Power fitting fault risk intelligent early warning system based on operation big data

Abstract

The present application relates to electric digital data processing and transmission line state monitoring technical field, especially to a kind of electric power fitting fault risk intelligent early warning system based on operation big data, the system includes processor and memory, processor executes the computer program of memory, to realize the following steps: according to the temperature rise response value of effective reference fitting of target fitting forms common response value;According to the difference between the temperature rise response value of target fitting and common response value, obtain the local response residual of target fitting;According to the current variation flowing through target fitting, identify operation excitation event, simultaneously according to the recovery condition of current, common response value and local response residual after event, judge whether it is assessable operation excitation event, and then according to the local response residual difference of target fitting before and after event, carry out fault risk early warning to target fitting, improve the stability and accuracy of early warning to fitting fault risk.

Description

Technical Field

[0001] This invention relates to the field of electrical digital data processing and transmission line condition monitoring technology, and in particular to an intelligent early warning system for power fitting fault risks based on big data operation. Background Technology

[0002] Tension clamps, splicing pipes, and current-carrying fittings in transmission lines bear the long-term responsibility of connecting conductors, transmitting mechanical loads, and transmitting current. When the contact interface of the fittings experiences a decline in crimping quality, oxidation of the contact surface, loosening of the connection, or localized corrosion, the contact resistance may increase, leading to localized heating under the influence of load current. Therefore, the temperature of the fittings and its temperature rise relative to the ambient temperature are important physical responses in monitoring the operational status of the transmission line.

[0003] Current methods for monitoring the thermal condition of electrical fittings typically involve obtaining fitting temperatures through infrared inspection, online temperature measurement terminals, or other temperature acquisition devices. Anomalies are then identified using absolute temperature, temperature rise relative to ambient temperature, relative temperature difference, or rules combining these with line operating parameters. While these methods can detect obvious overheating, transmission line loads fluctuate continuously during daily scheduling, seasonal changes, and transient operations. Fitting temperatures are simultaneously affected by line current, ambient heat exchange conditions, and the fitting's own condition. Even healthy fittings can experience significant temperature increases and decreases with load variations.

[0004] In continuous big data scenarios, elevated temperature rises at a single moment or within a short time window may originate from localized changes in the target fitting itself, or from load changes experienced by the entire line and subsequent thermal inertia. If long-term risk states are directly established based on fixed temperature thresholds, single relative temperature differences, or the number of abnormal events, the recoverable response caused by the load may be carried over to the long-term cumulative results. Conversely, if only the simultaneous temperature difference of similar fittings is used, it is difficult to confirm whether the target fitting has truly returned to a comparable thermal state after a single load excitation. Therefore, existing fitting thermal status monitoring still faces the problem of distinguishing recoverable temperature rises caused by load changes and thermal inertia from the fitting's own continuous abnormal trends. This can easily affect the stability of long-term fault risk assessment and the determination of operational and maintenance priorities, leading to a decrease in the accuracy of early warnings for fitting fault risks.

[0005] Therefore, improving the stability and accuracy of early warning of hardware failure risks has become an urgent problem to be solved. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide an intelligent early warning system for power fitting failure risks based on operational big data, in order to solve the problem of how to improve the stability and accuracy of early warning of fitting failure risks.

[0007] This invention provides an intelligent early warning system for power fitting fault risks based on big data, including a memory, a processor, and a computer program stored in the memory and running on the processor. The system is characterized in that the processor executes the computer program to perform the following steps: Based on the difference between the temperature of each fitting and the ambient temperature at each sampling time, the temperature rise response value of each fitting at each sampling time is obtained. Any fitting is recorded as the target fitting. Based on the device identification of the target fitting and the topology of its line, as well as the temperature rise response value of each fitting at each sampling time, the effective reference fitting of the target fitting at each sampling time is obtained. Based on the number of valid reference fittings and the discrete distribution of temperature rise response values ​​at each sampling time, the common response value at each sampling time is obtained. Based on the difference between the temperature rise response value of the target fitting and the common response value at each sampling time, the local response residual of the target fitting at each sampling time is obtained. Based on the change in current flowing through the target fitting, an operational stimulus event is identified, and a pre-event stability window, an event action window, and a post-event recovery window are established for the operational stimulus event. Based on the current change in the line where the target fitting is located, the common response value of the target fitting, and the recovery status of the local response residual within the post-event recovery window, it is determined whether the operational stimulus event is an evaluable operational stimulus event. Based on whether the operational stimulus event is an evaluable operational stimulus event, and the difference in local response residuals between the target hardware before the event and after the event, a fault risk warning is issued for the target hardware.

[0008] Preferably, obtaining the temperature rise response value of each fitting at each sampling time includes: For any sampling time, if neither the ambient temperature nor the hardware temperature of any hardware is marked as invalid, then the ambient temperature is subtracted from the hardware temperature of any hardware at any sampling time to obtain the temperature rise response value of any hardware at any sampling time. If the ambient temperature or the temperature of any hardware at any sampling time is marked as invalid, then the temperature rise response value of any hardware at any sampling time is set to empty.

[0009] Preferably, the step of obtaining the effective reference hardware for the target hardware at each sampling time includes: Based on the topology of the line where the target fitting is located, other fittings with the same fitting type and conductor specifications as the target fitting are obtained within the topological neighborhood of the current-carrying path of the same line as the target fitting, and are recorded as the reference fittings of the target fitting. For any sampling time, the reference fitting with a non-empty temperature rise response value is recorded as the valid reference fitting for the target fitting at that sampling time.

[0010] Preferably, obtaining the common response value at each sampling time includes: For any sampling time, calculate the median absolute deviation of the temperature rise response values ​​of all valid reference hardware of the target hardware at that sampling time, and multiply the median absolute deviation by a constant 1.4826 to obtain the reference discrete scale at that sampling time. If the number of valid reference hardware for the target hardware at any sampling time is greater than or equal to the preset minimum number of reference hardware, and the reference discrete scale at any sampling time is less than or equal to the preset discrete scale threshold, then the median of the temperature rise response values ​​of all valid reference hardware for the target hardware at any sampling time is obtained as the common response value at any sampling time. If the number of valid reference fittings for the target fitting at any sampling time is less than the preset minimum number of reference fittings, or the reference discrete scale at any sampling time is greater than the preset discrete scale threshold, then the common response value at any sampling time is set to empty.

[0011] Preferably, obtaining the local response residual of the target hardware at each sampling time includes: For any sampling time, if the common response value at any sampling time is not empty, then the temperature rise response value of the target hardware at any sampling time is subtracted from the common response value at any sampling time to obtain the local response residual of the target hardware at any sampling time. If the common response value at any sampling time is empty, then the local response residual of the target hardware at any sampling time is set to empty.

[0012] Preferably, the step of identifying operational excitation events based on changes in current flowing through the target fitting, and establishing a pre-event stability window, an event action window, and a post-event recovery window for the operational excitation event, includes: Obtain a preset stable window length before the event and a preset current stabilization threshold before the event. Based on the preset stable window length before the event, obtain candidate stable windows with each sampling time as the starting time. Calculate the range of the current flowing through the target fitting within each candidate stable window. Record the candidate stable windows with a range less than the preset current stabilization threshold before the event as stable windows. According to the starting time corresponding to each candidate stable window, determine whether each candidate stable window is a stable window in turn. If there is an unstable window, record the stable window before the unstable window as a candidate stable window before the event. Obtain the median of the current flowing through the target fitting at all sampling times within the stabilization window before the candidate event, and use it as the baseline for the line current before the event. For any sampling time after the stabilization window before the candidate event, calculate the relative change amplitude of the current flowing through the target fitting at that sampling time based on the absolute value of the difference between the current flowing through the target fitting at that sampling time and the baseline for the line current before the event. The calculation formula is as follows: , In the formula, This represents the relative change amplitude of the current flowing through the target fitting at any given sampling time. This represents the current flowing through the target fitting at any given sampling time. Indicates the line current reference before the event. This indicates the preset minimum reference current. Represents the absolute value symbol. This indicates taking the maximum value; The relative change amplitude of the current flowing through the target hardware is obtained sequentially at each sampling time after the stabilization window before the candidate event. The number of continuous samples for the preset event entry and the relative change threshold of the current entering the preset event are obtained. Based on the number of continuous samples for the preset event entry, candidate event windows are obtained sequentially with each sampling time after the stabilization window before the candidate event as the starting time. It is then determined whether each candidate event window meets the excitation event condition. The excitation event condition is that the relative change amplitude of the current flowing through the target hardware at each sampling time within the candidate event window is greater than or equal to the relative change threshold of the current entering the preset event. The starting time of the first candidate event window that meets the excitation event condition is recorded as the excitation event entry time. The stable window before the arrival time of the stimulus event is obtained and is closest to the arrival time of the stimulus event. This window is denoted as the stable window before the event of the running stimulus event. Obtain a preset event exit current relative change threshold and a preset event exit duration sample number. Based on the preset event exit duration sample number, sequentially obtain candidate exit windows starting from each sampling time after the first candidate event window that meets the excitation event condition. Sequentially determine whether each candidate exit window meets the excitation exit condition. The excitation exit condition is that the relative change amplitude of the current flowing through the target hardware at each sampling time within the candidate exit window is less than the preset event exit current relative change threshold. Record the end time of the first candidate exit window that meets the excitation exit condition as the excitation event exit time. The window formed between the entry time of the stimulus event and the exit time of the stimulus event is denoted as the event action window of the running stimulus event; Obtain the preset event recovery window length, and based on the preset event recovery window length, take the exit time of the stimulus event as the start time to obtain the event recovery window for the running stimulus event.

[0013] Preferably, determining whether the runtime stimulus event is an evaluable runtime stimulus event includes: If there is a sampling moment with a common empty response value within the pre-event stabilization window, event action window, and post-event recovery window of the running stimulus event, then the running stimulus event is determined to be an unevaluable running stimulus event. If there are no sampling moments with a common response value of empty within the pre-event stabilization window, event action window, and post-event recovery window of the running excitation event, then the median of the current flowing through the target fitting at all sampling moments within the pre-event stabilization window is calculated as the pre-event line current reference. The median of the current flowing through the target fitting at all sampling moments within the post-event recovery window is calculated as the post-event line current reference. The absolute value of the difference between the pre-event line current reference and the post-event line current reference is used as the numerator, and the maximum value between the pre-event line current reference and the preset minimum reference current is used as the denominator to obtain the relative difference of the current reference before and after the event. Calculate the median of the common response values ​​at all sampling times within the pre-event stability window, and use it as the pre-event common response benchmark; Calculate the median of the common response values ​​at all sampling times within the post-event recovery window, and use it as the common response benchmark after the event; Obtain a preset line current recovery threshold, a preset common response recovery threshold, and a preset residual stability threshold. Set the evaluation criteria for the excitation event as follows: the relative difference between the current reference before and after the event is less than or equal to the preset line current recovery threshold, the absolute value of the difference between the common response reference before the event and the common response reference after the event is less than or equal to the preset common response recovery threshold, and the range of the local response residuals at all sampling times within the recovery window after the event is less than or equal to the preset residual stability threshold. If the pre-event stabilization window and post-event recovery window of the running stimulus event meet the evaluable running stimulus event determination criteria, then the running stimulus event is determined to be an evaluable running stimulus event. If the pre-event stabilization window and post-event recovery window of the operational stimulus event do not meet the criteria for evaluating an operational stimulus event, the post-event recovery window is shifted backward by one sampling period to obtain a new post-event recovery window. Based on the new post-event recovery window, the post-event line current reference, the post-event common response reference, the relative difference between the current references before and after the event, and the range of the local response residuals at all sampling moments within the post-event recovery window are recalculated. If there are no sampling moments with empty common response values ​​within the new post-event recovery window, and the pre-event stabilization window and the new post-event recovery window meet the criteria for evaluating an operational stimulus event, the new post-event recovery window replaces the post-event recovery window as the post-event recovery window, and the operational stimulus event is determined to be an evaluable operational stimulus event. If there is a sampling moment with a common response value of empty within the recovery window after the new event, or if the stable window before the event and the recovery window after the new event do not meet the criteria for determining an evaluable operational stimulus event, then the recovery window after the new event is moved backward by one sampling period to obtain a new recovery window after the event, until the total duration of the backward sampling periods reaches the preset maximum recovery waiting time. If the criteria for determining an evaluable operational stimulus event are still not met, and the stable window before the event and the recovery window after the new event do not meet the criteria for determining an evaluable operational stimulus event, then the operational stimulus event is determined to be an unevaluable operational stimulus event.

[0014] Preferably, the fault risk warning for the target hardware includes: If the operational stimulus event is an evaluable operational stimulus event, the median of the local response residuals at all sampling times within the pre-event stability window and the post-event recovery window of the operational stimulus event is calculated respectively, and used as the pre-event response residual benchmark and the post-event response residual benchmark of the operational stimulus event. The pre-event response residual benchmark is subtracted from the post-event response residual benchmark to obtain the change in the event residual of the target hardware after the operational stimulus event. In the historical operation and maintenance records, obtain the historical evaluable operation incentive events of the target hardware and its valid reference hardware that have been confirmed to be normal after re-inspection, and record them as historical normal operation incentive events. If the number of historical normal operation incentive events is less than the preset minimum number of normal samples, save the event residual change of the operation incentive event, and set the degradation memory state value of the target hardware after the operation incentive event to the preset initial degradation memory state value. If the number of historical normal operation incentive events is greater than or equal to the preset minimum number of normal samples, the event residual change of each historical normal operation incentive event is obtained, the median and median absolute deviation of the event residual changes of all historical normal operation incentive events are calculated, a preset sensitivity coefficient is obtained, the preset sensitivity coefficient, the constant 1.4826 and the median absolute deviation are multiplied together to obtain the target value, the sum between the median of the event residual changes of all historical normal operation incentive events and the target value is calculated, and the sum result is obtained. The maximum value between the sum result and the constant 0 is selected as the boundary of normal event residual change. Based on the difference between the event residual change of the operational stimulus event and the boundary of the normal event residual change, the degradation memory state value of the target hardware after the operational stimulus event is obtained; If the operational stimulus event is an unevaluable operational stimulus event, and there is an evaluable operational stimulus event before the operational stimulus event, then the degradation memory state value of the target hardware after the most recent evaluable operational stimulus event is obtained, and used as the degradation memory state value of the target hardware after the operational stimulus event. If the running stimulus event is an unevaluable running stimulus event, and there is no evaluable running stimulus event before the running stimulus event, then the degradation memory state value of the target hardware after the running stimulus event is set to a preset initial degradation memory state value. Based on the degradation memory state value of the target hardware after the operational excitation event, a fault risk warning is issued for the target hardware.

[0015] Preferably, obtaining the degradation memory state value of the target hardware after the operational stimulus event includes: The target difference is obtained by subtracting the normal event residual change boundary from the event residual change of the running stimulus event. The maximum value between the target difference and the constant 0 is selected as the unrecovered response increment of the target hardware after the event of the running stimulus event. If there is no evaluable operational stimulus event before the operational stimulus event, the sum of the unrecovered response increment after the event and the preset initial degradation memory state value is taken as the degradation memory state value of the target hardware after the operational stimulus event. If there is at least one evaluable operational stimulus event before the operational stimulus event, the post-event unrecovered response increment of the target hardware after all evaluable operational stimulus events before the operational stimulus event is obtained using the normal event residual change boundary. This is recorded as the post-event unrecovered response increment. The sum of the preset initial degradation memory state value, all post-event unrecovered response increments, and the post-event unrecovered response increment of the operational stimulus event is calculated to obtain the degradation memory state value of the target hardware after the operational stimulus event.

[0016] Preferably, the step of providing a fault risk warning for the target hardware based on the degradation memory state value of the target hardware after the operational stimulus event includes: Obtain a preset first risk classification threshold and a preset second risk classification threshold, wherein the preset second risk classification threshold is greater than the preset first risk classification threshold; If the degradation memory state value of the target hardware after the operation stimulus event is less than the preset first risk classification threshold, then the target hardware will be subject to routine monitoring. If the degradation memory state value of the target hardware after the operation stimulus event is greater than or equal to the preset first risk classification threshold and less than the preset second risk classification threshold, then the target hardware will be added to the key monitoring set and the target hardware will be monitored. If the degradation memory state value of the target hardware after the running stimulus event is greater than or equal to the second risk classification threshold, a risk fault warning is issued for the target hardware, and a re-inspection task is generated.

[0017] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows: This invention first utilizes effective reference fittings located on the same current-carrying path as the target fitting, possessing the same fitting type and conductor specifications, to form a common response value. This weakens the common temperature rise caused by changes in line current and the surrounding environment in the local response residual of the target fitting. Furthermore, using the operational excitation event as the time unit, the change in event residual is calculated only after the line current and the average of the common response return to a comparable state before the event, and the local response residual of the target fitting stabilizes. This avoids the period before thermal inertia has recovered. Simultaneously, the difference in local response residuals before and after the event is used to form the post-event unrecovered response increment. The normal event residual change boundary is dynamically updated using data confirmed to be normal after re-inspection, and a degradation memory state value is calculated based on the event residual change to characterize the cumulative trend of newly added unrecovered local responses during the monitoring period. Based on the degradation memory state value, graded early warning and re-inspection tasks are generated, allowing data processing results to directly change the monitoring and re-inspection priority of the target fitting. This improves the stability and accuracy of early warning for fitting failure risks. Detailed Implementation

[0018] To illustrate the technical solution of the present invention, specific embodiments are described below. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure; rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0019] This invention provides an intelligent early warning system for power fitting fault risks based on operational big data, including a processor and a memory. The processor executes a computer program stored in the memory to implement an intelligent early warning method for power fitting fault risks based on operational big data. The method includes the following steps: Step S101: Based on the difference between the temperature of each fitting and the ambient temperature at each sampling time, obtain the temperature rise response value of each fitting at each sampling time, record any fitting as the target fitting, and obtain the valid reference fitting of the target fitting at each sampling time based on the device identifier of the target fitting, the topology of its line, and the temperature rise response value of each fitting at each sampling time.

[0020] A hardware device identifier is established for each monitored hardware (a unique coded data identifier that corresponds to a specific hardware). The hardware temperature of each hardware is acquired at each sampling time according to a preset sampling period for fault risk monitoring. Regarding the preset sampling frequency, considering the significant thermal inertia and relatively slow rate of temperature change in transmission line hardware, while simultaneously requiring the identification of line load change events, the sampling period is generally set within the range of 2 to 10 minutes in actual engineering. In this embodiment, the preset sampling period is set to 5 minutes; however, this is not a limitation and implementers can set it according to specific scenarios.

[0021] The ambient temperature at each sampling time is obtained from either a line environment temperature measuring device or a meteorological monitoring device whose temperature data is time-aligned with that of the fittings. For each sampling time, the fitting temperature is subtracted from the ambient temperature to obtain the fitting's temperature rise response value, in units consistent with the temperature unit. Taking the g-th fitting at sampling time t as an example, the temperature rise response value of the g-th fitting at sampling time t is... In the formula, This represents the temperature rise response value of the g-th fitting at sampling time t. This represents the temperature of the g-th fitting at sampling time t. This represents the ambient temperature at sampling time t. If the temperature of a certain fitting or the ambient temperature at a certain sampling time is missing or marked as invalid by the data source, the temperature rise response value of that fitting at that sampling time is set to null.

[0022] The fault risk monitoring process for each monitored fitting is consistent. In this embodiment of the invention, any monitored fitting is designated as the target fitting, and the analysis is performed using the target fitting as an example. Since a high temperature rise in the target fitting at a single moment or within a short time window may originate from changes in the target fitting's own state, or from load changes experienced by the entire line and subsequent thermal inertia, this scheme first obtains valid reference fittings of the same type and conductor specifications within the topological neighborhood of the same current-carrying path as the target fitting. Then, based on the temperature rise of the valid reference fittings, the influence of overall line load changes is deducted to obtain the additional temperature rise of the target fitting. Furthermore, using the time interval of line load changes as the unit of analysis, the degradation of the target fitting is analyzed based on the difference in additional temperature rise before and after the load change, thereby providing a fault risk warning for the target fitting and improving the stability and accuracy of the fault risk warning.

[0023] Regarding the acquisition of valid reference tools: Taking any sampling time as an example, within the topological neighborhood of the same current-carrying path as the target fitting, other fittings of the same type and conductor specifications as the target fitting are acquired and denoted as reference fittings of the target fitting at that sampling time. If the temperature rise response value of a certain reference fitting at any sampling time is empty, it should be excluded. Therefore, reference fittings with non-empty temperature rise response values ​​at any sampling time are denoted as valid reference fittings of the target fitting at that sampling time. Similarly, valid reference fittings of the target fitting are acquired at each sampling time.

[0024] Step S102: Based on the number of valid reference fittings and the discrete distribution of temperature rise response values ​​at each sampling time, obtain the common response value at each sampling time. Based on the difference between the temperature rise response value of the target fitting and the common response value at each sampling time, obtain the local response residual of the target fitting at each sampling time.

[0025] After obtaining the effective reference hardware of the target hardware at each sampling time through the above steps, the common response affected by the load change of the entire line is analyzed based on the temperature rise of all effective reference hardware at each sampling time. The common response value at each sampling time is obtained, and then subtracted to obtain the local response residual of the target hardware at each sampling time, which is used to quantify the additional temperature rise of the target hardware after subtracting the common response.

[0026] First, the common response value at each sampling time is obtained. In this scheme, the median of the temperature rise response values ​​of all valid reference hardware at each sampling time is used as the common response value at each sampling time, with the unit consistent with the temperature rise response value. To prevent abnormal contamination of the common response by a single valid reference hardware, the validity of the common response value at each sampling time needs to be further determined, and invalid common response values ​​are set to null. Specifically:

[0027] Taking sampling time t as an example, the discreteness of the temperature rise response value of the effective reference hardware at sampling time t is first checked: the median absolute deviation of the temperature rise response values ​​of all effective reference hardware of the target hardware at sampling time t is calculated, and the median absolute deviation is multiplied by a constant 1.4826 (used to convert the median absolute deviation into a measure equivalent to the standard deviation, this constant comes from the mathematical derivation of the properties of the normal distribution) to obtain the reference discrete scale at sampling time t, denoted as . .

[0028] Considering that if the number of valid reference fittings for the target fitting is small at sampling time t, the valid reference fittings may not accurately and fully reflect the common response affected by the load changes of the entire line, a preset minimum number of reference fittings is set. and preset discrete scale threshold If the number of valid reference fittings for the target fitting at sampling time t is greater than or equal to the preset minimum number of reference fittings. And the reference discrete scale at sampling time t If the median of the temperature rise response values ​​of all valid reference fittings of the target fitting at sampling time t is true, it indicates that the median can accurately reflect the common response affected by the load change of the entire line. Therefore, the median is used as the common response value at sampling time t. Otherwise, if the number of valid reference fittings of the target fitting at sampling time t is less than 1, the median is used as the common response value at sampling time t. Or the reference discrete scale at the sampling time. Set the common response value at sampling time t to null. Similarly, obtain the common response value at each sampling time.

[0029] Among them, the minimum number of preset reference fittings The settings in this solution are as follows. The method is as follows: Within the topological neighborhood of the target fitting belonging to the same current-carrying path, obtain other fittings of the same type and conductor specifications as the target fitting, and use half of all other fittings as... Alternatively, it can be set based on the number of valid reference fittings that can typically be obtained simultaneously over a long period of time at each historical sampling point, but it is necessary to ensure that... .

[0030] For the preset discrete scale threshold For historical normal sampling times (i.e., no missing data and no abnormalities in fittings), the reference discrete scale for each historical normal sampling time is calculated according to the method for obtaining the reference discrete scale at sampling time t, and its normal range is obtained (which can be obtained through box plot method). The maximum value in the normal range is used as the preset discrete scale threshold. .

[0031] After obtaining the common response value at each sampling time, the temperature rise response value of the target fitting at each sampling time is subtracted from the common response value to obtain the local response residual of the target fitting at each sampling time. The unit is the same as that of the temperature rise response value. This subtraction preserves the local response difference of the target fitting relative to the effective reference fitting, thus mitigating the impact of overall line load variations on the corresponding residual of the target fitting. Taking sampling time t as an example, if the common response value at sampling time t is not empty, the formula for calculating the local response residual of the target fitting at sampling time t is:

[0032] In the formula, This represents the local response residual of the target hardware at sampling time t. This represents the temperature rise response value of the target hardware at sampling time t. This represents the set of valid reference fittings for the target fitting at sampling time t. This represents the temperature rise response value of the i-th valid reference hardware of the target hardware at sampling time t. This indicates taking the median. This is the common response value at sampling time t.

[0033] If the common response value at sampling time t is empty, then the local response residual of the target hardware at sampling time t is set to empty, i.e. Similarly, the local residual of the target hardware at each sampling time is obtained.

[0034] Step S103: Based on the change in current flowing through the target fitting, identify the operation excitation event, and establish the pre-event stability window, event action window, and post-event recovery window of the operation excitation event. Based on the current change in the line where the target fitting is located, the common response value of the target fitting, and the recovery status of the local response residual within the post-event recovery window, determine whether the operation excitation event is an evaluable operation excitation event.

[0035] This solution analyzes operational excitation events by monitoring changes in the current flowing through the target fittings. A significant change in the line current is considered an "operational excitation event." This event is then used to reflect the abnormal trends of the target fittings based on their degradation after the event. Furthermore, based on the recovery status before and after the event, fault risk warnings are provided for the target fittings, improving the stability and accuracy of fault risk warnings.

[0036] This scheme manages the time process of line current by sampling the waiting state, event action state, and recovery waiting state of the running excitation event. First, a preset pre-stabilization window length is set. and the preset current stabilization threshold before the event Preset stable window length To ensure a relatively stable current window before an operational stimulus event, a continuously stable current segment is obtained from the historical operational data of the line where the target fitting is located. This segment is then combined with the sampling period to determine how many sampling moments need to be included within the window, thus obtaining the preset pre-stabilization window length. This solution sets a preset pre-stabilized window length. The number of sampling times is 5 (i.e., 5 sampling moments), which is not limited here and can be set by the implementer according to the specific scenario; a preset current stability threshold before the event is also included. The preset pre-stabilization window length is calculated from the historical operating data of the line where the target fitting is located. The range of the current flowing through the target fitting is determined based on the fluctuation range of the range. In this scheme, a preset current stabilization threshold before the event is set. It is set to 10A, with no restrictions here; implementers can set it according to the specific scenario.

[0037] The continuous maintenance length of the system is The candidate stable window, i.e., based on the preset pre-event stable window length. Candidate stable windows are obtained with each sampling time as the starting time. The range of the current flowing through the target fitting within each candidate stable window is calculated. The current with a range less than the preset pre-event current stabilization threshold is selected. The candidate stable windows are denoted as stable windows. Based on the start time corresponding to each candidate stable window, it is sequentially determined whether each candidate stable window is a stable window. If an unstable window exists, the stable window preceding the unstable window is denoted as the candidate pre-event stable window. The median of the current flowing through the target fitting at all sampling times within the candidate pre-event stable window is used as the pre-event line current benchmark, denoted as... If a stable window has not yet formed before candidate events, i.e., no range greater than or equal to 1 / 2 has occurred. If the window is unstable, the system will remain in a waiting state for events.

[0038] After establishing the stabilization window before the candidate event, the event action window for the running excitation event is constructed based on the current changes after the stabilization window. First, the current changes after the stabilization window are analyzed: for any sampling time after the stabilization window, assuming it is sampling time t, the current flowing through the target fitting at sampling time t is compared with the baseline line current before the event. The absolute value of the difference is used to calculate the relative change amplitude of the current flowing through the target fitting at sampling time t. The calculation formula is as follows:

[0039] , In the formula, This represents the relative change in current flowing through the target fitting at sampling time t. This represents the current flowing through the target fitting at sampling time t. Indicates the line current reference before the event. This indicates the preset minimum reference current. Represents the absolute value symbol. This indicates taking the maximum value; Among them, for the preset minimum reference current The settings, The setting is based on the low load current level in the historical normal operation data of the line where the target fitting is located. Its function is to: when the current is very low before the event, Smaller than average This avoids using a very small reference current as the denominator, which would amplify a small absolute current change into a large relative change. The embodiments of this invention provide... There are no restrictions here; implementers can set it according to the specific circumstances of the line.

[0040] Similarly, the relative change amplitude of the current flowing through the target fitting is obtained sequentially at each sampling time after the stabilization window before the candidate event. This scheme sets the event activation state only when the relative change amplitude exceeds a threshold and continuously reaches the preset number of continuous samples for the event to activate. This is to distinguish between actual operational excitation events and single-point glitches or very short instantaneous fluctuations. Therefore, a preset threshold for the relative change of current upon event activation also needs to be set. And preset events enter continuous sample count Preset event entry current relative change threshold. Historical relative current change amplitudes are generated by calculating the historical operating data of the line where the target fitting is located according to the same relative change amplitude relationship. These amplitudes are then pre-determined based on daily small load fluctuations and the load change range used as an operating excitation. In this embodiment of the invention, the following settings are configured: There are no restrictions here; implementers can set them according to specific scenarios. Preset number of continuous samples upon event entry. The sampling period and the duration of the single-point or short-term current fluctuations to be excluded are predetermined. More frequent sampling results in a larger number of samples for the same actual duration; sparser sampling results in a smaller number of samples. In this embodiment, a preset number of samples is set for each event to enter a continuous period. There are no restrictions here; implementers can set them according to the specific scenario.

[0041] Based on preset events, the number of continuous samples is entered. Candidate event windows are sequentially acquired, starting at each sampling time after the stabilization window before the candidate event. Each candidate event window is then sequentially judged to determine whether it meets the excitation event condition. The excitation event condition is that the relative change amplitude of the current flowing through the target fitting at each sampling time within the candidate event window is greater than or equal to a preset threshold for the relative change of the current entering the event. The start time of the first candidate event window that meets the conditions for the stimulus event is recorded as the stimulus event entry time. At this time, it is determined that the stimulus event has occurred. Therefore, the stable window before the stimulus event entry time and the candidate event closest to the stimulus event entry time is obtained and recorded as the stable window before the stimulus event for this operation.

[0042] This scheme sets the excitation event to terminate only when the relative change amplitude does not exceed the event exit threshold and the event exit duration sample number is reached consecutively. A preset event exit current relative change threshold is set. and the number of samples that exited the preset event Preset event exit current relative change threshold and the number of samples that exited the preset event The setting is based on the aforementioned preset event entry current relative change threshold. And preset events enter continuous sample count The settings are based on a similar principle: a preset threshold for the relative change in current upon event exit. The relative change amplitude is calculated during the recovery process of the historical normal load changes of the line where the target fitting is located, and the level of this amplitude is predetermined based on the load recovery to the normal operating range. It can usually be smaller than To create hysteresis and reduce repeated entry and exit events near the boundary, this embodiment of the invention sets... There are no restrictions here; implementers can set them according to specific scenarios. Preset number of samples for event exit. Based on the desired short-term fallback duration and sampling period, and to avoid misjudging the event as ended when the current only briefly returns to the threshold, this embodiment of the invention sets a preset number of continuous samples for the event to enter. There are no restrictions here; implementers can set them according to the specific scenario.

[0043] Based on a preset number of continuous samples for event exit, candidate exit windows are sequentially obtained, starting from each sampling moment after the first candidate event window that meets the excitation event condition. Each candidate exit window is then sequentially assessed to determine whether it meets the excitation exit condition, which is that the relative change amplitude of the current flowing through the target hardware at each sampling moment within the candidate exit window is less than a preset event exit current relative change threshold. The end time of the first candidate exit window that meets the incentive exit condition is recorded as the incentive event exit time.

[0044] The window formed between the entry time of the stimulus event and the exit time of the stimulus event is denoted as the event action window of the running stimulus event.

[0045] After the expiration event, the window enters a response waiting state; the window length is restored after setting a preset event. Based on a preset post-event recovery window length, the exit time of the stimulus event is used as the start time to obtain the post-event recovery window for the running stimulus event. Specifically, the preset post-event recovery window length... The settings are predetermined based on the duration required for hardware of the same type and wire specifications as the target hardware to stabilize after a historical normal load change, as well as the sampling period, to cover the normal thermal inertia recovery process. In this embodiment of the invention, a preset recovery window length is set after the event. There are no restrictions here; implementers can set them according to the specific scenario.

[0046] Thus, the pre-event stability window, event impact window, and post-event recovery window of the running stimulus event have been obtained. Based on these windows, robust statistics for comparison can be formed, specifically including: (1) calculating the median of the current flowing through the target fitting at all sampling times within the pre-event stability window, as the pre-event line current benchmark; (2) calculating the median of the current flowing through the target fitting at all sampling times within the post-event recovery window, as the post-event line current benchmark; (3) calculating the median of the common response value at all sampling times within the pre-event stability window, as the pre-event common response benchmark; (4) calculating the median of the common response value at all sampling times within the post-event recovery window, as the post-event common response benchmark.

[0047] Because of the thermal inertia of the fittings, the recovery of the current flowing through the target fitting to the level before the operational excitation event does not necessarily mean that the fitting temperature has also recovered; the fitting may still be in the process of cooling down. If the temperature rise after the event is directly compared with the temperature rise before the event, it is easy to identify the lingering heat as a fault. Therefore, this solution sets three recovery conditions: First, determine whether the current itself has returned to a level close to that before the event; second, determine whether the common response value of the effective reference fittings of the target fitting has returned to a level close to that before the event, thereby determining whether the common thermal state of the entire line has recovered; third, determine whether the local response residual of the target fitting itself has stabilized, to avoid the target fitting still being in a process of continuous cooling or fluctuation. Only when the above three conditions are met simultaneously can the degradation of the target fitting be accurately analyzed, and thus the abnormal condition of the target fitting be accurately identified. Therefore, this operational excitation event is marked as an "evaluable operational excitation event". If the above three conditions still cannot be met after waiting for a period of time, or if there is a moment when the common response value fails during the event, then the risk status of the target hardware will not be updated using this running incentive event, and it will be marked as an "unevaluable running incentive event".

[0048] Based on the above description, the specific process for determining whether a runtime stimulus event is an evaluable runtime stimulus event is as follows: If there are sampling moments with a common empty response value within the pre-event stabilization window, event action window, and post-event recovery window of a running stimulus event, then the running stimulus event is determined to be an unevaluable running stimulus event. If there are no sampling moments with a common response value of empty within the pre-event stabilization window, event action window, and post-event recovery window of the excitation event, then the absolute value of the difference between the pre-event and post-event line current references is used as the numerator, and the maximum value between the pre-event line current reference and the preset minimum reference current is used as the denominator to obtain the relative difference of the current references before and after the event, denoted as . ,Right now In the formula, Indicates the line current reference after the event. Indicates the line current reference before the event. This indicates the preset minimum reference current. Represents the absolute value symbol. This indicates taking the maximum value; The following parameters are set: preset line current recovery threshold, preset common response recovery threshold, and preset residual stability threshold. The preset line current recovery threshold is set based on the range of relative differences between the current benchmarks before and after historically determined normal operating excitation events. The threshold can be the midpoint of the range. In this embodiment, the preset line current recovery threshold is set to 0.02, a dimensionless ratio, which is not limited here; implementers can set it according to specific scenarios. The preset common response recovery threshold is set based on the range of absolute differences between the common response benchmarks before and after historically determined normal operating excitation events. In this embodiment, the preset common response recovery threshold is set to 0.10, with the unit consistent with the common response value, which is not limited here; implementers can set it according to specific scenarios. The preset residual stability threshold is set based on the fluctuation range of the range of local response residuals within the recovery window after historically determined normal operating excitation events. In this embodiment, the preset residual stability threshold is set to 0.10, with the unit consistent with the local response residual, which is not limited here; implementers can set it according to specific scenarios. The evaluation criteria for the excitation event are as follows: the relative difference between the current reference before and after the event is less than or equal to the preset line current recovery threshold, the absolute value of the difference between the common response reference before the event and the common response reference after the event is less than or equal to the preset common response recovery threshold, and the range of the local response residuals at all sampling times within the recovery window after the event is less than or equal to the preset residual stability threshold. If the pre-event stabilization window and post-event recovery window of the running stimulus event meet the evaluation criteria for the running stimulus event, then the running stimulus event is determined to be an evaluable running stimulus event. If the pre-event stabilization window and post-event recovery window of the operational stimulus event do not meet the criteria for determining an evaluable operational stimulus event, the post-event recovery window is shifted backward by one sampling period to obtain a new post-event recovery window. The post-event line current reference, the post-event common response reference, the relative difference between the current references before and after the event, and the range of the local response residuals at all sampling times within the post-event recovery window are recalculated. If there are no sampling times with empty common response values ​​within the new post-event recovery window, and the pre-event stabilization window and the new post-event recovery window meet the criteria for determining an evaluable operational stimulus event, the new post-event recovery window replaces the post-event recovery window as the post-event recovery window, and the operational stimulus event is determined to be an evaluable operational stimulus event. If there is a sampling moment with a common response value of empty within the recovery window after the new event, or if the stable window before the event and the recovery window after the new event do not meet the criteria for determining an evaluable running stimulus event, then the recovery window after the new event is moved backward by one sampling period to obtain a new recovery window after the new event. This process continues until the total duration of the backward-moved sampling periods reaches the preset maximum recovery waiting time. If the criteria for determining an evaluable running stimulus event are still not met and there is no sampling moment with a common response value of empty, then the running stimulus event is determined to be an unevaluable running stimulus event.

[0049] This completes the identification of operational stimulus events and the judgment of evaluable operational stimulus events.

[0050] Step S104: Based on whether the operation stimulus event is an evaluable operation stimulus event and the difference in local response residuals between the target hardware before the event and after the event, a fault risk warning is issued for the target hardware.

[0051] If the current operational stimulus event is an evaluable operational stimulus event, then based on the difference in local response residuals between the target fitting before the event and the recovery window after the event, the degenerate memory state value of the target fitting after this operational stimulus event is obtained, which is used to detect the fault risk of the target fitting. The implementation process is as follows: (1) If the current operation stimulus event is an evaluable operation stimulus event, let the current operation stimulus event be the kth evaluable operation stimulus event of the target hardware. Then, based on the difference in local response residuals between the target hardware in the stable window before the event and the recovery window after the event, the event residual change of the kth evaluable operation stimulus event is obtained.

[0052] Specifically: Calculate the median of the local response residuals at all sampling times within the pre-event stability window and post-event recovery window of the k-th evaluable operational stimulus event, respectively. Use these medians as the pre-event response residual benchmark and the post-event response residual benchmark for the k-th evaluable operational stimulus event. Subtract the pre-event response residual benchmark from the post-event response residual benchmark to obtain the change in the event residual of the target hardware after the k-th evaluable operational stimulus event, denoted as […]. .

[0053] (2) Calculate the boundary of the residual change of normal events.

[0054] Even for normal hardware, the change in event residuals before and after a load change cannot be completely consistent; measurement errors and minor environmental changes can all cause fluctuations. Therefore, the presence of risk cannot be determined solely based on the change in event residuals. This solution collects valid reference hardware that has been re-inspected and confirmed to be normal, and whose type and conductor specifications are consistent with the target hardware. It then statistically analyzes the change in event residuals of these normal hardware after experiencing evaluable operational stimulus events, thus forming a "normal event residual change boundary." Based on this boundary, it determines whether the change in event residuals for the kth evaluable operational stimulus event is abnormal, thereby assessing the failure risk of the target hardware.

[0055] The steps to obtain the boundary of residual change for normal events are as follows: In the historical operation and maintenance records, the historical evaluable operation incentive events of the target hardware and its valid reference hardware that have been confirmed to be normal after re-inspection are obtained and recorded as historical normal operation incentive events. If the number of historical normal operation incentive events is less than the preset minimum number of normal samples, it is considered that an accurate and effective "normal event residual change boundary" cannot be determined. At this time, the event residual change of the kth evaluable operation incentive event is temporarily saved, and the degradation memory state value of the target hardware after the kth evaluable operation incentive event is set to the preset initial degradation memory state value. In this embodiment of the invention, the preset initial degradation memory state value is set to 0. There is no restriction here, and the implementer can set it according to the specific scenario. In this embodiment of the invention, the minimum number of preset normal samples is set to 9. There is no limitation here. Implementers can set it according to specific scenarios, but it is necessary to ensure that the minimum number of preset normal samples can be used to calculate the median and median absolute deviation. If the number of historical normal operation incentive events is greater than or equal to 9, the event residual change of each historical normal operation incentive event is obtained. The median and median absolute deviation of the event residual changes of all historical normal operation incentive events are calculated. A preset sensitivity coefficient is obtained. The preset sensitivity coefficient, the constant 1.4826, and the median absolute deviation are multiplied together to obtain a target value. The sum between the median of the event residual changes of all historical normal operation incentive events and the target value is calculated to obtain a sum result. The maximum value between the sum result and the constant 0 is selected as the boundary of normal event residual change.

[0056] In one embodiment, the formula for calculating the boundary of normal event residual variation is: , In the formula, B represents the boundary of the residual change during normal events. This represents the median of the changes in event residuals across all historical, normally functioning stimulus events. This indicates the preset sensitivity coefficient. Let m represent the change in event residual for the j-th historical normal operation stimulus event, and m represent the total number of all historical normal operation stimulus events. Represents the absolute value symbol. This indicates taking the median. This is the absolute deviation of the median. This indicates taking the maximum value.

[0057] Among them, for the preset sensitivity coefficient The settings include a preset sensitivity coefficient. The value ranges from 1.5 to 3.0, and is used to control the sensitivity to minor anomalies. A larger value allows for a wider range of normal fluctuations and a lower probability of false alarms, but also a higher probability of missed alarms. Conversely, a smaller value narrows the allowable range of normal fluctuations, increases the probability of false alarms, but decreases the probability of missed alarms. This value is determined based on historical normal operation stimulus events and existing re-inspection results. In this embodiment of the invention, a preset sensitivity coefficient is set. There are no restrictions here; implementers can set them according to the specific scenario.

[0058] (3) Based on the difference between the event residual change of the kth evaluable operational stimulus event and the boundary of the normal event residual change, obtain the degradation memory state value of the target hardware after the operational stimulus event.

[0059] The target difference is obtained by subtracting the normal event residual change boundary from the event residual change of the kth evaluable operational stimulus event. The maximum value between the target difference and the constant 0 is selected as the unrecovered response increment of the target fitting after the kth evaluable operational stimulus event.

[0060] In one embodiment, the formula for calculating the unrecovered response increment of the target fitting after the kth evaluable operational stimulus event is as follows: In the formula, This indicates that the target fitting did not recover its response increment after the kth evaluable operational stimulus event. B represents the change in the event residual of the target fitting after the kth evaluable operational stimulus event, and B represents the boundary of the normal event residual change. This indicates taking the maximum value.

[0061] The unrecovered response increment after the event is used to quantify the local temperature rise increase that remains and exceeds the normal event fluctuation range after the event is recovered. Its dimension is consistent with the local response residual of the target fitting.

[0062] If there is no evaluable operational stimulus event before the kth evaluable operational stimulus event, i.e., k=1, then the sum of the unrecovered response increment after the event and the preset initial degradation memory state value is taken as the degradation memory state value of the target hardware at the kth evaluable operational stimulus event. If the target hardware has at least one evaluable operational stimulus event before the kth evaluable operational stimulus event, then the post-event unrecovered response increment of the target hardware after all evaluable operational stimulus events before the kth evaluable operational stimulus event is obtained using the normal event residual change boundary, and recorded as the post-event unrecovered response increment. The sum of the preset initial degradation memory state value, the post-event unrecovered response increment of all events, and the post-event unrecovered response increment of the kth evaluable operational stimulus event is calculated to obtain the degradation memory state value of the target hardware after the kth evaluable operational stimulus event.

[0063] In one embodiment, the formula for calculating the degradation memory state value of the target hardware after the kth evaluable operational stimulus event is as follows: In the formula, This represents the degradation memory state value of the target hardware after the kth evaluable operational stimulus event. This represents the degradation memory state value of the target hardware after the kth evaluable operational stimulus event. This indicates that the target hardware did not recover its response increment after the kth evaluable operational stimulus event. The preset initial degradation memory state value is denoted as... , That is, when k=1, .

[0064] It is worth noting that when the historical normal operation stimulus events are updated, or when the number of historical normal operation stimulus events reaches the preset minimum number of normal samples for the first time, the normal event residual change boundary should be recalculated using the event residual change of the updated historical normal operation stimulus events. Then, using the recalculated normal event residual change boundary, the event residual change of the system stored during the period when the number of historical normal operation stimulus events has not reached the preset minimum number of normal samples should be calculated. The unrecovered response increment after the event is recalculated and re-accumulated to form the degradation memory state value.

[0065] (4) If the current operational stimulus event is an unevaluable event, the degradation memory state value of the target hardware remains the degradation memory state value under the operational stimulus event prior to the current operational stimulus event. If there are no evaluable operational stimulus events prior to the current operational stimulus event, the preset initial degradation memory state value remains. In this embodiment of the invention, the preset initial degradation memory state value is 0.

[0066] Thus, the degradation memory state value of the target hardware after this operational stimulus event was obtained. Furthermore, based on the degradation memory state value of the target hardware after this operational stimulus event, a fault risk warning was issued for the target hardware, specifically:

[0067] A first preset risk grading threshold and a second preset risk grading threshold are set. The first preset risk grading threshold is used to upgrade from routine monitoring to enhanced monitoring, and the second preset risk grading threshold is used to generate re-inspection tasks. Therefore, the second preset risk grading threshold is greater than the first preset risk grading threshold. Historical operating data of valid reference hardware of the same type and wire specifications as the target hardware are obtained. Historical degradation memory state values ​​are calculated according to the above processing flow. These historical degradation memory state values ​​are then correlated with the corresponding actual maintenance and re-inspection results and existing operation and maintenance rules to determine the first preset risk grading threshold and the second preset risk grading threshold. In this embodiment of the invention, the first preset risk grading threshold is set to 0.50, and the second preset risk grading threshold is set to 1.20. This is not a limitation, and implementers can set them according to specific scenarios.

[0068] If the degradation memory state value of the target hardware after this operational incentive event is less than the preset first risk classification threshold, the target hardware will be subject to routine monitoring. If the degradation memory state value of the target hardware after this operational incentive event is greater than or equal to the preset first risk classification threshold and less than the preset second risk classification threshold, then the target hardware will be added to the key monitoring set for key monitoring. If the degradation memory state value of the target hardware after this operational incentive event is greater than or equal to the second risk classification threshold, a risk fault warning will be issued for the target hardware, and a re-inspection task will be generated.

[0069] Routine monitoring refers to the continuous monitoring of the target fittings according to the existing daily operation and maintenance methods of the line. The system normally collects and processes the temperature of the target fittings, the ambient temperature, and the line current of the current-carrying path where the target fittings are located. Routine inspections such as infrared thermography and visible light inspection are carried out according to the established inspection cycle without increasing the frequency of manual verification or the priority of on-site inspection of the target fittings.

[0070] Key monitoring refers to increasing the attention and inspection priority of target hardware in the operation and maintenance platform, increasing the frequency of manual verification of the temperature rise of target hardware, local response residuals of target hardware, and the trend of degradation memory state value changes, and appropriately shortening the infrared retest or on-site inspection interval of the target hardware according to the existing line operation and maintenance schedule; during the enhanced monitoring phase, power outage maintenance tasks are not directly generated. When the degradation memory state value further reaches the second risk level threshold, a re-inspection task will be generated according to this plan.

[0071] The re-inspection task must include at least the hardware equipment identification, installation location, degradation memory state value, warning time, and re-inspection type. The warning time is the time when the degradation memory state value first reaches the second risk level threshold after being updated or recalculated. The re-inspection type can be configured according to existing operation and maintenance resources, including one or more of infrared re-testing, visible light inspection, and power outage connection status inspection. The re-inspection task is sent to the corresponding line's operation and maintenance task queue through the operation and maintenance platform task interface, so that the calculation result of the degradation memory state value directly changes the actual re-inspection schedule of the target hardware. Routine monitoring, key monitoring, and re-inspection are existing technologies and will not be elaborated here.

[0072] The following set of illustrative data illustrates the continuous operation process, starting from the target fitting temperature, reference fitting temperature, ambient temperature, and line current, through the formation of a valid reference fitting set, the determination of operational excitation event recovery, the calculation of post-event unrecovered response increments, and the updating of degradation memory states, until the generation of the re-inspection task. The listed values ​​are illustrative, and the displayed results are rounded to two decimal places.

[0073] The tension clamp with the hardware identifier G0 is selected as the target hardware. Within the topological neighborhood of the same current-carrying path of the line, tension clamps with the same hardware type and conductor specifications as the target hardware are selected as candidate reference hardware. The preset minimum number of reference hardware is set to 3, the preset discrete scale threshold is 0.50, the preset pre-event stability window length and the preset post-event recovery window length are both set to 5, the preset pre-event current stability threshold is 10A, the preset minimum reference current is 50A, the preset event-entry current relative change threshold is 0.15, the preset event-exit current relative change threshold is 0.05, the preset event-entry duration sample number and the preset event-exit duration sample number are both 3, the preset line current recovery threshold is 0.02, the preset common response recovery threshold is 0.10, the preset residual stability threshold is 0.10, the preset minimum number of normal samples is set to 9, and the preset maximum recovery waiting time is 10 sampling periods. The preset sensitivity coefficient is then obtained. The threshold is set to 2.0, the first risk classification threshold is set to 0.50, and the second risk classification threshold is set to 1.20.

[0074] The event residual changes for the nine historical normal operation stimulus events were -0.10, -0.05, 0.00, 0.00, 0.05, 0.05, 0.10℃, 0.10, and 0.15, respectively. The median of the event residual changes for the historical normal operation stimulus events was also mentioned. The median absolute deviation is 0.05, and the median absolute deviation multiplied by 1.4826 is 0.07413. Substituting these values ​​into the normal event residual variation boundary calculation relationship, the normal event residual variation boundary B is obtained as 0.19826.

[0075] Before the first operational excitation event of the target fitting, the five currents flowing through the target fitting within the pre-event stabilization window are 518A, 520A, 521A, 519A and 520A respectively, with the difference between the maximum and minimum values ​​being 3A. The pre-event current stabilization threshold is no greater than 10A, and the pre-event line current reference is 520A. At a sampling moment within the pre-event stabilization window, the ambient temperature is 30.00℃, the target hardware temperature is 53.02℃, and the temperatures of the five valid reference hardwares are 52.10℃, 52.30℃, 52.40℃, 52.50℃, and 52.70℃, respectively. The median of the temperature rise response of the five reference hardwares is 22.40℃, and the reference dispersion scale is 0.14826, which is less than the preset dispersion scale threshold of 0.50. Therefore, the common response value at this sampling moment is 22.40℃, the temperature rise response of the target hardware is 23.02℃, and the local response residual of the target hardware is 0.62℃.

[0076] The common response under the same operating conditions at five sampling times within the pre-event stability window were 22.35℃, 22.40℃, 22.38℃, 22.45℃, and 22.42℃, respectively, with a median of 22.40℃. The local response residuals of the target hardware at the five sampling times were 0.60℃, 0.62℃, 0.61℃, 0.63℃, and 0.64℃, respectively, with a median of 0.62℃.

[0077] In the first operational stimulus event, the current flowing through the target fitting was 600A, 650A, and 690A at three consecutive sampling times, respectively. The relative change amplitudes of the currents relative to the pre-event line current reference of 520A were 0.15385, 0.25000, and 0.32692, respectively, all exceeding 0.15, and the system entered the event activation state. When the event ended, the current flowing through the target fitting was 544A, 540A, and 523A at three consecutive sampling times, respectively, with relative change amplitudes of 0.04615, 0.03846, and 0.00577, respectively, all returning to within 0.05, and the system entered the recovery waiting state. The five line currents within the post-event recovery window were 523A, 522A, 524A, 521A, and 523A, with a median of 523A. The relative difference between this and 520A was 0.00577, which is less than the preset line current recovery threshold of 0.02. The common response values ​​at the five sampling times within the post-event recovery window were 22.42℃, 22.45℃, 22.46℃, 22.44℃, and 22.47℃, with a median of 22.45℃. This differs from the pre-event common response benchmark of 22.40℃ by 0.05℃. The local response residuals of the five target fittings within the post-event recovery window were 0.69℃, 0.72℃, 0.74℃, 0.71℃, and 0.73℃, with a median of 0.72℃. The difference between the maximum and minimum values ​​was 0.05℃. The conditions for line current recovery, common response recovery, and local response residual stability of the target fitting are all met. This operational excitation event constitutes an evaluable operational excitation event, and the event residual change is... The value was 0.10℃. Since 0.10℃ did not exceed the normal event residual variation boundary of 0.19826℃, the response increment was not recovered after the event. A value of 0 indicates a degenerate memory state. Keep it at 0.

[0078] In the second evaluable operational stimulus event, the pre-event line current reference was 535A, and the post-event line current reference was 532A, with a relative difference of 0.00561. The pre-event common response reference was 23.10℃, and the post-event common response reference was 23.16℃, with a difference of 0.06℃. The range of the local response residual of the target fitting within the post-event recovery window was 0.06℃, and all three recovery conditions were met. The pre-event response residual reference was 0.66℃, and the post-event response residual reference was 1.08℃, with the event residual change... The temperature was 0.42℃, and no recovery increment occurred after the event. The value is 0.22174℃, representing the degradation memory state value. The value is 0 + 0.22174℃ = 0.22174℃, which is displayed as 0.22℃ with two decimal places. This is lower than the preset first risk level threshold, so the target hardware will continue to be monitored normally.

[0079] In the third evaluable operational stimulus event, the pre-event line current reference was 548A, and the post-event line current reference was 551A, with a relative difference of 0.00547. The pre-event common response reference was 24.00℃, and the post-event common response reference was 24.04℃, with a difference of 0.04℃. The range of the local response residual of the target fitting within the post-event recovery window was 0.07℃, and all three recovery conditions were met. The pre-event response residual reference was 0.71℃, and the post-event response residual reference was 1.45℃, with the event residual change... The temperature was 0.74℃, and no recovery increment occurred after the event. The value is 0.54174℃, representing the degradation memory state value. The value is 0.22174℃ + 0.54174℃ = 0.76348℃, which is displayed as 0.76℃ after rounding to two decimal places. This value is between the preset first risk classification threshold and the preset second risk classification threshold. Therefore, the hardware equipment identifier G0 is added to the key monitoring set.

[0080] In the fourth evaluable operational stimulus event, the pre-event line current reference was 560A, and the post-event line current reference was 557A, with a relative difference of 0.00536. The pre-event common response reference was 24.60℃, and the post-event common response reference was 24.66℃, with a difference of 0.06℃. The range of the local response residual of the target fitting within the recovery window after the event was 0.08℃, and all three recovery conditions were met. The pre-event response residual reference was 0.78℃, and the post-event response residual reference was 1.58℃, with the event residual change... The temperature was 0.80℃, and no recovery increment occurred after the event. The value is 0.60174℃, representing the degradation memory state value. The temperature is 0.76348℃ + 0.60174℃ = 1.36522℃, which is displayed as 1.37℃ to two decimal places, reaching the preset second risk level threshold. The fault risk warning module generates a re-inspection task containing the hardware equipment identifier G0, installation location, degradation memory state value, warning time, and re-inspection type, and sends it to the operation and maintenance platform task queue.

[0081] If, after re-inspection, the target hardware corresponding to the hardware equipment identifier G0 is confirmed to be in normal condition, the event residual changes of the first to fourth evaluable operational stimulus events (0.10℃, 0.42℃, 0.74℃, and 0.80℃) are added to the historical normal event residual change sample. After increasing the historical normal event residual change sample to 13, the median is 0.10℃, the median absolute deviation is 0.10℃, and the normal event residual change boundary is recalculated to be 0.39652℃. Using the normal event residual change boundary of 0.39652℃, the unrecovered response increments after the first to fourth evaluable operational stimulus events are recalculated to be 0, 0.02348℃, 0.34348℃, and 0.40348℃, respectively, and re-accumulated to obtain the degraded memory state value of 0.77044℃. Since 0.77044℃ is not less than the first risk classification threshold and less than the second risk classification threshold, the target hardware is treated according to enhanced monitoring. Therefore, after the re-examination confirms that the new normal event data is normal, it enters the dynamic update of the normal event residual change boundary, and the degenerate memory state value during the current period of accumulated degenerate memory state is recalculated according to the same normal event residual change boundary.

[0082] After updating the boundary of normal event residual changes and recalculating the degradation memory state value, if another line current increase event occurs, but any one of the following conditions—line current recovery condition, common response recovery condition, or local response residual stability condition of the target fitting—is still not met within the maximum recovery waiting time, then this operational stimulus event is marked as an unevaluable operational stimulus event. The system does not generate a post-event recovery response increment corresponding to this operational stimulus event, and the degradation memory state value in the state database remains at 0.77044℃. Subsequent new operational stimulus events that meet the evaluation conditions will continue to be updated.

[0083] As can be seen from the above illustrative operation process, the local response residuals of the target fittings, utilizing the common response value formed by effective reference fittings, mitigate the impact of common load changes. Furthermore, the recovery conditions of the operational excitation event limit the comparison points before and after the event, ultimately only writing positive increments exceeding the historical normal event residual change boundaries into the degradation memory state. The resulting degradation memory state value reflects the cumulative trend of newly added unrecovered local responses during the monitoring period and continuously drives the three operational states: routine monitoring, enhanced monitoring, and re-inspection tasks. This improves the stability and accuracy of early warning for fitting failure risks.

[0084] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A smart early warning system for power fitting fault risks based on big data operation, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: Based on the difference between the temperature of each fitting and the ambient temperature at each sampling time, the temperature rise response value of each fitting at each sampling time is obtained. Any fitting is recorded as the target fitting. Based on the device identification of the target fitting and the topology of its line, as well as the temperature rise response value of each fitting at each sampling time, the effective reference fitting of the target fitting at each sampling time is obtained. Based on the number of valid reference fittings and the discrete distribution of temperature rise response values ​​at each sampling time, the common response value at each sampling time is obtained. Based on the difference between the temperature rise response value of the target fitting and the common response value at each sampling time, the local response residual of the target fitting at each sampling time is obtained. Based on the change in current flowing through the target fitting, an operational stimulus event is identified, and a pre-event stability window, an event action window, and a post-event recovery window are established for the operational stimulus event. Based on the current change in the line where the target fitting is located, the common response value of the target fitting, and the recovery status of the local response residual within the post-event recovery window, it is determined whether the operational stimulus event is an evaluable operational stimulus event. Based on whether the operational stimulus event is an evaluable operational stimulus event, and the difference in local response residuals between the target hardware before the event and after the event, a fault risk warning is issued for the target hardware.

2. The intelligent early warning system for power fitting fault risk based on operational big data as described in claim 1, characterized in that, The acquisition of the temperature rise response value of each fitting at each sampling time includes: For any sampling time, if neither the ambient temperature nor the hardware temperature of any hardware is marked as invalid, then the ambient temperature is subtracted from the hardware temperature of any hardware at any sampling time to obtain the temperature rise response value of any hardware at any sampling time. If the ambient temperature or the temperature of any hardware at any sampling time is marked as invalid, then the temperature rise response value of any hardware at any sampling time is set to empty.

3. The intelligent early warning system for power fitting fault risk based on operational big data as described in claim 2, characterized in that, The method for obtaining the effective reference hardware for the target hardware at each sampling time includes: Based on the topology of the line where the target fitting is located, other fittings with the same fitting type and conductor specifications as the target fitting are obtained within the topological neighborhood of the current-carrying path of the same line as the target fitting, and are recorded as the reference fittings of the target fitting. For any sampling time, the reference fitting with a non-empty temperature rise response value is recorded as the valid reference fitting for the target fitting at that sampling time.

4. The intelligent early warning system for power fitting fault risk based on operational big data as described in claim 1, characterized in that, The process of obtaining the common response value at each sampling time includes: For any sampling time, calculate the median absolute deviation of the temperature rise response values ​​of all valid reference hardware of the target hardware at that sampling time, and multiply the median absolute deviation by a constant 1.4826 to obtain the reference discrete scale at that sampling time. If the number of valid reference hardware for the target hardware at any sampling time is greater than or equal to the preset minimum number of reference hardware, and the reference discrete scale at any sampling time is less than or equal to the preset discrete scale threshold, then the median of the temperature rise response values ​​of all valid reference hardware for the target hardware at any sampling time is obtained as the common response value at any sampling time. If the number of valid reference fittings for the target fitting at any sampling time is less than the preset minimum number of reference fittings, or the reference discrete scale at any sampling time is greater than the preset discrete scale threshold, then the common response value at any sampling time is set to empty.

5. The intelligent early warning system for power fitting fault risk based on operational big data as described in claim 4, characterized in that, The process of obtaining the local response residual of the target hardware at each sampling time includes: For any sampling time, if the common response value at any sampling time is not empty, then the temperature rise response value of the target hardware at any sampling time is subtracted from the common response value at any sampling time to obtain the local response residual of the target hardware at any sampling time. If the common response value at any sampling time is empty, then the local response residual of the target hardware at any sampling time is set to empty.

6. The intelligent early warning system for power fitting fault risk based on operational big data as described in claim 1, characterized in that, The step of identifying operational excitation events based on changes in current flowing through the target fitting, and establishing a pre-event stability window, an event action window, and a post-event recovery window for the operational excitation event, includes: Obtain the preset pre-event stable window length and the preset pre-event current stabilization threshold. Based on the preset pre-event stable window length, obtain candidate stable windows with each sampling time as the starting time. Calculate the range of the current flowing through the target fitting within each candidate stable window. Record the candidate stable window with a range less than the preset pre-event current stabilization threshold as a stable window. According to the starting time corresponding to each candidate stable window, determine whether each candidate stable window is a stable window in turn. If there is an unstable window, record the stable window before the unstable window as a candidate pre-event stable window. Obtain the median of the current flowing through the target fitting at all sampling times within the stabilization window before the candidate event, and use it as the baseline for the line current before the event. For any sampling time after the stabilization window before the candidate event, calculate the relative change amplitude of the current flowing through the target fitting at that sampling time based on the absolute value of the difference between the current flowing through the target fitting at that sampling time and the baseline for the line current before the event. The calculation formula is as follows: ， In the formula, This represents the relative change amplitude of the current flowing through the target fitting at any given sampling time. This represents the current flowing through the target fitting at any given sampling time. Indicates the line current reference before the event. This indicates the preset minimum reference current. Represents the absolute value symbol. This indicates taking the maximum value; The relative change amplitude of the current flowing through the target hardware is obtained sequentially at each sampling time after the stabilization window before the candidate event. The number of continuous samples for the preset event entry and the relative change threshold of the current entering the preset event are obtained. Based on the number of continuous samples for the preset event entry, candidate event windows are obtained sequentially with each sampling time after the stabilization window before the candidate event as the starting time. It is then determined whether each candidate event window meets the excitation event condition. The excitation event condition is that the relative change amplitude of the current flowing through the target hardware at each sampling time within the candidate event window is greater than or equal to the relative change threshold of the current entering the preset event. The starting time of the first candidate event window that meets the excitation event condition is recorded as the excitation event entry time. The stable window before the arrival time of the stimulus event is obtained and is closest to the arrival time of the stimulus event. This window is denoted as the stable window before the event of the running stimulus event. Obtain a preset event exit current relative change threshold and a preset event exit duration sample number. Based on the preset event exit duration sample number, sequentially obtain candidate exit windows starting from each sampling time after the first candidate event window that meets the excitation event condition. Sequentially determine whether each candidate exit window meets the excitation exit condition. The excitation exit condition is that the relative change amplitude of the current flowing through the target hardware at each sampling time within the candidate exit window is less than the preset event exit current relative change threshold. Record the end time of the first candidate exit window that meets the excitation exit condition as the excitation event exit time. The window formed between the entry time of the stimulus event and the exit time of the stimulus event is denoted as the event action window of the running stimulus event; Obtain the preset event recovery window length, and based on the preset event recovery window length, take the exit time of the stimulus event as the start time to obtain the event recovery window for the running stimulus event.

7. The intelligent early warning system for power fitting fault risk based on operational big data as described in claim 1, characterized in that, The step of determining whether the operational stimulus event is an evaluable operational stimulus event includes: If there is a sampling moment with a common empty response value within the pre-event stabilization window, event action window, and post-event recovery window of the running stimulus event, then the running stimulus event is determined to be an unevaluable running stimulus event. If there are no sampling moments with a common response value of empty within the pre-event stabilization window, event action window, and post-event recovery window of the running excitation event, then the median of the current flowing through the target fitting at all sampling moments within the pre-event stabilization window is calculated as the pre-event line current reference. The median of the current flowing through the target fitting at all sampling moments within the post-event recovery window is calculated as the post-event line current reference. The absolute value of the difference between the pre-event line current reference and the post-event line current reference is used as the numerator, and the maximum value between the pre-event line current reference and the preset minimum reference current is used as the denominator to obtain the relative difference of the current reference before and after the event. Calculate the median of the common response values ​​at all sampling times within the pre-event stability window, and use it as the pre-event common response benchmark; Calculate the median of the common response values ​​at all sampling times within the post-event recovery window, and use it as the common response benchmark after the event; Obtain a preset line current recovery threshold, a preset common response recovery threshold, and a preset residual stability threshold. Set the evaluation criteria for the excitation event as follows: the relative difference between the current reference before and after the event is less than or equal to the preset line current recovery threshold, the absolute value of the difference between the common response reference before the event and the common response reference after the event is less than or equal to the preset common response recovery threshold, and the range of the local response residuals at all sampling times within the recovery window after the event is less than or equal to the preset residual stability threshold. If the pre-event stabilization window and post-event recovery window of the running stimulus event meet the evaluable running stimulus event determination criteria, then the running stimulus event is determined to be an evaluable running stimulus event. If the pre-event stabilization window and post-event recovery window of the operational stimulus event do not meet the criteria for evaluating an operational stimulus event, the post-event recovery window is shifted backward by one sampling period to obtain a new post-event recovery window. Based on the new post-event recovery window, the post-event line current reference, the post-event common response reference, the relative difference between the current references before and after the event, and the range of the local response residuals at all sampling moments within the post-event recovery window are recalculated. If there are no sampling moments with empty common response values ​​within the new post-event recovery window, and the pre-event stabilization window and the new post-event recovery window meet the criteria for evaluating an operational stimulus event, the new post-event recovery window replaces the post-event recovery window as the post-event recovery window, and the operational stimulus event is determined to be an evaluable operational stimulus event. If there is a sampling moment with a common response value of empty within the recovery window after the new event, or if the stable window before the event and the recovery window after the new event do not meet the criteria for determining an evaluable operational stimulus event, then the recovery window after the new event is moved backward by one sampling period to obtain a new recovery window after the event, until the total duration of the backward sampling periods reaches the preset maximum recovery waiting time. If the criteria for determining an evaluable operational stimulus event are still not met, and the stable window before the event and the recovery window after the new event do not meet the criteria for determining an evaluable operational stimulus event, then the operational stimulus event is determined to be an unevaluable operational stimulus event.

8. The intelligent early warning system for power fitting fault risk based on operational big data as described in claim 7, characterized in that, The aforementioned fault risk warning for the target hardware includes: If the operational stimulus event is an evaluable operational stimulus event, the median of the local response residuals at all sampling times within the pre-event stability window and the post-event recovery window of the operational stimulus event is calculated respectively, and used as the pre-event response residual benchmark and the post-event response residual benchmark of the operational stimulus event. The pre-event response residual benchmark is subtracted from the post-event response residual benchmark to obtain the change in the event residual of the target hardware after the operational stimulus event. In the historical operation and maintenance records, obtain the historical evaluable operation incentive events of the target hardware and its valid reference hardware that have been confirmed to be normal after re-inspection, and record them as historical normal operation incentive events. If the number of historical normal operation incentive events is less than the preset minimum number of normal samples, save the event residual change of the operation incentive event, and set the degradation memory state value of the target hardware after the operation incentive event to the preset initial degradation memory state value. If the number of historical normal operation incentive events is greater than or equal to the preset minimum number of normal samples, the event residual change of each historical normal operation incentive event is obtained, the median and median absolute deviation of the event residual changes of all historical normal operation incentive events are calculated, a preset sensitivity coefficient is obtained, the preset sensitivity coefficient, the constant 1.4826 and the median absolute deviation are multiplied together to obtain the target value, the sum between the median of the event residual changes of all historical normal operation incentive events and the target value is calculated, and the sum result is obtained. The maximum value between the sum result and the constant 0 is selected as the boundary of normal event residual change. Based on the difference between the event residual change of the operational stimulus event and the boundary of the normal event residual change, the degradation memory state value of the target hardware after the operational stimulus event is obtained; If the operational stimulus event is an unevaluable operational stimulus event, and there is an evaluable operational stimulus event before the operational stimulus event, then the degradation memory state value of the target hardware after the most recent evaluable operational stimulus event is obtained, and used as the degradation memory state value of the target hardware after the operational stimulus event. If the running stimulus event is an unevaluable running stimulus event, and there is no evaluable running stimulus event before the running stimulus event, then the degradation memory state value of the target hardware after the running stimulus event is set to a preset initial degradation memory state value. Based on the degradation memory state value of the target hardware after the operational excitation event, a fault risk warning is issued for the target hardware.

9. The intelligent early warning system for power fitting fault risk based on operational big data as described in claim 8, characterized in that, The acquisition of the degradation memory state value of the target hardware after the operational stimulus event includes: The target difference is obtained by subtracting the normal event residual change boundary from the event residual change of the running stimulus event. The maximum value between the target difference and the constant 0 is selected as the unrecovered response increment of the target hardware after the event of the running stimulus event. If there is no evaluable operational stimulus event before the operational stimulus event, the sum of the unrecovered response increment after the event and the preset initial degradation memory state value is taken as the degradation memory state value of the target hardware after the operational stimulus event. If there is at least one evaluable operational stimulus event before the operational stimulus event, the post-event unrecovered response increment of the target hardware after all evaluable operational stimulus events before the operational stimulus event is obtained using the normal event residual change boundary. This is recorded as the post-event unrecovered response increment. The sum of the preset initial degradation memory state value, all post-event unrecovered response increments, and the post-event unrecovered response increment of the operational stimulus event is calculated to obtain the degradation memory state value of the target hardware after the operational stimulus event.

10. The intelligent early warning system for power fitting fault risk based on operational big data as described in claim 8, characterized in that, The method of providing fault risk warning for the target hardware based on the degradation memory state value of the target hardware after the operational stimulus event includes: Obtain a preset first risk classification threshold and a preset second risk classification threshold, wherein the preset second risk classification threshold is greater than the preset first risk classification threshold; If the degradation memory state value of the target hardware after the operation stimulus event is less than the preset first risk classification threshold, then the target hardware will be subject to routine monitoring. If the degradation memory state value of the target hardware after the operation stimulus event is greater than or equal to the preset first risk classification threshold and less than the preset second risk classification threshold, then the target hardware will be added to the key monitoring set and the target hardware will be monitored. If the degradation memory state value of the target hardware after the running stimulus event is greater than or equal to the second risk classification threshold, a risk fault warning is issued for the target hardware, and a re-inspection task is generated.

Images