the enhancement threshold, then it will be determined as a fault event and the circuit breaker will trip.

[0026] Furthermore, the preset maximum allowable cut-off time and the current still exceed the preset parameter value.

[0027] Secondly, the present invention provides a multi-parameter fusion analysis and protection system based on load characteristic identification, which adopts the following technical solution: the multi-parameter fusion analysis and protection system based on load characteristic identification includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned multi-parameter fusion analysis and protection method based on load characteristic identification is implemented.

[0028] The embodiments of the present invention have at least the following beneficial effects:

[0029] 1. Based on the self-normalized transient energy envelope shape matching, the influence of load power level and fault distance on energy amplitude is effectively reduced. The correlation between the time distribution shape of energy injection and the preset template is used only to accurately identify the normal load input process, thereby avoiding protection maloperation caused by high harmonic current of nonlinear load start-up and improving the selectivity of protection.

[0030] 2. The phase difference between fundamental voltage and current is predicted by a first-order trend prediction model, and the continuous phase destruction is determined by the over-limit of the prediction residuals after multiple consecutive predictions. This can sensitively capture the random phase jump caused by hidden faults such as low-amplitude fault arcs, and make up for the inadequacy of pure energy envelope matching in detecting minor faults. When the phase is abnormal but the energy envelope matching degree is high, enhanced threshold cross-validation is introduced to prevent false tripping of special normal loads, thus balancing detection sensitivity and protection reliability.

[0031] 3. An analysis window that dynamically shortens with the rate of change of fault current is adopted, allowing for rapid output during severe short circuits and thorough observation during smooth transients, achieving a self-consistent balance between speed and selectivity. Within the window, energy envelope matching and phase verification are executed in parallel along two paths, and multiple levels of evidence, such as normal operation confirmation, morphological anomalies, phase destruction, and forced tripping at the maximum allowable clearing time, are integrated according to priority. By combining the complementary characteristics of transient energy behavior and phase consistency, the limitations of single-feature decision-making are overcome, and the speed, selectivity, and reliability of the protection are synergistically improved. Attached Figure Description

[0032] Figure 1 The flowchart illustrates the steps of the multi-parameter fusion analysis protection method based on load characteristic identification in this invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] The following is in conjunction with the appendix Figure 1 The specific embodiments of the present invention will be described in detail below.

[0035] S1: In response to overcurrent, the analysis window is dynamically shortened based on the rate of change of current.

[0036] The synchronous sampling module within the terminal uses sampling time intervals. Real-time acquisition of instantaneous voltage values The units are volts (V) and instantaneous current. The unit is amperes (A), and the most recently stored... One sampled data.

[0037] Continuous monitoring of the effective value of the sliding window of the current The unit is ampere (A).

[0038] when Exceeding the preset initial overcurrent trigger threshold When an overcurrent occurs, the trigger time is recorded. and the corresponding sampling index Then it enters the analysis state.

[0039] The method for real-time calculation of the effective value of the sliding window is a well-known technique and will not be elaborated further.

[0040] Initial overcurrent trigger threshold Get experience points , This is the effective value of the line's rated current, expressed in amperes (A). This threshold can also be adjusted by the implementer based on the statistical value of the maximum normal inrush current on site.

[0041] After triggering the analysis state, a preset duration is set after the trigger. The absolute value of the rate of change of current is obtained by performing first-order difference calculation on the instantaneous sampled current values. .

[0042] Specifically, after the triggering time, continuous data collection... For each current sampling point, calculate the absolute value of the current difference between adjacent sampling points, find their arithmetic mean, and then divide by the sampling time interval. ,get Approximate value.

[0043] Preset duration Get experience points (milliseconds), corresponding number of sampling points .

[0044] Current change rate The calculation formula is:

[0045]

[0046] In the formula, For the triggering of the first The instantaneous current value at each sampling point, in amperes (A); The sampling time interval is expressed in seconds (s). The number of sampling points used to calculate the rate of change is dimensionless.

[0047] The unit is amperes per second (A / s), which reflects the overall rate of current rise during the initial period after triggering.

[0048] Current change rate The magnitude of the fault current directly reflects the rapid increase in fault current; the faster the current rises, the shorter the observation time allowed by the protection device.

[0049] Therefore, based on The window shortening coefficient is calculated using a preset linear mapping function. The mapping function has the following form:

[0050]

[0051] In the formula, The maximum current change rate is a preset benchmark value, corresponding to the maximum current rise rate under a typical metallic short-circuit fault, and is taken as an empirical value. It can also be adjusted by the implementer based on the line impedance and short-circuit capacity.

[0052] From this mapping relationship, it can be seen that when As it approaches zero, near ;when Increase and approach or exceed hour, Approaching and maintaining the lower limit .

[0053] Window shortening factor The value of is restricted to to Between them, there is no dimension.

[0054] Window shortening factor With preset base window length Multiply by this to obtain the dynamic analysis window duration applicable to this overcurrent event. The calculation formula is:

[0055]

[0056] In the formula, the basic window length Get experience points (milliseconds), this value is determined based on the longest duration of a typical load start-up transient process, and can also be adjusted by the implementer according to the characteristics of the field load.

[0057] This is determined by This is the duration of the analysis window corresponding to this overcurrent event.

[0058] S2: Perform energy envelope morphology matching and fundamental phase verification in parallel within the window.

[0059] The analysis window duration is obtained by responding to overcurrent triggering and dynamically shortening it based on the rate of current change. Subsequently, within the limited timeframe of this window, an energy envelope matching path and a phase consistency verification path are simultaneously opened in parallel. The two paths process the real-time sampled data independently and do not block each other, continuously generating and updating the normal input event flags respectively. Signs of severe morphological abnormalities and phase destruction markers .

[0060] The parallel execution is implemented as follows: when an interrupt occurs at each sampling point, the interrupt service routine simultaneously updates the energy accumulation sequence and caches voltage and current data; while the calculation of the fundamental phase difference utilizes a timed interrupt per power frequency cycle to perform a discrete Fourier transform on the cached data for one cycle, wherein the interrupt time is every The power frequency period is All of these can be adjusted by the implementer according to the specific implementation scenario.

[0061] The two paths share the same set of sampled data but maintain their own independent state variables, and are guaranteed not to block each other by enabling a global interrupt.

[0062] In the energy envelope matching path, from the trigger time Start by taking the instantaneous voltage value at each sampling point The units are volts (V) and instantaneous current values. The unit is amperes (A), multiplied by the sampling time interval. The unit is seconds (s), and then the transient energy accumulation sequence is obtained by accumulating the points one by one.

[0063]

[0064] in, The sampling point number after triggering. The total number of sampling points within the window and , The unit is joules (J); because different load power levels will lead to differences in the absolute amplitude of energy, and shape matching needs to be based on the relative temporal distribution of energy injection, therefore... Perform self-normalization as follows:

[0065] in, Accumulate energy at the end of the window. Take the smallest positive number The implementation can be adjusted by the implementer according to the specific implementation scenario. The complete normalized energy envelope morphology sequence is obtained at the end of the analysis window. It is generated later and used for event logging and template updates when conditions are met.

[0066] During the analysis window, for the current sampling point The processor has been obtained by to The processor directly calculates the Pearson correlation coefficient between the transient energy accumulation sequence prefix and the corresponding equal-length prefix truncated from the starting point of each template. .

[0067] This direct correlation calculation does not require prior knowledge of the accumulated energy at the window endpoint. Specifically, according to the original normalization formula, let... The original instruction manual stated that the normalized energy envelope morphology sequence was taken from... Therefore, under the load input energy direction described in this application .

[0068] because For all elements in the prefix of the same transient energy accumulation sequence that form a common proportionality factor, according to the definition of the Pearson correlation coefficient, we have:

[0069]

[0070] Therefore, the correlation coefficient calculated directly using the current transient energy cumulative sequence prefix within the analysis window is the same as the correlation coefficient obtained after common normalization using the cumulative energy at the end of the window. This calculation does not use the current cumulative energy. It accumulates energy at the end of the alternative window and does not read future data.

[0071] Before using templates for matching, a standard template library needs to be established in advance. The offline calibration method for the pre-defined templates is as follows:

[0072] 1. Select typical loads that may appear on site (such as LED lights, variable frequency air conditioners, resistance heaters, motors, etc.). Each load should be put into operation independently at rated voltage, and repeated more than 10 times. Here, 10 is an empirical value and can be adjusted by the implementer according to the specific implementation scenario.

[0073] 2: Each input begins from the overcurrent trigger moment, and data collection... The voltage and current data over the specified time period are used to obtain the energy envelope sequence using the self-normalization method described above. These are empirical values ​​and can be adjusted by the implementer based on the specific implementation scenario.

[0074] 3: Average multiple sequences of the same load point by point to obtain the standard template sequence for that load. .

[0075] 4: Template length is fixed. point.

[0076] 5: When the actual analysis window At that time, the template Before cutting off from the starting point One point is used for correlation calculation; when (This will not happen, because) If ), then linear interpolation expansion is performed on the template.

[0077] After the device is put into operation, the matched templates can be updated by performing an exponentially weighted moving average on each event confirmed as normal operation (and without enabling the enhancement threshold). The update coefficient is taken as... To track slow changes in load characteristics, where the update coefficient is taken as... These are empirical values ​​and can be adjusted by the implementer based on the specific implementation scenario.

[0078] Therefore, the Pearson correlation coefficient The calculation formula is:

[0079]

[0080] And take the maximum value As the current form matching degree; when Reaching or exceeding the normal input confirmation threshold At that time, the normal input event marker will be activated. Place And freeze the state of that path, if Below the threshold for severe morphological abnormalities Then the cumulative duration of low correlation, when Restore to The duration of low correlation will be reset to zero when the above values ​​are continuously below a certain threshold; The duration reached the defect confirmation delay. At that time, the severe morphological abnormality will be marked. Place .

[0081] in, Get experience points , which are preset parameters. Get experience points , which are preset parameters. Get experience points These are preset parameters, all of which can be adjusted by the implementer.

[0082] In the phase consistency verification path, the fundamental frequency discrete Fourier transform is used to calculate the current fundamental frequency voltage phase angle once per power frequency cycle within the analysis window. Phase angle with fundamental current The measured fundamental phase difference was obtained. The unit is degrees (°); when the first At the end of each cycle, based on the first-order trend prediction model, using the current measured values... Historical steady-state phase difference statistical mean The unit is degrees (°) and the trend inertia coefficient. Calculate the predicted value of the phase difference for the next cycle. Among them, the trend inertia coefficient Get experience points , dimensionless, is a preset parameter.

[0083]

[0084] In the The cycle arrived and actual measurements were obtained. Then, calculate the predicted residuals:

[0085]

[0086] The residual The unit is degrees (°), which measures the degree of deviation between the actual phase value and the first-order trend prediction.

[0087] If the calculated predicted residual Exceeding the phase destruction threshold This will destroy the confirmation counter. Add one, otherwise Zeroing out, among which, Get experience points These are preset parameters, but can also be adjusted by the implementer according to the on-site noise level.

[0088] when The count value reached the number of consecutive confirmations When a persistent phase anomaly is detected, the phase destruction flag is set. Place And immediately stop updating. Only use the statistical values ​​that were locked before the fault occurred; The method of obtaining the value is to calculate the arithmetic mean of the measured phase differences over a period of continuous operation of the device and when the predicted residuals do not exceed the limit. Get experience points These are preset parameters, which can also be adjusted by the implementer.

[0089] In the dynamic analysis window Throughout the entire duration, the two paths mentioned above always operate independently in parallel. The energy envelope matching path only uses the transient energy accumulation sequence prefix obtained up to the current sampling point to perform equivalent Pearson correlation calculation, while the phase verification path uses the cached power frequency cycle data to perform first-order trend prediction and residual continuity confirmation. Neither path reads future sampling data, and each path updates the flag it is responsible for in real time according to its own discrimination criteria. The generated flag is maintained until the end of the window.

[0090] S3: When the window ends, determine whether to trip based on priority and perform normal tripping.

[0091] In the dynamic analysis window At the end of the event, stop updating the data of the energy envelope matching path and the phase verification path, and read the latched normal input event flag. Phase destruction marker Signs of severe morphological abnormalities and the current maximum morphological matching degree Based on the following combination of conditions, it determines whether the current overcurrent event belongs to normal load input, and thus decides whether to suppress the trip command.

[0092] Otherwise, if there is a phase anomaly and the morphological matching degree is not higher than the enhancement threshold, or if there is a morphological anomaly, the circuit breaker will trip.

[0093] when and At that time, the similarity between the energy envelope shape and the known template had reached the first threshold, and the fundamental phase difference did not show continuous disruption under the first-order trend prediction. Both paths pointed to normal load input. Therefore, the event was determined to be a normal transient, and no tripping signal was generated. , and Reset, and optionally use the normalized energy envelope morphology sequence acquired this time to smoothly update the template with the highest matching degree to reflect the slow change of load characteristics.

[0094] When the phase destruction sign At that time, regardless of the normal input event marker For each value, a more stringent energy-matching enhancement threshold than the first threshold is introduced. . Must be greater than , Experience points .

[0095] If the current shape matches If the load is normal and the circuit breaker is not tripped, it is considered a normal load input; otherwise, it is considered a fault event and the circuit breaker will trip.

[0096] In particular, when and When the confidence level of the conclusion that the circuit breaker will not trip is extremely high, updates to the template library based on this data should be disabled; when but In this case, it indicates a rare occurrence where the circuit breaker does not trip, but the event is recorded.

[0097] If any of the above non-tripping conditions is met, the trip actuator is not activated, and the circuit returns to the effective value of the current sliding window. The monitoring status is checked, waiting for the next overcurrent trigger; otherwise, if or If the above conditions are not met (i.e., the event is considered abnormal and the circuit breaker is tripped); if none of the above conditions are met (i.e., the event is considered abnormal and the circuit breaker is tripped), the event is considered abnormal and the circuit breaker is tripped. and If the current event is not a fault, no trip command will be generated, and the system will return to the monitoring state, while recording the unmatched event.

[0098] In summary, at the end of the dynamic analysis window, decisions should be made in the following priority order:

[0099] like and If so, the circuit breaker will not trip.

[0100] like and If so, the circuit breaker will not trip.

[0101] like or If so, the circuit breaker will trip.

[0102] otherwise( and ),without tripping.

[0103] After any of the above non-trip conclusions takes effect, the terminal returns to the valid value of the current sliding window. The online monitoring status was cleared. , , The indicator indicates that if a trip is required, the trip operation will be performed.

[0104] Regardless of the conclusion, all transient flags are cleared when the window ends.

[0105] If the current window is terminated prematurely due to a forced trip, the same clearing action will be performed.

[0106] When a trip is deemed necessary, a trip command is immediately generated, driving the trip actuator to disconnect the protected circuit. After the trip command is output, the type of fault, trigger time, and key discrimination parameters are latched for event logging. Simultaneously, the sampling and calculation process within the current analysis window is terminated, and the terminal switches to post-trip protection status, awaiting manual reset or remote command.

[0107] S4: Force tripping when the preset maximum allowable disconnection time is reached and the current still exceeds the limit.

[0108] In the event that the preset maximum allowable disconnection time has been reached and the current is still exceeded, an emergency disconnection timer independent of the analysis window is started synchronously from the overcurrent trigger moment. The preset duration of this timer is an empirical value of 100 milliseconds, and the unit is milliseconds (ms). This is a preset parameter and can be adjusted by the implementer according to the thermal stability limit of the cable.

[0109] Once the timer starts incrementing, it is unaffected by the window shortening factor, thus forming the final cut-off time boundary independent of transient waveforms and phase analysis.

[0110] The emergency tripping timer and the dynamic analysis window run in parallel and do not wait for each other: if the timer reaches the maximum allowable tripping time before the dynamic analysis window and the current is still over the limit, the current analysis window is immediately terminated and a forced trip is executed; if the dynamic analysis window ends first and does not trip temporarily, but the current is still over the limit when the timer reaches the maximum allowable tripping time, a forced trip is still triggered to ensure that the thermal stability safety boundary takes priority.

[0111] When the timer's accumulated value reaches (Unit: milliseconds ms) Read the latest effective value of the current sliding window. (Unit: Ampere A) and compared with the absolute threshold for severe overcurrent. (Unit: Ampere, A) Comparison, Get experience points , The effective value of the line's rated current (in amperes A) and a multiple thereof These are preset parameters and can be adjusted.

[0112] like This indicates that even after the maximum permissible disconnection time, a severe overcurrent exceeding the cable's thermal stability limit still exists in the circuit; this should be ignored. , and The transient flag directly generates a forced trip command to drive the trip actuator to disconnect the protected circuit. In this case, a forced trip will not be triggered and the issue will be handled by other criteria.

[0113] After the forced trip command takes effect, the trigger type of this event is latched as timeout forced trip, and the trigger time is recorded. The time of resection and the effective value of the current at the time of resection. Used for event logging and fault tracing, and subsequently, transient flags are set. , and All data is cleared and sampling and data processing in the current analysis window are terminated. The terminal then enters the post-trip protection state and waits for manual reset or remote command.

[0114] This invention also discloses a multi-parameter fusion analysis and protection system based on load characteristic identification, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the multi-parameter fusion analysis and protection method based on load characteristic identification of this invention. The system also includes other components well-known to those skilled in the art, such as a communication bus and communication interface; their configuration and functions are known in the art and will not be described further here.

[0115] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A multi-parameter fusion analysis protection method based on load characteristic identification, characterized in that, include: In response to overcurrent, the analysis window is dynamically shortened based on the rate of change of current; wherein, the rate of change of current... Analysis window duration and number of window sampling points Determined according to the following formula: In the formula, For the first time after overcurrent triggering The instantaneous current value at each current sampling point The sampling time interval, The number of sampling points used to calculate the rate of change of current. The preset maximum reference value for the rate of change of current. This is the preset base window length; Energy envelope morphology matching and fundamental phase verification are performed in parallel within the window. Morphology matching is initiated by accumulating instantaneous power values. During the window, the Pearson correlation is directly calculated between the transient energy accumulation sequence prefix obtained up to the current sampling point and the corresponding equal-length prefix of the template. This direct Pearson correlation calculation is algebraically equivalent to dividing each element in the transient energy accumulation sequence prefix by the sum of the cumulative energy at the end of the same window and a preset positive constant before performing the Pearson correlation calculation. After the window ends and the cumulative energy at the end of the window is obtained, the complete transient energy accumulation sequence is self-normalized. If the morphology matching degree reaches the first threshold, normal load input is confirmed; if it remains below the second threshold, morphology anomaly is marked. Phase verification performs first-order trend prediction on the fundamental phase difference; if the prediction residual continuously exceeds the limit, phase anomaly is judged. If, at the end of the window, it is confirmed that normal load is being supplied and there is no phase abnormality, or if the similarity is higher than the enhanced threshold which is stricter than the first threshold, then the circuit breaker will not trip. Otherwise, if there is a phase anomaly and the morphological matching degree is not higher than the enhancement threshold, or if there is a morphological anomaly, the circuit breaker will trip. It will force a trip when the preset maximum allowable disconnection time is reached and the current still exceeds the limit.

2. The multi-parameter fusion analysis and protection method based on load characteristic identification according to claim 1, characterized in that, The dynamic shortening of the analysis window based on the rate of change of current includes: Obtain the absolute value of the rate of change of current within a preset time after triggering; The window shortening coefficient is obtained by using a preset linear mapping function based on the absolute value of the current change rate. The duration of the analysis window is calculated as the product of the window shortening coefficient and the preset base window length.

3. The multi-parameter fusion analysis and protection method based on load characteristic identification according to claim 1, characterized in that, The self-normalization includes: Starting from the overcurrent trigger moment, the product of the instantaneous voltage and instantaneous current values ​​at each sampling point is multiplied by the sampling interval and then accumulated to obtain the transient energy accumulation sequence. During the analysis window, for the prefix formed by the currently obtained transient energy accumulation sequence, the Pearson correlation is directly calculated with the corresponding equal-length prefix of the preset load template. After the analysis window ends and the accumulated energy at the window endpoint is obtained, each element in the complete transient energy accumulation sequence is divided by the sum of the accumulated energy at the window endpoint and a preset positive constant to obtain the normalized energy envelope pattern sequence.

4. The multi-parameter fusion analysis and protection method based on load characteristic identification according to claim 1, characterized in that, The first threshold and the second threshold are preset parameters.

5. The multi-parameter fusion analysis and protection method based on load characteristic identification according to claim 1, characterized in that, The similarity comparison with the template includes: The energy envelope pattern sequence obtained by processing the instantaneous value of accumulated power after self-normalization is correlated with multiple preset load templates one by one to obtain the correlation coefficient corresponding to each template. The maximum value among all correlation coefficients is taken as the similarity.

6. The multi-parameter fusion analysis and protection method based on load characteristic identification according to claim 1, characterized in that, The first-order trend prediction includes: Obtain the measured phase difference between the fundamental voltage and fundamental current in the current cycle; Based on the measured phase difference, the preset trend inertia coefficient, and the statistical mean of historical steady-state phase differences, the predicted phase difference value for the next cycle is calculated. The prediction residual is calculated based on the predicted phase difference value and the measured phase difference in the next cycle; If the number of times the predicted residual exceeds the preset phase disruption threshold reaches the preset confirmation number, then a phase anomaly is determined.

7. The multi-parameter fusion analysis and protection method based on load characteristic identification according to claim 6, characterized in that, The statistical mean of the historical steady-state phase difference was obtained through the following method: When the predicted residuals for multiple consecutive cycles do not exceed the limit and the current is in a stable range, the measured phase difference for the corresponding cycle is arithmetically averaged and updated to the statistical mean of the historical steady-state phase difference.

8. The multi-parameter fusion analysis and protection method based on load characteristic identification according to claim 1, characterized in that, The application of the enhancement threshold includes: When the fundamental phase verification determines that the phase is abnormal, if the similarity obtained by the energy envelope morphology matching is higher than the enhancement threshold, the circuit breaker will not trip. If the similarity is lower than or equal to the enhancement threshold, it is determined to be a fault event and the circuit breaker trips.

9. The multi-parameter fusion analysis and protection method based on load characteristic identification according to claim 1, characterized in that, The preset maximum allowable cut-off time and the current still exceed the preset parameter value.

10. A multi-parameter fusion analysis and protection system based on load characteristic identification, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the multi-parameter fusion analysis and protection method based on load characteristic identification according to any one of claims 1-9.

Images