Fault prediction system based on electrical behavior feature mapping

Abstract

The invention discloses a fault prediction system based on electrical behavior feature mapping, and relates to the technical field of fault prediction. Comprising an electrical behavior data acquisition and processing module used for acquiring electrical behavior data in real time, performing data preprocessing and extracting electrical behavior characteristics; the phase coupling integrity evaluation module is used for extracting cross-phase phase difference deviation and fluctuation characteristics, forming a coupling integrity index and judging a phase coupling steady state; the broadband disturbance feature recognition module is used for extracting disturbance enhancement features, forming an arc disturbance index and judging the arc steady state; and the early-stage arc evolution evaluation module is used for fusing the coupling integrity index and the arc disturbance index, constructing an arc joint deviation amount, forming an early-stage evolution score in a window, judging an arc evolution stage and generating intervention measures. The problems that traditional monitoring cannot capture early hidden coupling damage of the arc and evolution of the arc from the early stage to the development stage is difficult to recognize in time are solved.

Description

Technical Field

[0001] This invention relates to the field of fault prediction technology, specifically a fault prediction system based on electrical behavior feature mapping. Background Technology

[0002] In various power transmission and distribution systems, industrial power systems, and electrical equipment, the electrical operating status directly affects equipment safety, energy efficiency, and power supply stability. With the increasing complexity of electrical equipment structures and the continuous growth of electrical loads, the electrical behavior within systems is becoming more diverse and dynamic, and various electrical faults are exhibiting increasingly concealed forms and more complex evolution processes. Therefore, monitoring, identifying, and predicting electrical characteristics has become an important direction for ensuring the reliable operation of power systems, and is of great significance for improving equipment safety management capabilities and reducing outages.

[0003] For example, invention patent CN114839456B discloses a device and method for diagnosing and predicting aircraft electrical circuit faults. The device includes: a remote power controller (RPC) control engine based on a digital signal processor and a power line interface module. The power line interface module transmits the excitation signal generated by the RPC control engine to the aircraft electrical circuit. The response signal generated when the aircraft electrical circuit is excited is transmitted to the RPC control engine through the power line interface module. The RPC control engine is used to diagnose and predict aircraft electrical circuit faults. This application eliminates the need for numerous sensors and the need to transmit, collect, and process large amounts of data from these sensors, making the diagnosis and prediction of aircraft electrical circuit faults more reliable, economical, and practical.

[0004] For example, the invention patent with publication number CN119716355A discloses an electrical equipment fault prediction system and method based on data fusion. This system accurately captures the electromagnetic signal characteristic parameters of the operating environment of electrical equipment, scientifically evaluates the electromagnetic interference index, and thus intelligently adjusts the data acquisition strategy, effectively improving the accuracy and efficiency of data acquisition. At the same time, by monitoring the operating status parameters of electrical equipment, it can also deeply extract its characteristics and calculate the data acquisition compliance index, providing a strong basis for data preprocessing. Finally, by highly integrating the multi-dimensional operating status parameters of electrical equipment, it can accurately identify operating degradation factors, achieve accurate prediction of the fault level of electrical equipment, and issue timely warnings. This not only significantly improves the accuracy and timeliness of fault prediction, but also effectively reduces operation and maintenance costs, enhances the stability and security of the power system, and provides strong technical support for the intelligent management and maintenance of electrical equipment.

[0005] However, during the long-term operation of existing electrical systems, some practical challenges related to early arc detection have gradually emerged: early arcs do not immediately cause a large current increase during their generation stage, but rather appear as weak signs such as phase noise diffusion, decreased cross-phase coherence, and enhanced broadband micro-disturbances. However, these arc precursors exhibit asynchronous diffusion paths across the three phases. Traditional monitoring methods based on single phases or fixed windows cannot capture the process of phase coupling relationship disruption, leading to the system's inability to identify the weak correlation between "early arc development and arc progression." This means that arc hazards may only be detected when they reach a significant stage, missing early warning opportunities.

[0006] Therefore, in order to address the above problems, there is an urgent need for a fault prediction system based on electrical behavior feature mapping. Summary of the Invention

[0007] Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a fault prediction system based on electrical behavior feature mapping, which solves the problem that traditional monitoring cannot capture early-stage implicit coupling damage of electric arcs and is difficult to identify the evolution of electric arcs from the early stage to the development stage in a timely manner.

[0008] Technical solution To achieve the above objectives, the present invention provides the following technical solution: a fault prediction system based on electrical behavior feature mapping, comprising: an electrical behavior data acquisition and processing module, used to acquire electrical behavior data in real time, preprocess the electrical behavior data, and extract electrical behavior features based on the preprocessed electrical behavior data; a phase coupling integrity assessment module, used to extract phase difference deviation and fluctuation characteristics across phases based on electrical behavior features; combine phase difference deviation and fluctuation characteristics with historical baselines to form a coupling integrity index; determine phase coupling steady state based on the coupling integrity index; a broadband disturbance feature identification module, used to extract disturbance enhancement features from electrical behavior features and form an arc disturbance index with historical baselines; determine arc steady state based on the arc disturbance index; and an early arc evolution assessment module, used to fuse the deviation of the coupling integrity index from the phase coupling steady state and the amplification features of the arc disturbance index from the arc steady state to construct a joint arc deviation, and form an early evolution score within a sliding window; determine the arc evolution stage based on the early evolution score, and generate corresponding intervention measures according to different stages.

[0009] Furthermore, the process of real-time acquisition of electrical behavior data, data preprocessing of the electrical behavior data, and extraction of electrical behavior features based on the preprocessed electrical behavior data is as follows: Real-time acquisition of electrical behavior data, including three-phase power frequency current, three-phase power frequency voltage, three-phase broadband current, and zero-sequence current; wherein the three phases include phase A, phase B, and phase C; time alignment of the electrical behavior data using the same sampling clock, and execution of DC bias removal and power frequency synchronization filtering to eliminate steady-state offset and electromagnetic background noise; low-pass filtering of the three-phase power frequency current and three-phase power frequency voltage to obtain their respective stable three-phase power frequency components, and band-pass filtering of the three-phase broadband current to extract the three-phase... Non-power frequency broadband components are processed, and outlier removal and undersampling compensation are performed on the filtered electrical behavior data. Z-score normalization is applied to the electrical behavior data. Based on a fixed sliding time window, the RMS values ​​of the three-phase power frequency current, three-phase power frequency voltage, and zero-sequence current are calculated to obtain the effective values ​​of the three-phase power frequency current, three-phase power frequency voltage, and zero-sequence current. The square integral of the three-phase broadband current within the non-power frequency band is calculated to obtain the energy of the three-phase broadband current. The instantaneous phase angle of the three-phase power frequency is obtained through Hilbert transform based on the three-phase power frequency voltage. An electrical fault prediction database is established, and the original and preprocessed electrical behavior data are written into the database.

[0010] Furthermore, based on electrical behavior characteristics, the specific process for extracting the phase difference deviation and fluctuation characteristics across phases is as follows: Based on a sliding time window, the instantaneous phase angles of the three-phase power frequency are obtained. The instantaneous phase difference of phase AB is obtained by calculating the phase difference between phase A and phase B in the instantaneous phase angles of the three-phase power frequency. The instantaneous phase difference of phase BC is obtained by calculating the phase difference between phase B and phase C. The instantaneous phase difference of phase CA is obtained by calculating the phase difference between phase C and phase A. The instantaneous phase difference sequences of phase AB, phase BC, and phase CA within the window are obtained. The average value and standard deviation of the three instantaneous phase difference sequences within the window are calculated respectively to obtain the average phase difference of phase AB, the average phase difference of phase BC, the average phase difference of phase CA, the standard deviation of phase difference of phase AB, the standard deviation of phase difference of phase BC, and the standard deviation of phase difference of phase CA.

[0011] Furthermore, combining phase difference deviation and fluctuation characteristics, the specific process of forming the coupling integrity index by referring to historical baselines is as follows: Calculate the differences between the average phase difference of phases AB, BC, and CA within the window and the ideal three-phase phase difference constant to obtain the ideal phase difference deviations of phases AB, BC, and CA; calculate the sum of squares of the ideal phase difference deviations of phases AB, BC, and CA, and calculate the average to obtain the three-phase average phase deviation within the window; obtain the historical three-phase average phase deviation sequence, and select the median as the phase difference. The mean deviation from the baseline is calculated; the sum of the squares of the standard deviations of the phase differences of phases AB, BC, and CA within the window is calculated, and the average value is obtained to obtain the three-phase phase difference fluctuation within the window; the historical three-phase phase difference fluctuation sequence is obtained, and the median is selected as the phase difference fluctuation baseline; the phase deviation normalization term is obtained by dividing the current three-phase average phase deviation by the mean deviation from the baseline; the phase fluctuation normalization term is obtained by dividing the current three-phase phase difference fluctuation by the phase difference fluctuation baseline; the phase coupling integrity value is obtained by subtracting the phase fluctuation normalization term from the phase deviation normalization term and taking the opposite number as the exponent.

[0012] Furthermore, the specific process for determining the steady state of phase coupling based on the coupling integrity index is as follows: the phase coupling integrity value is calculated in real time and compared with the stability threshold; if the phase coupling integrity value is greater than or equal to the stability threshold, it is marked as a normal phase coupling state; if the phase coupling integrity value is less than the stability threshold, it is marked as an abnormal phase coupling state; the phase coupling integrity value and the phase coupling state mark are written into the electrical fault prediction database.

[0013] Furthermore, the specific process of extracting disturbance enhancement features from electrical behavior characteristics and forming an arc disturbance index by comparing with historical baselines is as follows: The three-phase broadband current energy is obtained, and the broadband current energy of phase A, phase B, and phase C is added together to obtain the total three-phase broadband current energy. The median of the historical total three-phase broadband current energy is selected as the broadband current energy baseline value. The broadband current energy ratios of phase A, phase B, and phase C are obtained by dividing the three-phase broadband current energy by the total three-phase broadband current energy. The deviations of the three-phase broadband energy ratios by one-third are obtained by subtracting one-third from the three-phase broadband energy ratios to obtain the deviations of the three-phase broadband energy ratios. The energy ratio deviation is calculated by taking the square root of the sum of the squares of the three broadband energy ratio deviations to obtain the broadband current energy imbalance. The energy amplification term is obtained by dividing the current total three-phase broadband current energy by the broadband current energy baseline value and adding it to the constant. The effective values ​​of the three-phase power frequency current and the zero-sequence current are obtained, and the effective values ​​of the three-phase power frequency currents A, B, and C are added to obtain the total effective value of the three phases. The zero-sequence leakage relative intensity value is obtained by dividing the zero-sequence current effective value by the sum of the total effective value of the three phases and the minimum constant value. The arc disturbance assessment value is obtained by multiplying the energy amplification term, the sum of the constant and the broadband current energy imbalance, and the sum of the constant and the zero-sequence leakage relative intensity value.

[0014] Furthermore, the specific process of determining the arc steady state based on the arc disturbance index is as follows: the arc disturbance assessment value is calculated in real time and compared with the disturbance threshold; if the arc disturbance assessment value is less than or equal to the disturbance threshold, it is marked as a normal arc disturbance state; if the arc disturbance assessment value is greater than the disturbance threshold, it is marked as an abnormal arc disturbance state; the arc disturbance assessment value and the arc disturbance state mark are written into the electrical fault prediction database.

[0015] Furthermore, the specific process of constructing the arc joint deviation by integrating the deviation of the coupling integrity index relative to the phase coupling steady state and the amplification characteristics of the arc disturbance index relative to the arc steady state, and forming an early evolution score within a sliding window, is as follows: The median of the phase coupling integrity value sequence corresponding to the historical phase coupling normal state is selected from the electrical fault prediction database as the phase coupling integrity baseline value, and the median of the arc disturbance evaluation value sequence corresponding to the historical arc disturbance normal state is selected as the arc disturbance baseline value; the current phase coupling integrity value and arc disturbance evaluation value are obtained, and the difference between the phase coupling integrity baseline value and the phase coupling integrity value is divided by the phase coupling integrity baseline value, and non-negative processing is performed to obtain the phase coupling deviation; the arc disturbance evaluation value is divided by the sum of the arc disturbance baseline value and the minimum constant value to obtain the relative amplification of the arc disturbance; the phase coupling deviation is multiplied by the relative amplification of the arc disturbance to obtain the arc joint deviation; based on the sliding time window, each arc joint deviation within the window is accumulated, and the accumulated result is subjected to hyperbolic tangent operation to obtain the early arc evolution value.

[0016] Furthermore, the specific process of determining the arc evolution stage based on the early evolution score is as follows: The early arc evolution value... Compare with the multi-level evolution thresholds S1 and S2: when When S1 ≤ S1, the current window is determined to be in the normal arc evolution stage; when S1 < When S2 ≤ S2, the current window is determined to be in the early suspected arc stage; when When >S2, it is determined that the current window is in the arc development stage.

[0017] Furthermore, the specific process for generating corresponding intervention measures according to different stages is as follows: When in the normal arc evolution stage, only routine monitoring is maintained, no alarms are triggered, and the fault risk level is output as normal; when in the early suspected arc stage, the fault risk level is output as early suspected arc; an early risk warning is pushed to the operation and maintenance end; the suspected fault location is identified and marked according to the deviation of the three ideal phase differences and the deviation of the broadband energy ratio, and the phase coupling integrity value, arc disturbance assessment value, and early arc evolution value of the current window are output; at the same time, the data sampling and monitoring frequency is increased, and inspection suggestions are pushed; when in the arc development stage, the fault risk level is output as arc development trend is obvious, an arc warning signal is issued, and a development trend prediction curve is generated based on the rate of change of the early arc evolution value; at the same time, intervention action suggestions are generated, including: partial discharge measurement, loop infrared temperature measurement, node fastening inspection, and load switching; an arc evolution trend report is generated, and the early arc evolution value and the corresponding fault risk level are written into the electrical fault prediction database.

[0018] Beneficial effects The present invention has the following beneficial effects: (1) This invention, by constructing a dual-feature coupling integrity index of phase difference deviation and phase fluctuation, no longer relies on single-phase quantity or fixed window monitoring mode, and can capture the weak synchronous destruction process between three phases. Therefore, it can identify the trend of interphase coupling loosening and destruction before the arc causes the current to rise, significantly improving the early identification sensitivity.

[0019] (2) The present invention constructs an arc disturbance index by combining the characteristics of multiple sources of disturbance such as broadband current energy, energy ratio deviation, imbalance degree, and zero-sequence leakage intensity. This can effectively distinguish between background noise and true breakdown precursors, solve the problem that traditional frequency band analysis is difficult to identify weak disturbances from noise, and significantly improve the accuracy of early disturbance identification and anti-interference ability.

[0020] (3) This invention achieves continuous quantitative expression of the early arc evolution trajectory by jointly modeling the coupling deviation and the disturbance amplification, and by forming early arc evolution values ​​through sliding window accumulation and hyperbolic tangent transformation. It can not only classify and judge the three stages of normal, suspected, and developing, but also generate trend predictions based on the evolution slope, thus achieving true forward-looking risk assessment.

[0021] (4) This invention marks the faulty phases by using phase difference deviation and broadband energy deviation structure, providing clear reference for suspected locations for operation and maintenance. At the same time, it automatically triggers action suggestions such as inspection, sampling enhancement, partial discharge detection, infrared temperature measurement, and node tightening based on different evolution stages, realizing an intelligent closed loop from data diagnosis to intervention decision-making, thereby improving safety and maintainability.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1 This is a block diagram of a fault prediction system based on electrical behavior feature mapping. Figure 2 A graph showing the relationship between arc disturbance assessment values ​​and broadband energy evolution; Figure 3 This is a flowchart of the overall process for detecting early arc evolution and determining risk classification. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. As those skilled in the art will understand, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figures 1-3 This invention provides a technical solution: a fault prediction system based on electrical behavior feature mapping, such as... Figure 1 As shown, it includes: an electrical behavior data acquisition and processing module, used to acquire electrical behavior data in real time, preprocess the electrical behavior data, and extract electrical behavior features based on the preprocessed electrical behavior data; a phase coupling integrity assessment module, used to extract phase difference deviation and fluctuation characteristics across phases based on electrical behavior characteristics; combine phase difference deviation and fluctuation characteristics with historical baselines to form a coupling integrity index; determine phase coupling steady state based on the coupling integrity index; a broadband disturbance feature identification module, used to extract disturbance enhancement features from electrical behavior characteristics and form an arc disturbance index with historical baselines; determine arc steady state based on the arc disturbance index; and an early arc evolution assessment module, used to fuse the deviation of the coupling integrity index from the phase coupling steady state and the amplification features of the arc disturbance index from the arc steady state to construct the arc joint deviation, and form an early evolution score within a sliding window; determine the arc evolution stage based on the early evolution score, and generate corresponding intervention measures according to different stages.

[0026] Specifically, the process of real-time acquisition of electrical behavior data, data preprocessing of the electrical behavior data, and extraction of electrical behavior features based on the preprocessed electrical behavior data is as follows: Real-time acquisition of electrical behavior data includes three-phase power frequency current, three-phase power frequency voltage, three-phase broadband current, and zero-sequence current; the three phases include phase A, phase B, and phase C; the three-phase power frequency current is acquired by current transformers arranged in the three-phase circuit, the three-phase power frequency voltage is acquired by voltage transformers; the three-phase broadband current is acquired by broadband current sensors with high-frequency response characteristics, such as Rogowski coils, and the zero-sequence current is acquired by the three-phase current through a zero-sequence current transformer (CT). All sensor sampling signals undergo unified A / D conversion by the front-end data acquisition device and are accompanied by an absolute timestamp. To ensure the synchronization of multi-source measurements, a unified sampling clock is used in the sampling process, with the sampling frequency preferably set in the range of 10kHz to 200kHz to simultaneously cover the power frequency component and the broadband disturbance component, and strict alignment of the three-phase signals is achieved through a precise time protocol of the synchronous clock. Electrical behavior data is time-aligned using the same sampling clock, and DC bias removal and power frequency synchronization filtering are performed to eliminate steady-state offsets and electromagnetic background noise. DC bias removal is achieved using a high-pass filter, while power frequency synchronization filtering employs a linear-phase low-pass filter with a cutoff frequency set between 300Hz and 500Hz to ensure stable extraction of the fundamental power frequency component. Low-pass filtering is used to process the three-phase power frequency current and voltage to obtain their respective stable three-phase power frequency components. Bandpass filtering is used to process the three-phase broadband current to extract the three-phase non-power frequency broadband components. The bandwidth of the bandpass filter is preferably between 2kHz and 50kHz to capture typical arc broadband characteristics. The filter uses an FIR structure to ensure phase consistency. Outlier removal and undersampling compensation are then performed on the filtered electrical behavior data. Outliers are identified and removed using median filtering, and undersampling compensation uses linear interpolation to ensure uninterrupted signal sequence, facilitating subsequent window calculations.z-score normalization is performed on the electrical behavior data, using the mean and standard deviation of the sliding window as a benchmark to avoid feature distortion caused by differences in magnitude across different time periods. The sliding time window is adaptively set according to the sampling frequency, preferably between 20ms and 100ms, so that each window contains 1 to 5 power frequency cycles and covers at least 200 sampling points. This ensures the stability of the statistics while taking into account the evolution timescale of arc precursor disturbances on the order of tens of milliseconds, meeting the constraints of real-time detection and device computing power. Based on the fixed sliding time window, calculations are performed on the three-phase power frequency current, three-phase power frequency voltage, and zero-sequence current. RMS (Real-Time Simulation) yields the RMS values ​​of three-phase power frequency current, three-phase power frequency voltage, and zero-sequence current. The three-phase broadband current energy is obtained by calculating the square integral of the three-phase broadband current over the non-power frequency band, with the integration interval consistent with the bandpass filter bandwidth. This integral is used to quantify the broadband disturbance intensity and serves as an important input parameter for the arc disturbance index. Based on the three-phase power frequency voltage, the instantaneous phase angles of the three-phase power frequency are obtained through Hilbert transform. To avoid phase angle jumps, the instantaneous phase angle sequence of the three-phase power frequency after Hilbert transform needs to be unwrapped to ensure continuous and usable phase differences, thus ensuring accurate calculation of cross-phase synchronization indicators. An electrical fault prediction database is established, using timestamps as unique indexes to write the raw and preprocessed electrical behavior data into the database.

[0027] In this implementation scheme, a multi-source electrical measurement system with a unified sampling clock is constructed to comprehensively acquire three-phase power frequency current, voltage, broadband current, and zero-sequence current. Combined with DC removal, synchronous filtering, outlier elimination, undersampling compensation, and z-score normalization, signal stability and reliability are ensured. Adaptive sliding windows are used to extract key features such as RMS, broadband energy, and instantaneous phase angle, and Hilbert transform and phase expansion are used to ensure phase angle continuity. All raw and feature data are written to the fault prediction database according to timestamps. This method can accurately capture the power frequency and broadband disturbance characteristics of arc precursors, providing high-quality data support for early identification of electrical faults and improving the real-time performance and reliability of predictions.

[0028] Specifically, based on electrical behavior characteristics, the process for extracting the phase difference deviation and fluctuation characteristics across phases is as follows: Using a sliding time window, preferably between 20ms and 100ms, the instantaneous phase angles of the three-phase power frequency are obtained. These instantaneous phase angles are obtained through Hilbert transform, and phase expansion is performed on the Hilbert phase angle sequence to ensure continuous usability of phase angle changes, thereby avoiding jump distortions during the calculation of cross-phase phase differences. The instantaneous phase difference of phases AB is obtained by calculating the phase difference between phases A and B in the instantaneous phase angles of the three-phase power frequency, and the instantaneous phase difference of phases BC is obtained by calculating the phase difference between phases B and C. Phase difference is calculated by determining the phase difference between phase C and phase A to obtain the instantaneous phase difference of phase CA. This yields the instantaneous phase difference sequences of phases AB, BC, and CA within the window. The instantaneous phase difference sequences characterize the real-time synchronization degree between each phase within the current window. If the equipment is in a healthy three-phase symmetrical operating condition, the three phase difference sequences AB, BC, and CA should be stably distributed around 120° with only minor fluctuations. If there is poor contact, partial discharge, or arc precursor, the phase difference sequences will show slow shifts or short-term increased jitter, becoming a sensitive quantity for subsequent anomaly detection. The average and standard deviation of the three instantaneous phase difference sequences within the window are calculated to obtain the average phase difference of phases AB, BC, and CA, as well as the standard deviations of the phase differences of phases AB, BC, and CA. Among them, the average phase differences of AB, BC, and CA are used to quantify the systematic deviation of the three-phase phase relationship relative to the ideal 120° interval, and are slow variables that measure whether the phase coupling deviates from the symmetrical operating condition; the standard deviations of the phase differences of AB, BC, and CA reflect the fluctuation intensity of each phase difference within the window, and are fast variables that characterize phase jitter, intermittent disturbances, and arc precursor instability; the average value and standard deviation will serve as the basic input parameters in the subsequent calculation of the phase coupling integrity index, and by simultaneously constraining the deviation and fluctuation, the stability and reliability of phase coupling anomaly identification will be improved.

[0029] In this implementation scheme, by using a unified high-precision sampling system, strict clock synchronization, targeted filtering and separation strategies, and statistical feature extraction based on sliding windows, the three-phase operating condition information can be stably recovered in a noisy environment. The instantaneous phase angle obtained by Hilbert transform, combined with phase difference offset and fluctuation measurement, can simultaneously capture the chronic offset and rapid disturbance characteristics between the three phases. The combination of broadband energy, zero-sequence component, and phase coupling index constitutes a multi-dimensional sensitive quantity for early arc precursors, which can achieve stable and repeatable early identification before the fault causes significant current anomalies, significantly improving the accuracy of fault prediction and response lead, and enhancing the feasibility and engineering implementation of the technical solution.

[0030] Specifically, the process of forming the coupling integrity index by combining phase difference deviation and fluctuation characteristics with historical baselines is as follows: The differences between the average phase difference of phase AB, phase BC, and phase CA within the window and the ideal three-phase phase difference constant are calculated to obtain the ideal phase difference deviations of phase AB, BC, and CA; where the ideal three-phase phase difference constant is the theoretical value of 120°. This method reflects the target phase relationship of three-phase symmetry and is used to measure whether phase coupling deviates from the symmetrical operating state. Fixed values ​​avoid calculation ambiguity. The sum of squares of the ideal phase difference deviations of phases AB, BC, and CA is calculated, and the average is obtained to get the three-phase average phase deviation within the window. A historical three-phase average phase deviation sequence is obtained, and the median is selected as the baseline for the mean phase difference deviation. The sum of squares of the standard deviations of phase differences of phases AB, BC, and CA within the window is calculated, and the average is obtained to get the three-phase phase difference fluctuation within the window. Fluctuation is used to characterize the intensity of short-term disturbances and is highly sensitive to early random jitter of the arc, making it an indispensable statistical feature for revealing weak early disturbances. A historical three-phase phase difference fluctuation sequence is obtained, and the median is selected as the baseline for phase difference fluctuation. Using the median as the historical baseline reduces the impact of extreme disturbances on the statistical baseline and improves the stability and repeatability of the indicator. The phase deviation normalization term is obtained by dividing the current three-phase average phase deviation by the mean phase difference deviation from the baseline; the phase fluctuation normalization term is obtained by dividing the current three-phase phase difference fluctuation by the phase difference fluctuation baseline. These two normalization processes are used to eliminate differences in dimensions and magnitudes, making the characteristics of different equipment and different operating scenarios comparable, while improving the transferability of the model in various power distribution systems. The phase coupling integrity value is obtained by subtracting the phase fluctuation normalization term from the phase deviation normalization term and taking the opposite number as the exponent. The reason for choosing the natural exponential function as the mapping function is that it has monotonicity and continuity, which can compress the combined result of "phase deviation, the larger the deviation, the worse" and "fluctuation suppression, the more stable, the better" to the (0,1) interval, so that the phase coupling integrity value presents good physical interpretation. The closer it is to 1, the healthier the three-phase coupling; the closer it is to 0, the more broken the coupling. In addition, in order to avoid numerical instability caused by the denominator approaching zero, a very small positive number needs to be set in the calculation. As a lower bound, it ensures that exponentiation operations can be performed stably and avoids numerical overflow.

[0031] The specific formula for the phase coupling integrity value is as follows: ; In the formula, The phase coupling integrity value is used to quantify the stability of the phase relationship in a three-phase electrical system and is a core indicator for identifying potential arcing hazards in the early stages of cross-phase failure. This represents the average phase difference between phases A and B; This represents the average phase difference between phases BC; This represents the average phase difference of the CA phase; it is used to measure the degree to which the overall phase difference of the three pairs of phase differences within the current window deviates from the ideal 120°. This represents the standard deviation of the phase difference between phases A and B; This represents the standard deviation of the phase difference between phases BC; It represents the standard deviation of the phase difference of phase CA; it is used to reflect the short-time fluctuation amplitude of the three pairs of phase differences AB, BC, and CA, and to characterize the disturbance of phase stability between phases. It represents the deviation of the average phase difference from the baseline, and is used to characterize the average phase difference deviation level of the device in a healthy state. It is a historical reference baseline for judging whether the current deviation is abnormal. It represents the baseline quantity of phase difference fluctuation, used to reflect the typical fluctuation level of the three-phase phase difference under healthy conditions, and serves as a normalized reference baseline for the current fluctuation intensity.

[0032] In this implementation scheme, by quantifying and normalizing the three-phase average phase deviation and phase difference fluctuation respectively, and constructing an exponential phase coupling integrity value in conjunction with historical benchmarks, early sensitive identification of three-phase synchronization degradation is achieved. This method can simultaneously reflect the long-term trend of phase shift and the transient characteristics of short-term disturbances, avoiding misjudgments caused by a single indicator; the exponential mapping gives the results good numerical stability and interpretability, making it easy to use for operational status classification; at the same time, constructing a benchmark quantity based on the median improves noise resistance and transferability under different scenarios, thereby significantly enhancing the early detection capability of phase coupling anomalies and providing a more reliable quantitative basis for arc precursor detection and fault prediction.

[0033] Specifically, the process of determining the steady state of phase coupling based on the coupling integrity index is as follows: The phase coupling integrity value is calculated in real time and compared with a stability threshold. The stability threshold is preferably determined by statistically analyzing the distribution of phase coupling integrity values ​​under long-term operating conditions, and can be taken as the P10 to P20 quantile to ensure sensitivity to slight synchronicity degradation. If the phase coupling integrity value is greater than or equal to the stability threshold, it is marked as a normal phase coupling state, indicating that the phase relationship of the three-phase voltages is maintained within a healthy operating range and there is no obvious trend of phase decoupling. If the phase coupling integrity value is less than the stability threshold, it is marked as an abnormal phase coupling state, indicating signs of early phase disturbances and potential cross-phase faults. To avoid misjudgments caused by transient disturbances, the state flag change can be triggered only after two to three consecutive windows have all been determined to be abnormal, thereby improving the reliability of steady-state determination. The phase coupling integrity value and the phase coupling state flag are then written into the electrical fault prediction database.

[0034] In this implementation scheme, the steady-state determination method based on the coupling integrity index achieves quantitative and real-time monitoring of the three-phase phase synchronization relationship, enabling the detection of early signs of inter-phase coupling degradation before the three-phase voltage amplitude becomes abnormal. By using normalized deviation and fluctuation as comprehensive inputs, it accurately distinguishes between non-substantial fluctuations caused by transient noise and persistent offsets caused by electrical decoupling, effectively improving the robustness of steady-state determination. Combined with a stability threshold determined by the distribution of healthy samples, the determination criteria have clear physical meaning and statistical basis, avoiding reliance on manual experience. The continuous window triggering mechanism further suppresses false alarms caused by transient disturbances, improving the stability of phase coupling state identification. Finally, the steady-state markers are structured and written into the database, providing reliable preliminary information for subsequent arc disturbance index calculation and early evolution trend assessment, thereby significantly enhancing the early diagnosis and trend identification capabilities of the entire fault prediction.

[0035] Specifically, the process of extracting disturbance enhancement features from electrical behavior characteristics and forming an arc disturbance index by comparing with historical baselines is as follows: The three-phase broadband current energy is obtained, and the broadband current energy of phase A, phase B, and phase C is added together to obtain the total three-phase broadband current energy. This total three-phase broadband current energy is used to characterize the overall disturbance amplitude, preventing misjudgment due to single-phase energy anomalies. The median of the historical total three-phase broadband current energy is selected as the broadband current energy baseline value. The broadband current energy ratios of phase A, phase B, and phase C are obtained by dividing the three-phase broadband current energy by the total three-phase broadband current energy. One-third is subtracted from each of the three-phase broadband energy ratios to obtain the deviations in the broadband energy ratios of phase A, phase B, and phase C. The following steps are performed: First, calculate the square root of the sum of the squares of the deviations in the three broadband energy proportions to obtain the broadband current energy imbalance degree. This degree characterizes whether the three-phase energy distribution exhibits abnormal deviations, thus identifying the phase concentration characteristics of the arc. Second, divide the current total three-phase broadband current energy by the broadband current energy baseline value, add it to a constant, and take the natural logarithm to obtain the energy amplification term. This term enhances the sensitivity to sudden energy increases and keeps the input numerically stable, avoiding large fluctuations that could disrupt the model's monotonicity. Third, obtain the effective values ​​of the three-phase power frequency current and the zero-sequence current. Add the effective values ​​of the A, B, and C phase power frequency currents to obtain the total effective value of the three phases. Divide the effective value of the zero-sequence current by the sum of the total effective value of the three phases and the minimum constant value to obtain the relative intensity value of the zero-sequence leakage. The value of this term is used to avoid numerical divergence caused by a zero or excessively small denominator. The arc disturbance assessment value is obtained by multiplying the energy amplification term, the sum of constant 1 and broadband current energy imbalance, and the sum of constant 1 and zero-sequence leakage relative intensity. All three terms are shifted by "constant 1" to avoid negative inputs that could lead to unclear product direction, thus enhancing the model's interpretability and monotonicity. The arc disturbance assessment value serves as a joint indicator reflecting energy amplification, phase shift, and leakage intensity; a larger value indicates stronger arc disturbance and can be used as a core input for subsequent early arc evolution scoring.

[0036] The specific formula for evaluating arc disturbance is as follows: ; In the formula, It represents the arc disturbance assessment value, which serves as the core indicator for quantifying the intensity of arc disturbance and is used to comprehensively reflect the combined amplitude of three types of disturbance characteristics: energy amplification, three-phase imbalance, and zero-sequence leakage. It represents the total energy of the three-phase broadband current, characterizes the overall energy level of the broadband current, is a key indicator for judging whether the energy of the arc disturbance has increased, and also serves as a normalized benchmark for the energy ratio of each phase. It represents the baseline value of broadband current energy, allowing the current energy level to be compared with the normal state, thereby reflecting whether there is an abnormal increase in energy; This represents the broadband current energy of phase A; This represents the broadband current energy of phase B; It represents the C-phase broadband current energy; it is used to calculate the proportion of the three-phase broadband energy and compare it with one-third to measure whether the energy of one phase is abnormally amplified or weakened, that is, to characterize the imbalance of the three-phase broadband disturbance. It represents the effective value of zero-sequence current, measures whether there is leakage current or current asymmetry caused by arcing, and is one of the important characteristics of arc disturbance; It represents the total effective value of the three phases, serving as the normalization reference for the zero-sequence current; This represents a very small constant value, used to avoid numerical divergence caused by a denominator that is zero or too small. Its value is [value missing]. .

[0037] In this embodiment, Table 1 is a data table of arc disturbance assessment values. The baseline value of broadband current energy is set to 100. The table details the total effective value of the three phases, the effective value of the zero-sequence current, the broadband current energy of phase A, phase B, phase C, the total broadband current energy of the three phases, and the arc disturbance assessment value at five different times. Specifically, at time 1, the total effective value of the three phases is 293, the effective value of the zero-sequence current is 0.8, the broadband current energy of phase A is 34, the broadband current energy of phase B is 33, the broadband current energy of phase C is 33, the total broadband current energy of the three phases is 100, and the arc disturbance assessment value is 0.7007. At time 2, the total effective value of the three phases is 323, the effective value of the zero-sequence current is 2.5, the broadband current energy of phase A is 48, the broadband current energy of phase B is 46, the broadband current energy of phase C is 46, the total broadband current energy of the three phases is 140, and the arc disturbance assessment value is 0.892. At time 3, the total effective value of the three phases is 350, the effective value of the zero-sequence current is 4, the broadband current energy of phase A is 90, the broadband current energy of phase B is 70, the broadband current energy of phase C is 60, the total broadband current energy of the three phases is 220, and the arc disturbance assessment value is 1.2920. At time 4, the total effective value of the three phases is 292, the effective value of the zero-sequence current is 12, the broadband current energy of phase A is 36, the broadband current energy of phase B is 32, the broadband current energy of phase C is 32, the total broadband current energy of the three phases is 100, and the arc disturbance assessment value is 0.7452. At time 5, the total effective value of the three phases is 377, the effective value of the zero-sequence current is 28, the broadband current energy of phase A is 120, the broadband current energy of phase B is 80, the broadband current energy of phase C is 60, the total broadband current energy of the three phases is 260, and the arc disturbance assessment value is 1.6047.

[0038] Table 1. Data Table of Arc Disturbance Assessment Values

[0039] like Figure 2 The figure shows the relationship between the arc disturbance assessment value and the broadband energy evolution. It illustrates the synchronous change trend of the arc disturbance assessment value and the total three-phase broadband current energy over five increasing monitoring times. The figure uses a dual vertical axis structure; the left vertical axis reflects the intensity of the disturbance assessment value, and the right vertical axis reflects the broadband energy scale. The two curves together describe the disturbance enhancement at different stages. (See Table 1 and...) Figure 2It can be seen that from time 1 to time 3, both increase synchronously, then decrease significantly at time 4, and rise rapidly again at time 5, indicating that the enhancement of broadband energy is usually accompanied by an increase in disturbance intensity. The first peak occurs at time 3, which is the main inflection point of disturbance enhancement, with the total three-phase broadband current energy reaching 220 and the arc disturbance assessment value reaching 1.292, indicating that strong energy imbalance and zero-sequence leakage characteristics have appeared within the electrical system at this time. At time 4, the energy drops significantly, corresponding to a decrease in broadband energy to 100, indicating a temporary decline from a high disturbance state. Time 5 reaches the highest point of the entire process, indicating a potentially high-risk stage, with broadband energy rising to 260 and the disturbance assessment value rising to 1.6047, representing the strongest disturbance performance of all times, possibly corresponding to a significant arc development trend, requiring early warning and intervention.

[0040] In this implementation scheme, by jointly modeling the three-dimensional characteristics of total energy, three-phase imbalance, and zero-sequence leakage, the method can simultaneously capture key arc disturbance signs such as energy amplification, phase concentration characteristics, and changes in grounding leakage paths. This makes the arc disturbance index more sensitive and noise-resistant to early weak arcs. The introduction of historical baselines and minimal constant terms further enhances the numerical robustness of the evaluation index, making the method stable and feasible in complex actual operating scenarios, and significantly improving the early arc identification capability.

[0041] Specifically, the process of determining the arc steady state based on the arc disturbance index is as follows: The arc disturbance assessment value is calculated in real time and compared with a disturbance threshold. The disturbance threshold can be determined based on statistical results of historical normal operation data, preferably the P90 to P95 quantiles of the historical arc disturbance assessment value distribution, to ensure effective differentiation between normal disturbances and early arc anomalies. The real-time calculation process is synchronized with the sliding time window, ensuring that the judgment granularity is consistent with the equipment's dynamic response. If the arc disturbance assessment value is less than or equal to the disturbance threshold, it is marked as a normal arc disturbance state; if the arc disturbance assessment value is greater than the disturbance threshold, it is marked as an abnormal arc disturbance state. To avoid frequent state switching due to data jumps, a hysteresis confirmation mechanism of 2 to 3 windows can be added before the threshold comparison to improve the stability of the state determination. The arc disturbance assessment value and the arc disturbance state marker are written into the electrical fault prediction database to form a traceable disturbance change trajectory.

[0042] In this implementation scheme, by introducing statistical thresholds, hysteresis mechanisms, and structured state markers, this method can form stable and repeatable arc steady-state determination results based on the arc disturbance index, which significantly improves the robustness of arc anomaly identification. At the same time, the determination results are written into the database, making the arc steady-state determination traceable and quantifiable, providing a reliable foundation for subsequent assessment of the early evolution trend of the arc.

[0043] Specifically, the process of constructing a joint arc deviation by integrating the deviation of the coupling integrity index from the phase coupling steady state and the amplification characteristics of the arc disturbance index relative to the arc steady state, and forming an early evolution score within a sliding window, is as follows: the median of the phase coupling integrity value sequence corresponding to the historical phase coupling normal state is selected from the electrical fault prediction database as the phase coupling integrity baseline value, and the median of the arc disturbance assessment value sequence corresponding to the historical arc disturbance normal state is selected as the arc disturbance baseline value; using the median as the baseline can eliminate the influence of extreme samples, making the deviation more robust and ensuring that the evolution score is not affected by occasional disturbances. Obtain the current phase coupling integrity value and arc disturbance assessment value. Divide the difference between the phase coupling integrity baseline value and the current phase coupling integrity value by the baseline value, and perform non-negativity processing to obtain the phase coupling deviation. Non-negativity processing involves comparing the calculated result with zero and taking the larger value to avoid abnormal situations caused by sign reversal. The phase coupling deviation can be used as an intensity indicator to accumulate within a window, reflecting the degree of phase coupling disruption. Divide the arc disturbance assessment value by the sum of the arc disturbance baseline value and the minimum constant value to obtain the relative arc disturbance amplification. The minimum constant is used to avoid numerical explosion caused by the denominator approaching zero, and its value is set to [value missing]. To balance numerical stability and sensitivity, the relative amplification value reflects the degree of broadband perturbation anomaly enhancement, a key characteristic of early arc development. Multiplying the phase coupling deviation by the relative amplification of the arc perturbation yields the arc joint deviation value. This product reflects the design intuition that an arc evolution trend is only considered when phase decoupling and energy anomalies coexist, suppressing false alarms caused by a single anomaly. The product method is essentially a logical AND emphasis mechanism, allowing strong anomalies to contribute more, while weak anomalies cannot trigger score enhancement alone, thus reducing the probability of misjudgment. Based on a sliding time window, with a window length preferably between 100ms and 1s, each arc joint deviation value within the window is accumulated. Accumulation rather than averaging reflects the impact of time continuity on the arc's performance. The key contributions of early arc evolution are: short-term burst noise does not cause cumulative increase, while continuous deviation gradually drives the score up, which is more consistent with the physical process of arc initiation; the accumulation method also enhances the sensitivity of the method to weak long-term anomalies; and the early arc evolution value is obtained by performing hyperbolic tangent operation on the accumulation result. The use of tanh has a clear mathematical motivation: its S-shaped characteristic can provide a highly sensitive linear response when the deviation is small, while it automatically enters the smooth saturation region when the deviation is large, thereby avoiding false alarms caused by infinite growth of the score; tanh's interval compression capability, limiting the output to -1 to 1, ensures the numerical stability of the early arc evolution value, allowing the score to be used as a unified scale input to the subsequent stage division logic, which is beneficial for controlling false alarms and over-response.

[0044] The specific formula for the early arc evolution value is as follows: ; In the formula, This represents the early arc evolution value, reflecting the overall intensity of the arc at the end of the window from its early stage to its evolution and development. Indicates the length of the sliding time window, used to form a trend assessment; This represents the baseline value for phase coupling integrity, indicating the three-phase coupling reference level of the equipment under healthy operating conditions; It represents the phase coupling integrity value, which quantitatively reflects the synchronicity of the current three-phase phase relationship; This represents the arc disturbance assessment value, reflecting the intensity of typical arc disturbances such as broadband current energy anomalies and enhanced zero-sequence leakage. This represents the baseline value of the arc disturbance, indicating the typical level of the arc disturbance during the healthy phase; This represents a very small constant value, ensuring that calculations are still possible even with extremely low arc disturbances and avoiding division by zero errors. The value is [value to be filled in]. .

[0045] In this implementation scheme, stable, continuous, and physically consistent quantification of the early arc evolution process is achieved through joint modeling of phase coupling deviation and broadband perturbation amplification, a deviation accumulation mechanism, and piecewise sensitivity compression of tanh. The results maintain high sensitivity to weak precursors while suppressing sporadic noise, significantly improving the reliability and interpretability of early arc trend identification.

[0046] Specifically, the process of determining the arc evolution stage based on the early evolution score is as follows: The early arc evolution value... Compare with the multi-level evolution thresholds S1 and S2: when When S1 ≤ S1, the current window is determined to be in the normal arc evolution stage; when S1 < When S2 ≤ S2, the current window is determined to be in the early suspected arc stage; when When S2 is reached, the current window is determined to be in the arc development stage. The multi-level evolution thresholds S1 and S2 are obtained based on statistical analysis of historical long-term operating samples. S1 characterizes the upper bound of the early arc evolution value under typical healthy operating conditions, and can be taken as the 90th to 95th percentile of the early arc evolution value distribution. S2 distinguishes between suspected arcs and the development stage, and can be taken as the 50th to 70th percentile of the early arc evolution value distribution. To ensure that the thresholds can be repeatedly set, all thresholds in this invention are managed consistently based on the number of sampling points, the size of the sliding window, and the device model, and a version number is recorded in the database.

[0047] In this implementation plan, by introducing statistical thresholds and clearly defining physical stages, this method can achieve step-by-step identification of arc initiation, suspected arc, and arc development without increasing the false alarm rate, significantly improving the reliability, stability, and engineering feasibility of early arc alarms.

[0048] Specifically, the process of generating corresponding intervention measures according to different stages is as follows: When in the normal arc evolution stage, only routine monitoring is maintained, no alarm is triggered, and the fault risk level is output as normal; when in the early suspected arc stage, the fault risk level is output as early suspected arc; an early risk warning is pushed to the operation and maintenance end; the suspected fault location is identified and marked according to the deviation of three ideal phase differences and the deviation of broadband energy proportion, and the phase most likely to be faulty is determined by quantitative comparison. Specifically, firstly, the deviation of broadband current energy proportion of the three phases A, B, and C is calculated, and the phase with the largest deviation is selected as the candidate phase for broadband energy anomaly; then, the phase difference between the two phases involved in the candidate phase is calculated respectively. For example, if the candidate phase is A, the corresponding deviation is the average phase difference between AB and CA in the current window. If both phase difference deviations exceed the threshold of the corresponding historical steady-state mean plus 1.5 times the standard deviation, the candidate phase is directly determined as the suspected fault phase. If only one phase difference deviation exceeds the threshold, the standard deviation of the phase difference fluctuation between two adjacent phases is compared, and the phase corresponding to the side with greater fluctuation is determined as the suspected fault phase. It outputs the phase coupling integrity value, arc disturbance assessment value, and early arc evolution value of the current window; simultaneously, it increases the data sampling and monitoring frequency, preferably by 1.5 to 2 times the original frequency, to enhance the ability to capture rapidly evolving disturbances while maintaining compatibility with existing data structures, and pushes inspection suggestions; when in the arc development stage, it outputs a fault risk level indicating a clear arc development trend, issues an arc warning signal, and generates a development trend prediction curve based on the rate of change of early arc evolution values, where the rate of change is obtained by the difference between two adjacent early arc evolution values, so that the trend curve can reflect the arc growth rate and form a visual representation. The system optimizes the risk escalation slope and generates intervention action suggestions, including partial discharge measurement, loop infrared thermography, node tightening inspection, and load switching. These intervention action suggestions are automatically prioritized based on risk severity, with priority given to actions that can be confirmed without power outages, such as infrared thermography, followed by operational measures like node tightening and load switching, thus improving the operability and safety of interventions. An arc evolution trend report is generated, including early arc evolution values, trend curves, stage judgment results, and suggested action records, providing traceability and supporting operation and maintenance decisions. The early arc evolution values ​​and corresponding fault risk levels are then written into the electrical fault prediction database.

[0049] like Figure 3The diagram shows the overall flowchart for early arc evolution detection and risk classification. It illustrates the overall processing flow for early arc evolution detection. First, three-phase current, voltage, broadband current, and zero-sequence current are acquired in real time, and A / D conversion and time synchronization are performed using a unified sampling clock. After DC removal, power frequency synchronization filtering, outlier removal, and undersampling compensation, key features such as RMS, voltage phase angle, and broadband current energy are extracted. Then, phase difference deviation and phase difference fluctuation are calculated separately, and a phase coupling integrity index is obtained based on historical baselines; simultaneously, broadband energy imbalance and zero-sequence leakage intensity are calculated to form an arc disturbance index. These two are further merged into a joint deviation, accumulated through a sliding window, and the early arc evolution value is obtained through a hyperbolic tangent function. Finally, based on the evolution value and multi-level thresholds, a stage judgment is made, outputting the risk level of normal, suspected, or developing stages. For suspected and developing stages, location indication, early warning prompts, trend judgments, and intervention suggestions are provided. Simultaneously, the judgment results are written into a fault prediction database, achieving closed-loop processing for early warning of arc faults.

[0050] In this implementation plan, by implementing graded responses at different evolution stages, dynamically improving monitoring resolution, quantifying and locating suspected locations, constructing trend curves by combining change rates, and outputting executable intervention actions, the arc risk is made to form a closed loop from detection and location to assessment and intervention. This significantly improves the observability, interpretability, and controllability of early arc evolution and enhances the proactive safety capabilities of electrical systems.

[0051] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0052] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. As those skilled in the art will understand, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A fault prediction system based on electrical behavior feature mapping, characterized in that, include: The electrical behavior data acquisition and processing module is used to acquire electrical behavior data in real time, preprocess the electrical behavior data, and extract electrical behavior features based on the preprocessed electrical behavior data. The phase coupling integrity assessment module is used to extract phase difference deviation and fluctuation characteristics across phases based on electrical behavior features; and to form a coupling integrity index by combining phase difference deviation and fluctuation characteristics and comparing with historical baselines. Determine the steady state of phase coupling based on the coupling integrity index; A broadband disturbance feature identification module is used to extract disturbance enhancement features from electrical behavior features and form an arc disturbance index by comparing with historical baselines; the arc steady state is determined based on the arc disturbance index. The early arc evolution assessment module is used to integrate the deviation of the coupling integrity index from the phase coupling steady state and the amplification characteristics of the arc disturbance index relative to the arc steady state to construct the joint arc deviation and form an early evolution score within a sliding window; the arc evolution stage is determined based on the early evolution score, and corresponding intervention measures are generated according to different stages.

2. The fault prediction system based on electrical behavior feature mapping according to claim 1, characterized in that, The specific process of real-time acquisition of electrical behavior data, data preprocessing of electrical behavior data, and extraction of electrical behavior features based on the preprocessed electrical behavior data is as follows: Real-time acquisition of electrical behavior data, including: three-phase power frequency current, three-phase power frequency voltage, three-phase broadband current, and zero-sequence current; where the three phases include phase A, phase B, and phase C; The electrical behavior data is time-aligned using the same sampling clock, and DC bias removal and power frequency synchronization filtering are performed to eliminate steady-state offset and electromagnetic background noise. Low-pass filtering is used to process the three-phase power frequency current and voltage to obtain their respective stable three-phase power frequency components. Band-pass filtering is used to process the three-phase broadband current to extract the three-phase non-power frequency broadband components. Outlier removal and undersampling compensation are then performed on the filtered electrical behavior data. Z-score normalization is performed on the electrical behavior data. Based on a fixed sliding time window, the RMS values ​​of the three-phase power frequency current, three-phase power frequency voltage, and zero-sequence current are calculated to obtain the effective values ​​of the three-phase power frequency current, three-phase power frequency voltage, and zero-sequence current. The square integral of the three-phase broadband current in the non-power frequency band is calculated to obtain the three-phase broadband current energy. The instantaneous phase angle of the three-phase power frequency is obtained through Hilbert transform based on the three-phase power frequency voltage. An electrical fault prediction database is established, and the original and preprocessed electrical behavior data are written into the database.

3. The fault prediction system based on electrical behavior feature mapping according to claim 1, characterized in that, The specific process for extracting the phase difference deviation and fluctuation characteristics across phases based on electrical behavior features is as follows: Based on a sliding time window, the instantaneous phase angles of the three-phase power frequency are obtained. The instantaneous phase difference of phase AB is obtained by calculating the phase difference between phase A and phase B in the instantaneous phase angles of the three-phase power frequency. The instantaneous phase difference of phase BC is obtained by calculating the phase difference between phase B and phase C. The instantaneous phase difference of phase CA is obtained by calculating the phase difference between phase C and phase A. The instantaneous phase difference sequences of phase AB, phase BC, and phase CA within the window are obtained. The average value and standard deviation of the three instantaneous phase difference sequences within the window are calculated to obtain the average phase difference of phase AB, the average phase difference of phase BC, the average phase difference of phase CA, the standard deviation of phase difference of phase AB, the standard deviation of phase difference of phase BC, and the standard deviation of phase difference of phase CA.

4. The fault prediction system based on electrical behavior feature mapping according to claim 1, characterized in that, The specific process of forming the coupling integrity index by combining phase difference deviation and fluctuation characteristics with historical baselines is as follows: The deviations of the ideal phase difference between phases AB, BC, and CA within the window are calculated respectively from the ideal three-phase phase difference constant to obtain the deviations of the ideal phase difference between phases AB, BC, and CA. Calculate the sum of squares of the ideal phase difference deviations of phases AB, BC, and CA, and average them to obtain the average phase deviation of the three phases within the window; obtain the historical three-phase average phase deviation sequence, and select the median as the mean deviation of the phase difference from the baseline. Calculate the sum of squares of the standard deviations of phase differences of phases AB, BC, and CA within the window, and average them to obtain the three-phase phase difference fluctuation within the window; obtain the historical three-phase phase difference fluctuation sequence, and select the median as the benchmark for phase difference fluctuation; The phase deviation normalization term is obtained by dividing the current three-phase average phase deviation by the mean deviation of the phase difference from the reference. The phase fluctuation normalization term is obtained by dividing the current three-phase phase difference fluctuation by the phase difference fluctuation reference quantity. The phase coupling integrity value is obtained by subtracting the phase fluctuation normalization term from the phase deviation normalization term and taking the opposite number as the exponent for natural exponentiation.

5. The fault prediction system based on electrical behavior feature mapping according to claim 1, characterized in that, The specific process for determining the phase coupling steady state based on the coupling integrity index is as follows: Calculate the phase coupling integrity value in real time and compare it with the stability threshold: If the phase coupling integrity value is greater than or equal to the stability threshold, it is marked as a normal phase coupling state; If the phase coupling integrity value is less than the stability threshold, it is marked as a phase coupling abnormal state; The phase coupling integrity value and phase coupling status flag are written into the electrical fault prediction database.

6. The fault prediction system based on electrical behavior feature mapping according to claim 1, characterized in that, The specific process of extracting disturbance enhancement features from electrical behavior characteristics and forming an arc disturbance index by comparing with historical baselines is as follows: The three-phase broadband current energy is obtained, and the broadband current energy of phase A, phase B, and phase C is added together to obtain the total broadband current energy. The median of the historical total broadband current energy is selected as the baseline value of broadband current energy. The broadband energy percentages of phase A, phase B, and phase C are obtained by dividing the three-phase broadband current energy by the total three-phase broadband current energy, respectively. The deviation of the broadband energy ratio of phase A, phase B, and phase C is obtained by subtracting one-third from the three-phase broadband energy ratios respectively. The broadband current energy imbalance is obtained by taking the square root of the sum of the squares of the three broadband energy ratio deviations. The energy amplification term is obtained by dividing the current total three-phase broadband current energy by the broadband current energy baseline value, adding it to the constant phase, and taking the natural logarithm. Obtain the effective values ​​of the three-phase power frequency current and the effective value of the zero-sequence current, and add the effective values ​​of the three-phase power frequency currents A, B and C to obtain the total effective value of the three phases. Divide the effective value of the zero-sequence current by the sum of the total effective value of the three phases and the minimum constant value to obtain the relative intensity value of the zero-sequence leakage. The arc disturbance assessment value is obtained by multiplying the energy amplification term, the sum of constant 1 and broadband current energy imbalance, and the sum of constant 1 and zero-sequence leakage relative intensity value.

7. The fault prediction system based on electrical behavior feature mapping according to claim 1, characterized in that, The specific process for determining the steady state of the arc based on the arc disturbance index is as follows: Real-time calculation of arc disturbance assessment values ​​and comparison with disturbance thresholds: If the arc disturbance assessment value is less than or equal to the disturbance threshold, it is marked as a normal arc disturbance state. If the arc disturbance assessment value is greater than the disturbance threshold, it is marked as an abnormal arc disturbance state; Arc disturbance assessment values ​​and arc disturbance status markers are written into the electrical fault prediction database.

8. The fault prediction system based on electrical behavior feature mapping according to claim 1, characterized in that, The specific process of constructing the arc joint deviation amount by considering the deviation of the fusion coupling integrity index from the phase coupling steady state and the amplification characteristics of the arc perturbation index relative to the arc steady state, and forming an early evolution score within the sliding window, is as follows: The median of the phase coupling integrity value sequence corresponding to the historical phase coupling normal state was selected from the electrical fault prediction database as the phase coupling integrity baseline value, and the median of the arc disturbance evaluation value sequence corresponding to the historical arc disturbance normal state was selected as the arc disturbance baseline value. Obtain the current phase coupling integrity value and arc disturbance assessment value. Divide the difference between the phase coupling integrity baseline value and the phase coupling integrity value by the phase coupling integrity baseline value and perform non-negation processing to obtain the phase coupling deviation. Divide the arc disturbance assessment value by the sum of the arc disturbance baseline value and the minimum constant value to obtain the relative amplification of the arc disturbance. The phase coupling deviation is multiplied by the relative amplification of the arc disturbance to obtain the arc joint deviation. Based on the sliding time window, the arc joint deviation of each time window is accumulated, and the accumulated result is subjected to hyperbolic tangent operation to obtain the early arc evolution value.

9. The fault prediction system based on electrical behavior feature mapping according to claim 1, characterized in that, The specific process for determining the arc evolution stage based on the early evolution score is as follows: Early arc evolution value Compare with the multi-level evolution thresholds S1 and S2: when When S1 ≤ S1, the current window is determined to be in the normal stage of arc evolution; When S1 < When S2 ≤ S2, the current window is determined to be in the early suspected arc stage; when When >S2, it is determined that the current window is in the arc development stage.

10. The fault prediction system based on electrical behavior feature mapping according to claim 1, characterized in that, The specific process for generating corresponding intervention measures based on different stages is as follows: When the arc evolution is in a normal stage, only routine monitoring is maintained, no alarm is triggered, and the fault risk level is output as normal. When the fault is in the early suspected arc stage, the fault risk level is output as early suspected arc; an early risk warning is pushed to the operation and maintenance terminal. Based on the deviation of three ideal phase differences and the deviation of broadband energy ratio, suspected fault locations are identified and marked, and the phase coupling integrity value, arc disturbance assessment value, and early arc evolution value of the current window are output; at the same time, the data sampling and monitoring frequency is increased, and inspection suggestions are pushed. When the arc is in the development stage, the output fault risk level is that the arc development trend is obvious, an arc warning signal is issued, and a development trend prediction curve is generated based on the rate of change of the early arc evolution value. Simultaneously, intervention action suggestions are generated, including: partial discharge measurement, loop infrared temperature measurement, node fastening check, and load switching. Generate an arc evolution trend report and write early arc evolution values ​​and corresponding fault risk levels into the electrical fault prediction database.

Images