A fault location sensor for pole-mounted circuit breakers

Abstract

This invention relates to the field of fault location technology, and more particularly to a fault location sensor for pole-mounted circuit breakers. The technical solution is as follows: A fault location sensor for pole-mounted circuit breakers includes: a sensor module for synchronously acquiring multi-physical quantity signals at the installation point, wherein the multi-physical quantity signals include at least two of high-frequency current signals, power frequency current / voltage signals, electric field / magnetic field signals, and temperature signals; and a processing module connected to the sensor module for performing on-site diagnosis and intelligent analysis of faults in the pole-mounted circuit breaker or its associated lines based on the synchronously acquired multi-physical quantity signals and by running a built-in diagnostic program. This invention effectively distinguishes various fault events by integrating multi-dimensional information such as high-frequency traveling waves, power frequency electrical quantities, electromagnetic fields, and temperature, and using a built-in intelligent diagnostic algorithm for comprehensive judgment, avoiding ineffective inspections and accurately guiding maintenance personnel to the actual fault point.

Description

Technical Field

[0001] This invention relates to the field of fault location technology, and in particular to a fault location sensor for pole-mounted circuit breakers. Background Technology

[0002] With the continuous expansion of the power distribution network and the increasing demands for power supply reliability, pole-mounted circuit breakers, as key equipment for line segment protection, directly affect the fault repair time and power restoration speed through their rapid and accurate fault location.

[0003] Currently, fault monitoring and location for pole-mounted circuit breakers suffer from the following shortcomings: First, fault detection methods are limited, relying primarily on current protection operation signals or simple current over-limit judgments. This method struggles to effectively distinguish between genuine short-circuit / grounding faults and interference events such as lightning strikes or large-capacity load switching. In complex operating conditions, such as during the rainy season, it easily generates numerous false alarms, causing maintenance personnel to waste time and effort. Second, fault location relies on manual inspection and experience-based judgment. When a line trips, maintenance teams must inspect section by section, relying on visual inspection, handheld equipment testing, or experience analysis to roughly determine the faulty section. This is inefficient, especially in inclement weather or at night, making inspections difficult and time-consuming, thus extending the average power outage time for users. Third, existing monitoring devices have low integration and complex installation. Sensors, acquisition terminals, and communication modules are often separate, requiring additional power and signal lines. Construction in outdoor pole-mounted environments is difficult and costly, and the reliability of the connections is significantly affected by the environment, limiting their large-scale deployment and application. In addition, the existing devices have limited intelligence, and most of them only realize the "remote signaling" function, lacking the ability to perform on-site analysis and diagnosis, and cannot provide key information such as fault type, waveform, and accurate time stamp, making it difficult for the back-end main station to perform in-depth analysis and accurate positioning. Summary of the Invention

[0004] This invention proposes a fault location sensor for pole-mounted circuit breakers, which solves the problems in the prior art where fault location of pole-mounted circuit breakers mainly relies on the detection of a single electrical parameter and manual inspection, resulting in low location efficiency, high false alarm rate, and inability to effectively distinguish between real faults and transient interference.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A fault location sensor for a pole-mounted circuit breaker includes: The sensor module is used to synchronously acquire multiple physical quantity signals at the installation point, including at least two of the following: high-frequency current signal, power frequency current / voltage signal, electric field / magnetic field signal, and temperature signal. The processing module, connected to the sensor module, is used to perform on-site diagnosis and intelligent analysis of faults in the pole-mounted circuit breaker or its associated lines based on the synchronously acquired multi-physical quantity signals by running a built-in diagnostic program, and generate structured fault information containing fault type judgment. A communication module, connected to the processing module, is used to upload the structured fault information; A power supply module is used to supply power to the sensor module, processing module and communication module; And a protective housing for integrating the sensor module, processing module, communication module and power supply module therein.

[0006] Furthermore, the processing module runs a built-in diagnostic program, specifically performing the following steps: S1. The synchronously acquired multi-physical quantity signals are preprocessed and feature extracted to obtain a multi-dimensional feature vector containing transient features, steady-state features and non-electrical features; S2. Input the multidimensional feature vector into a preset fault diagnosis model for processing to obtain the judgment result of the current event type and the corresponding confidence level; S3. Based on the judgment result and the confidence level, generate structured fault information and trigger data recording and communication strategies corresponding to the fault level.

[0007] Furthermore, in step S1, the preprocessing and feature extraction include: For the high-frequency current signal, wavelet transform or correlation algorithm is applied to accurately extract the arrival time, amplitude, polarity and main frequency components of the first wavefront of the fault transient traveling wave, which constitute the transient features; The power frequency current / voltage signal is subjected to full-cycle Fourier calculation to extract the effective values ​​and phase angles of the fundamental wave and specific harmonics, calculate the amplitude and phase angle of the zero-sequence current and negative-sequence current, and calculate the rate of change within a preset time window to form the steady-state characteristics. For the electric / magnetic field signal, calculate its effective value, pulse count, and the ratio of peak value to average value within a specific frequency band, as characteristic quantities reflecting the insulation state and switching action; For the temperature signal, calculate its real-time value, historical average value, and linear fitting slope within a preset time period to constitute the non-electrical feature; After normalizing all extracted transient features, steady-state features, electric / magnetic field features, and non-electrical features, they are combined in a preset order to form the multidimensional feature vector used for model input.

[0008] Furthermore, in step S2, the preset fault diagnosis model is a multi-threshold logic tree model; The multi-threshold logic tree model has a hierarchical judgment structure: First, determine whether the amplitude of the traveling wave front in the transient feature exceeds the first current threshold, and whether the time difference between the arrival time of the wave front in the transient feature and the zero-crossing point of the power frequency is less than a preset time window. If so, proceed to the short circuit / ground fault discrimination branch: further determine whether the zero-sequence current in the steady-state characteristics exceeds the second current threshold. If so, it is judged as a ground fault; otherwise, it is judged as a phase-to-phase short circuit fault. If not, proceed to the interference event discrimination branch: determine whether the amplitude of the traveling wave front in the transient feature exceeds the third current threshold but the effective value of the power frequency current in the steady-state feature does not increase significantly, and whether the pulse count in the electric / magnetic field feature increases sharply in a very short time. If so, it is judged as a lightning strike event; otherwise, combine the current change rate and waveform harmonic content in the steady-state feature to determine whether it is a load impact event. For all discrimination branches, the temperature change rate in the non-electrical characteristics is used for auxiliary verification and alarm level adjustment.

[0009] Furthermore, in step S2, the preset fault diagnosis model is a trained machine learning classification model. The machine learning classification model is a support vector machine, random forest, or one-dimensional convolutional neural network model. Before deployment, the model was trained using historical multi-physical quantity sample data containing various known fault types (including short-circuit faults, ground faults, lightning strikes, load switching, and arc faults) and normal operating conditions to establish a mapping relationship from the multi-dimensional feature vectors to event type classification. The processing module is also configured to receive model parameter update files trained based on new field data from the background system during operation, so as to realize online iterative optimization of the fault diagnosis model.

[0010] Furthermore, in step S3, the triggering of the data recording and communication strategy corresponding to the fault level includes: If the judgment result is a short circuit fault or a ground fault, and the confidence level is higher than the first threshold, then the full waveform recording of the fault is initiated, and the structured fault information containing detailed waveform data, fault type, timestamp and feature values ​​is immediately uploaded through the communication module. If the judgment result is a lightning strike event, or the confidence level is lower than the first threshold but higher than the second threshold, a simplified event log is recorded and uploaded to a non-real-time channel via the communication module or temporarily stored and then uploaded periodically. If the judgment result is a load shock or a normal event, or the confidence level is lower than the second threshold, a fault alarm will not be triggered.

[0011] Furthermore, the sensor module includes: High-frequency current sensors, using Rogowski coils or high-frequency current transformers, have a bandwidth of not less than 1MHz and a sampling rate of not less than 10MHz. They are used to capture transient traveling wave signals generated by faults to achieve fault location. The power frequency current sensor uses a high-precision current transformer with an accuracy class of no less than 0.5S. It is used to collect steady-state current signals for fault severity and type analysis. Electric field sensor and / or magnetic field sensor, using a combination of a three-dimensional magnetic field sensor and an electric field induction element, is used for contactless monitoring of the opening and closing status of circuit breakers and detection of partial discharge signals; In addition, infrared temperature sensors or contact digital temperature sensors, whose temperature probes are oriented toward or attached to circuit breaker contacts or cable joints, are used to monitor the temperature rise at connection points and provide early warning of overheating faults.

[0012] Furthermore, the communication module includes: The wireless communication unit integrates a cellular network module that supports long-range wireless communication and a low-power wireless module that supports local self-organizing networks; The processing module is configured to: when a fault is diagnosed, control the cellular network module to upload the structured fault information and waveform data to a remote backend system; at the same time, control the low-power wireless module to communicate with other fault location sensors installed on the same power distribution line in real time or at regular intervals, and exchange fault characteristics and time information to collaboratively achieve a preliminary judgment of the fault section.

[0013] Furthermore, it also includes: The high-precision clock module uses a timing chip that supports satellite synchronization; The high-precision clock module provides a unified and synchronized time reference at the microsecond level or higher for the multi-channel synchronous sampling of the sensor module, the event records generated by the processing module, and the collaborative positioning of the communication module and adjacent sensors.

[0014] Furthermore, the power supply module includes: The current transformer self-powered unit includes an energy-collecting CT and a high-efficiency power management circuit, which is used to extract power from the power frequency current of the measured conductor and charge the backup energy storage unit. Backup energy storage units use supercapacitors or high-temperature lithium batteries; The power supply module is configured to: supply power to the self-powered unit and charge the backup energy storage unit when the line is operating normally; and automatically and seamlessly switch to power supply from the backup energy storage unit when the line loses power due to a fault, ensuring that the sensor continues to work and complete data upload during the critical period after the fault.

[0015] The positive effects of this invention are: It has high fault identification accuracy and strong anti-interference ability. By integrating multi-dimensional information such as high-frequency traveling waves, power frequency electrical quantities, electromagnetic fields and temperature, and using built-in intelligent diagnostic algorithms for comprehensive judgment, it can effectively distinguish between real short circuit / grounding faults and interference events such as lightning strikes and inrush currents. It significantly reduces false alarms from the root cause, avoids ineffective inspections, and accurately guides maintenance forces to the real fault point.

[0016] The sensor incorporates a high-precision satellite-synchronized clock, providing microsecond-level accurate time stamps for the fault traveling wave front. By comparing time information between multiple sensors or cooperating with the main station, dual-end ranging of the traveling wave can be achieved, improving the fault location accuracy from the traditional "a section (several kilometers)" to "near a tower (tens of meters)". Furthermore, it can quickly exchange information with adjacent sensors via local low-power wireless communication, enabling on-site collaborative assessment of the fault section and significantly shortening the fault location time.

[0017] It possesses edge intelligence and adaptive capabilities: the processing module completes the entire process from feature extraction to fault diagnosis locally on the device, achieving edge intelligence. This reduces reliance on communication bandwidth and backend computing resources, and even if communication is interrupted, complete fault information can still be recorded locally. Simultaneously, the diagnostic model supports remote online updates, enabling continuous iterative optimization using on-site operational data, giving the system self-learning and self-evolution capabilities to adapt to the characteristics and changes of different lines.

[0018] An intelligent power supply solution combining CT self-powering and supercapacitor / battery backup power is adopted. During normal operation or transient faults, the line current provides self-power. After the line loses power due to a permanent fault, it can seamlessly switch to backup power supply, ensuring that the sensor can continue to work in the most critical few minutes after the fault trip, fully record the fault waveform and upload the data, thus solving the industry problem of the device "losing its voice" due to power loss after a fault.

[0019] Not only can it locate faults after they occur, but it can also provide early warnings of potential defects such as insulation degradation, overheating of connection points, and abnormal mechanical conditions by continuously monitoring information such as partial discharge signals, temperature trends, and circuit breaker status. This promotes the transformation of the operation and maintenance mode from "post-event maintenance" to "pre-event warning and condition-based maintenance," thereby improving the health management level of power distribution network equipment. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the structure of the present invention; In the picture: 1. Protective housing; 2. Fixing bracket; 3. Fixing bolts. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention. Example

[0022] like Figure 1 As shown, the fault location sensor for the pole-mounted circuit breaker has a protective housing 1, which integrates the sensor module, processing module, communication module, and power supply module. In use, the protective housing 1 is installed into the mounting bracket 2, and then screwed onto any position of the pole-mounted circuit breaker using fixing bolts 3, enabling convenient and quick installation and adjustment of the installation position.

[0023] The modules are as follows: The sensor module is used to synchronously acquire multiple physical quantity signals at the installation point, including at least two of the following: high-frequency current signal, power frequency current / voltage signal, electric field / magnetic field signal, and temperature signal. The processing module, connected to the sensor module, is used to perform on-site diagnosis and intelligent analysis of faults in the pole-mounted circuit breaker or its associated lines based on the synchronously acquired multi-physical quantity signals by running a built-in diagnostic program, and generate structured fault information containing fault type judgment. A communication module, connected to the processing module, is used to upload the structured fault information; The power supply module is used to supply power to the sensor module, processing module and communication module.

[0024] This sensor is a highly integrated outdoor intelligent monitoring device. The sensor module, acting as the system's sensory organ, utilizes sensing elements based on various principles to simultaneously collect physical signals (electrical, magnetic, and thermal) from the primary circuit breaker and its body. The processing module, acting as the "brain," incorporates a high-performance microprocessor (such as an ARM Cortex-M series or an MCU with an AI acceleration core) and runs dedicated fault diagnosis algorithm firmware. It receives and fuses multi-channel sensor data in real time, performing intelligent analysis from raw data to fault type determination locally on the device (i.e., at the "edge"), generating a structured diagnostic report rather than the raw data stream. The communication module (such as 4G / NB-IoT+LoRa), acting as the nervous system, is responsible for uploading diagnostic results and key data to a remote master station or conducting peer-to-peer communication with adjacent nodes. The power supply module ensures the device continues to operate under extreme conditions such as power outages. All modules are sealed within a protective housing (typically cast aluminum alloy with an IP67 protection rating), forming an integrated structure that is directly mounted on the circuit breaker body, enabling plug-and-play functionality.

[0025] This integrated sensor design solves the problems of complex wiring, low reliability, and high construction costs caused by the dispersed installation of sensors, data acquisition units, and communication terminals in traditional solutions. It offloads data processing and intelligent diagnostic capabilities from the main backend station to field devices, achieving "edge intelligence," significantly reducing communication dependence and backend computing pressure, and substantially improving response speed. It also clarifies the hardware foundation for acquiring information from multiple physical quantities, providing data support for subsequent high-precision, high-reliability fault identification and location.

[0026] The processing module runs a built-in diagnostic program, specifically performing the following steps: S1. The synchronously acquired multi-physical quantity signals are preprocessed and feature extracted to obtain a multi-dimensional feature vector containing transient features, steady-state features and non-electrical features; S2. Input the multidimensional feature vector into a preset fault diagnosis model for processing to obtain the judgment result of the current event type and the corresponding confidence level; S3. Based on the judgment result and the confidence level, generate structured fault information and trigger data recording and communication strategies corresponding to the fault level.

[0027] Specifically, the diagnostic program is stored in the processing module's FLASH memory. After power-on, the program enters a loop: Step S1: The raw ADC sampled data is first digitally filtered (e.g., low-pass, band-stop) to remove noise and power frequency interference. Then, different feature extraction functions are called for different signals. For example, wavelet transform is used to extract traveling wave features for high-frequency current signals; Fast Fourier Transform (FFT) is used to calculate amplitude and phase for power frequency signals. All extracted feature values ​​(digital quantities) are organized into a fixed-dimensional array, i.e., the "feature vector".

[0028] Step S2: Use the constructed feature vector as input parameters to call the fault diagnosis model function. This function may be a complex set of if-else logic judgments (corresponding to a logic tree model), or an inference interface of a trained machine learning model (such as calling the TensorFlow Lite Micro library). After the function executes, it returns two core outputs: event_type (e.g., 0-normal, 1-phase-to-ground, 2-phase-to-phase short circuit, 3-lightning strike, etc.) and confidence (a floating-point number between 0 and 1, representing the confidence level of the judgment).

[0029] Step S3: Based on event_type and confidence, the program enters different branches. For example, if event_type=1 and confidence>0.9, the "high-priority fault handling subroutine" is triggered. This subroutine will immediately retrieve and save the complete fault waveform data from the cache, package it into a JSON format message containing time, location, type, characteristic value, and confidence level, and mark it as "urgent" before sending it through the communication module.

[0030] This paper defines a standard process for edge intelligent diagnostics, breaking down complex fault analysis tasks into three clear steps: feature extraction, model inference, and decision output. This makes the diagnostic process repeatable and verifiable, and provides a clear interface for subsequent model upgrades (such as updating the function in step S2), forming the core framework of the software method claims.

[0031] In step S1, the preprocessing and feature extraction include: For the high-frequency current signal, wavelet transform or correlation algorithm is applied to accurately extract the arrival time, amplitude, polarity and main frequency components of the first wavefront of the fault transient traveling wave, which constitute the transient features; The power frequency current / voltage signal is subjected to full-cycle Fourier calculation to extract the effective values ​​and phase angles of the fundamental wave and specific harmonics, calculate the amplitude and phase angle of the zero-sequence current and negative-sequence current, and calculate the rate of change within a preset time window to form the steady-state characteristics. For the electric / magnetic field signal, calculate its effective value, pulse count, and the ratio of peak value to average value within a specific frequency band, as characteristic quantities reflecting the insulation state and switching action; For the temperature signal, calculate its real-time value, historical average value, and linear fitting slope within a preset time period to constitute the non-electrical feature; After normalizing all extracted transient features, steady-state features, electric / magnetic field features, and non-electrical features, they are combined in a preset order to form the multidimensional feature vector used for model input.

[0032] High-frequency signal processing: Wavelet transform (such as using Daubechies wavelets) can perform localized analysis of signals simultaneously in the time and frequency domains, making it particularly suitable for extracting abrupt change points (wavefront arrival time t0) and singular values ​​(wavefront amplitude A) of non-stationary signals such as transient traveling waves. Correlation algorithms (such as cross-correlation with a standard traveling wave template) can effectively improve the accuracy of wavefront timing detection in noisy environments. The dominant frequency components of the traveling wave (such as those obtained through wavelet coefficients or spectral analysis) help distinguish fault types (e.g., lightning strikes have richer frequency components).

[0033] Power frequency signal processing: A full-cycle Fourier transform (DFT / FFT) algorithm is employed to ensure accurate calculation of the amplitude and phase of the fundamental frequency (50Hz) and all harmonics even during asynchronous sampling, unaffected by non-integer harmonics. Zero-sequence current 3I0 and negative-sequence current I2 are key characteristics distinguishing between ground faults and phase-to-phase / three-phase imbalances. Calculating rates of change (e.g., di / dt) can sensitively capture sudden changes in current.

[0034] Electromagnetic signal processing: Partial discharge (PD) and circuit breaker operation generate high-frequency electromagnetic pulses. By extracting the signal in this frequency band using a digital bandpass filter (e.g., 300kHz-3MHz), and calculating its RMS value, pulse count per unit time, and peak / average ratio (Crest Factor), the severity and pattern of the discharge can be quantified for insulation warning.

[0035] Temperature signal processing: We focus not only on the real-time temperature T, but also on its trend. Linear regression is performed on the temperature data points from the most recent N minutes to obtain the slope k. A positive and large k value indicates a rapid temperature rise, potentially indicating poor contact. Comparison with historical averages can identify anomalies.

[0036] Feature fusion: The feature values ​​of the different dimensions (time, current, temperature, dimensionless ratio, etc.) are mapped to the [0,1] or standard normal distribution interval through max-min normalization or Z-score normalization. Then, they are concatenated into a one-dimensional feature vector in a predetermined order (e.g., [t0, A, f_peak, Ia, Ib, Ic, 3I0, I2, PD_count, T, k, ...]) as the standard input of the model.

[0037] It extracts the most effective and computable information for fault diagnosis from the original signal, integrates time domain, frequency domain, and time-frequency domain analysis methods, and the extracted feature set has clear physical meaning. It can comprehensively characterize various abnormal states such as short circuit, grounding, overload, partial discharge, and overheating, providing high-quality and standardized input materials for high-precision intelligent diagnosis.

[0038] In step S2, the preset fault diagnosis model is a multi-threshold logic tree model; The multi-threshold logic tree model has a hierarchical judgment structure: First, determine whether the amplitude of the traveling wave front in the transient feature exceeds the first current threshold, and whether the time difference between the arrival time of the wave front in the transient feature and the zero-crossing point of the power frequency is less than a preset time window. If so, proceed to the short circuit / ground fault discrimination branch: further determine whether the zero-sequence current in the steady-state characteristics exceeds the second current threshold. If so, it is judged as a ground fault; otherwise, it is judged as a phase-to-phase short circuit fault. If not, proceed to the interference event discrimination branch: determine whether the amplitude of the traveling wave front in the transient feature exceeds the third current threshold but the effective value of the power frequency current in the steady-state feature does not increase significantly, and whether the pulse count in the electric / magnetic field feature increases sharply in a very short time. If so, it is judged as a lightning strike event; otherwise, combine the current change rate and waveform harmonic content in the steady-state feature to determine whether it is a load impact event. For all discrimination branches, the temperature change rate in the non-electrical characteristics is used for auxiliary verification and alarm level adjustment.

[0039] The logic tree model is represented in code as a set of nested if-else conditional statements. Its implementation flow is as follows: Level 1 Judgment (Transient Start-up): Check if the amplitude of the high-frequency traveling wave front, A_wave, is greater than I_th1 (e.g., 10 times the peak value of the rated current), and if the time difference Δt between the arrival time of the wave front and the most recent zero-crossing point of the power frequency voltage is less than T_window (e.g., 1 ms). This condition is used to quickly capture strong transient processes accompanied by sudden changes in power frequency current. If satisfied, it is initially judged as a "serious fault" and proceeds to the next level. Otherwise, it jumps to the "interference event discrimination branch".

[0040] Second-level judgment (fault type subdivision): Under the "serious fault" branch, further check whether the steady-state zero-sequence current 3I0 is greater than I_th2 (ground fault initiation threshold). If 3I0>I_th2, it is judged as a ground fault; otherwise, it is judged as a phase-to-phase short circuit fault.

[0041] Interference event identification: For events that do not meet the first-level criteria but have obvious transient signals (A_wave > I_th3, I_th3 value is small), check whether the steady-state power frequency current effective value I_rms is within the normal range. If I_rms does not increase significantly, but the electric field / magnetic field pulse count PD_count increases sharply within 1-2 power frequency cycles, it is determined to be a lightning overvoltage. Otherwise, combine the current change rate di / dt (gradual) and harmonic content (such as prominent 5th and 7th harmonics) characteristics to determine whether it is a load impact such as motor starting.

[0042] Temperature-assisted verification: At the end of all decision branches, read the temperature change slope k_temp. If k_temp exceeds the warning threshold, regardless of the current event, a "risk of overheating" label will be added to the final structured information, and the alarm level may be upgraded.

[0043] This provides a diagnostic method that is well-defined, highly interpretable, and requires low computational resources. Its hierarchical structure mimics the reasoning logic of human experts: first, assess the magnitude of the impact (transient), then determine the nature of the problem (steady-state component), while also considering other clues (electromagnetic, temperature). By setting scientifically designed multiple thresholds, it can effectively distinguish easily confused events (such as short circuits vs. lightning strikes), exhibiting high reliability and making it particularly suitable for implementation in embedded devices with high reliability requirements and limited computing power.

[0044] In step S2, the preset fault diagnosis model is a trained machine learning classification model; The machine learning classification model is a support vector machine, random forest, or one-dimensional convolutional neural network model. Before deployment, the model was trained using historical multi-physical quantity sample data containing various known fault types (including short-circuit faults, ground faults, lightning strikes, load switching, and arc faults) and normal operating conditions to establish a mapping relationship from the multi-dimensional feature vectors to event type classification. The processing module is also configured to receive model parameter update files trained based on new field data from the background system during operation, so as to realize online iterative optimization of the fault diagnosis model.

[0045] Model selection and training: SVM / Random Forest: Suitable for processing the "feature vectors" constructed in claim 3. During the R&D phase, the model is trained on a PC server using a large amount of labeled historical data (feature vectors + fault type labels). After training, the model parameters (such as the support vectors and coefficients of SVM, and the tree structure of random forest) are exported as files.

[0046] 1D-CNN: Can directly process raw signal sequences or sequences with simple preprocessing. The convolutional layers of CNN can automatically learn from the data and extract deeper features, and may have a stronger ability to recognize waveform patterns.

[0047] Model Deployment: The trained model parameter file is integrated into the firmware of the processing module. During device operation, the diagnostic program calls a lightweight inference engine (such as the CMSIS-NN library) to load the model parameters and inputs real-time generated feature vectors (for SVM / random forest) or signal fragments (for 1D-CNN). The engine automatically outputs the classification results and probabilities.

[0048] Online iterative optimization: While the equipment is operating in the field, it transmits diagnostic results and raw data (when communication permits) back to the cloud. The cloud uses the newly collected data, combined with manually reviewed labels, to periodically retrain or fine-tune the model, generating parameter update files. The main station securely distributes the new model parameter files to the field equipment via a communication module (e.g., 4G), and the equipment automatically updates its local model when idle. This forms a closed loop of "field data acquisition -> cloud model optimization -> edge model update".

[0049] An adaptive and high-precision diagnostic path is provided. Machine learning models, especially deep learning models, can learn complex and nonlinear feature relationships and may discover patterns that are difficult to describe by human-defined rules, thus achieving higher classification accuracy under complex working conditions. The online update mechanism enables the entire system to "get smarter with use," adapting to different line characteristics and gradually changing operating environments, which is key to the long-term viability and advanced nature of the technical solution.

[0050] In step S3, the triggering of the data recording and communication strategy corresponding to the fault level includes: If the judgment result is a short circuit fault or a ground fault, and the confidence level is higher than the first threshold, then the full waveform recording of the fault is initiated, and the structured fault information containing detailed waveform data, fault type, timestamp and feature values ​​is immediately uploaded through the communication module. If the judgment result is a lightning strike event, or the confidence level is lower than the first threshold but higher than the second threshold, a simplified event log is recorded and uploaded to a non-real-time channel via the communication module or temporarily stored and then uploaded periodically. If the judgment result is a load shock or a normal event, or the confidence level is lower than the second threshold, a fault alarm will not be triggered.

[0051] This is a graded response mechanism based on diagnostic results. The program internally sets two confidence thresholds, for example, Conf_high=0.85 and Conf_medium=0.6.

[0052] High-priority fault response: When the diagnostic result is a short circuit / ground fault and the confidence level is >0.85, the highest priority response is triggered. Immediately lock and store all raw sampling data (full waveform recording) for the N cycles before the fault and the M cycles after the fault. Simultaneously, immediately wake up the cellular network module and upload a message containing the waveform recording file and structured diagnostic results to the main station with the highest priority (e.g., highest QoS level), which may trigger immediate alarms such as SMS messages.

[0053] Medium-level event response: For lightning strikes or low-confidence faults (0.6 < confidence < 0.85), only simplified logs (such as time, event type, and key characteristic values) are recorded, without recording complete waveforms. These logs are uploaded during idle periods via low-power NB-IoT or periodically via 4G, or stored locally initially and uploaded in batches when communication is idle. No immediate alarms are triggered; event records are only generated on the main station system.

[0054] Low-level / normal response: For load shocks or events with a confidence level <0.6, only internal state variables are updated, without any recording or active reporting, and only a state summary is carried in the heartbeat packet.

[0055] It enables intelligent management of communication and equipment resources. This avoids channel congestion and bandwidth waste caused by uploading all data at once, and also reduces the burden of processing invalid data in the backend system. It ensures that the most critical fault information can be reported in the fastest and most complete form, thereby optimizing the efficiency and reliability of the entire system.

[0056] The sensor module includes: High-frequency current sensors, using Rogowski coils or high-frequency current transformers, have a bandwidth of not less than 1MHz and a sampling rate of not less than 10MHz. They are used to capture transient traveling wave signals generated by faults to achieve fault location. The power frequency current sensor uses a high-precision current transformer with an accuracy class of no less than 0.5S. It is used to collect steady-state current signals for fault severity and type analysis. Electric field sensor and / or magnetic field sensor, using a combination of a three-dimensional magnetic field sensor and an electric field induction element, is used for contactless monitoring of the opening and closing status of circuit breakers and detection of partial discharge signals; In addition, infrared temperature sensors or contact digital temperature sensors, whose temperature probes are oriented toward or attached to circuit breaker contacts or cable joints, are used to monitor the temperature rise at connection points and provide early warning of overheating faults.

[0057] High-frequency current sensor (Rogowski coil): This is an air-core coil whose output signal is di / dt (the derivative of current with respect to time). Its core advantages are no magnetic saturation and a wide response bandwidth (up to several MHz). The small voltage signal output by the coil is restored to a current signal by an integrator circuit, and then sampled by a high-speed ADC (≥10MSPS) to perfectly capture the nanosecond-level rising fault traveling wave, which is the physical basis for achieving traveling wave localization.

[0058] Power frequency current sensor (high-precision CT): It adopts high-performance magnetic cores such as permalloy and has extremely high measurement accuracy (0.5S level) and linearity near the power frequency. It is used to accurately measure the effective value and phase of steady-state current and serves as the basis for fault type identification (such as judging grounding through zero-sequence current) and protection coordination.

[0059] Electromagnetic sensors: Three-dimensional magnetoresistive sensors (such as the HMC5883L) can measure changes in spatial magnetic field vectors, used for non-contact determination of circuit breaker contact positions (open / closed). Electric field sensing elements detect changes in the electric field near a conductor; combined with a high-frequency amplifier, they can capture high-frequency electromagnetic waves radiated by pulse currents generated by partial discharge. The combination of these two technologies enables monitoring of switch status and insulation status.

[0060] Temperature sensors: Infrared thermometry is non-contact, installed inside the sensor housing, and measured by looking through the viewing window at the circuit breaker contacts to measure their surface temperature. Contact-type digital sensors (such as the DS18B20) require thermal grease to be tightly attached to the heat-prone connection points. Real-time temperature monitoring is a direct means of preventing thermal failures caused by increased contact resistance.

[0061] A comprehensive, high-performance sensing system has been constructed. The combination of Rogowski coils and high-precision CT scanners addresses the dual high requirements of transient and steady-state measurements. The addition of electromagnetic and temperature sensors extends the monitoring scope from "faults" to "condition warnings," achieving a leap from "passively responding to faults" to "actively preventing risks," embodying the concept of condition monitoring.

[0062] The communication module includes: The wireless communication unit integrates a cellular network module that supports long-range wireless communication and a low-power wireless module that supports local self-organizing networks; The processing module is configured to: when a fault is diagnosed, control the cellular network module to upload the structured fault information and waveform data to a remote backend system; at the same time, control the low-power wireless module to communicate with other fault location sensors installed on the same power distribution line in real time or at regular intervals, and exchange fault characteristics and time information to collaboratively achieve a preliminary judgment of the fault section.

[0063] The communication module integrates two communication chips: one is a 4G Cat.1 or NB-IoT module (such as Quectel EC200U), which is responsible for communicating with the Internet cloud platform; the other is a LoRa or Zigbee module (such as SX1278), which is responsible for point-to-point or self-organizing network communication between similar sensors on adjacent towers within a range of 1-2 kilometers.

[0064] Cellular uplink: Used to report diagnostic results, waveform data, equipment status, etc. to the remote monitoring master station or cloud platform, and is the link of "cloud-edge collaboration".

[0065] Low-power wireless peer-to-peer communication: When a sensor (e.g., S1) detects a fault traveling wave, it immediately broadcasts a short message via LoRa containing a "fault signature code" and the "precise fault time t1" (from a high-precision clock in claim 9). Upon receiving this message, adjacent upstream and downstream sensors (S0, S2) record the times t0 and t2 when they detected the same fault traveling wave. By comparing the order of t0, t1, and t2 or calculating the time difference, several sensors can preliminarily determine whether the fault point is located between S0 and S1 or between S1 and S2 without the involvement of a master station, achieving distributed collaborative analysis and quickly locating the fault section.

[0066] The dual-mode communication architecture combines the advantages of wide-area coverage with local low-latency collaboration. The cellular network ensures the remote accessibility of data, while local low-power wireless communication enables "neighborhood collaboration" between sensors, greatly accelerating the initial location process of faulty sections, reducing complete reliance on master station computing, and improving system robustness and response speed.

[0067] Also includes: The high-precision clock module uses a timing chip that supports satellite synchronization; The high-precision clock module provides a unified and synchronized time reference at the microsecond level or higher for the multi-channel synchronous sampling of the sensor module, the event records generated by the processing module, and the collaborative positioning of the communication module and adjacent sensors.

[0068] Specifically, a BeiDou / GPS dual-mode timing chip (such as the Hexin Xingtong UM220) is used. This chip receives 1PPS (pulses per second) signals and UTC time messages transmitted by satellites every second. The processing module uses this clock signal to provide a precise sampling clock reference for the synchronous sampling of multiple ADCs, ensuring strict simultaneity of data from all channels; on the other hand, it assigns a uniform microsecond-level timestamp to each sampling point and each event.

[0069] For traveling wave positioning: The fault traveling wave propagates along the line at near the speed of light. Sensors S1 and S2 on either side of the fault point record the local arrival times T1 and T2 of the traveling wave. Since T1 and T2 are based on the same UTC time base, the master station, upon receiving these two times, can calculate the precise distance to the fault point using the formula L = v * (T2 - T1) / 2 (assuming the fault point is in the middle) or a two-ended formula. Time synchronization error is the decisive factor in the accuracy of traveling wave ranging; microsecond-level errors will lead to positioning errors on the order of hundreds of meters. Therefore, a high-precision clock is crucial for achieving accurate positioning within tens of meters.

[0070] It provides a unified "heartbeat" and "time scale" for the entire system. It is a cornerstone technology for realizing multi-sensor data fusion analysis (ensuring data time alignment) and precise wave-based positioning, elevating the local time of sensors to an absolute time system, enabling distributed devices to work collaboratively.

[0071] The power supply module includes: The current transformer self-powered unit includes an energy-collecting CT and a high-efficiency power management circuit, which is used to extract power from the power frequency current of the measured conductor and charge the backup energy storage unit. Backup energy storage units use supercapacitors or high-temperature lithium batteries; The power supply module is configured to: supply power to the self-powered unit and charge the backup energy storage unit when the line is operating normally; and automatically and seamlessly switch to power supply from the backup energy storage unit when the line loses power due to a fault, ensuring that the sensor continues to work and complete data upload during the critical period after the fault.

[0072] Self-powered unit: An independent energy-harvesting current transformer (CT) is wound around the primary conductor, inducing alternating current from the conductor current. This alternating current is converted into a stable direct current voltage (such as 12V or 5V) through a high-efficiency power management circuit (including a rectifier bridge, surge protection, MPPT maximum power point tracking, DC-DC voltage regulator, etc.) to power the entire sensor and charge the backup energy storage unit. MPPT technology can optimize energy harvesting efficiency under different primary currents (such as 10A-600A).

[0073] Backup energy storage units: These employ supercapacitors (high power density, long cycle life) or high-temperature lithium batteries (high energy density). They are connected to the self-powered DC bus via a charging management circuit.

[0074] Intelligent switching: The power management circuit continuously monitors the output voltage of the self-supply unit. When the line is normal and the supply voltage is stable, the power supply unit provides power and float charges the supercapacitor / battery. When a line fault trips, the primary current disappears, and the supply voltage drops, the power management circuit seamlessly switches to supercapacitor / battery discharge within milliseconds, maintaining system operation for several minutes to several hours, ensuring the recording, processing, and uploading of fault data. After the fault is cleared and power is restored, the system automatically switches back to self-supply mode.

[0075] This solution addresses the ultimate challenge of providing long-term, stable power to outdoor pole-mounted equipment. The CT-based power supply achieves energy self-sufficiency and requires no maintenance. Backup energy storage addresses the industry pain point of sudden device failure and loss of critical fault data due to line power outages, ensuring continuous and reliable monitoring—a key feature of practical design.

[0076] The above-described embodiments are detailed and specific, illustrating preferred embodiments of the present invention. They are only used to illustrate the technical ideas and features of the present invention, with the aim of enabling those skilled in the art to understand the content of the present invention and implement it accordingly. However, they are not limited to the present invention, and the patent scope of the present invention cannot be limited by this embodiment alone. That is, any equivalent changes or modifications made to the spirit disclosed in the present invention, without departing from the structure of the present invention, such as local improvements within the system and modifications or transformations between subsystems, are still within the patent scope of the present invention.

Claims

1. A fault location sensor for a pole-mounted circuit breaker, characterized in that, include: The sensor module is used to synchronously acquire multiple physical quantity signals at the installation point, including at least two of the following: high-frequency current signal, power frequency current / voltage signal, electric field / magnetic field signal, and temperature signal. The processing module, connected to the sensor module, is used to perform on-site diagnosis and intelligent analysis of faults in the pole-mounted circuit breaker or its associated lines based on the synchronously acquired multi-physical quantity signals by running a built-in diagnostic program, and generate structured fault information containing fault type judgment. A communication module, connected to the processing module, is used to upload the structured fault information; A power supply module is used to supply power to the sensor module, processing module and communication module; And a protective housing (1) for integrating the sensor module, processing module, communication module and power supply module therein.

2. The fault location sensor for a pole-mounted circuit breaker according to claim 1, characterized in that, The processing module runs a built-in diagnostic program, specifically performing the following steps: S1. The synchronously acquired multi-physical quantity signals are preprocessed and feature extracted to obtain a multi-dimensional feature vector containing transient features, steady-state features and non-electrical features; S2. Input the multidimensional feature vector into a preset fault diagnosis model for processing to obtain the judgment result of the current event type and the corresponding confidence level; S3. Based on the judgment result and the confidence level, generate structured fault information and trigger data recording and communication strategies corresponding to the fault level.

3. A fault location sensor for a pole-mounted circuit breaker according to claim 2, characterized in that, In step S1, the preprocessing and feature extraction include: For the high-frequency current signal, wavelet transform or correlation algorithm is applied to accurately extract the arrival time, amplitude, polarity and main frequency components of the first wavefront of the fault transient traveling wave, which constitute the transient features; The power frequency current / voltage signal is subjected to full-cycle Fourier calculation to extract the effective values ​​and phase angles of the fundamental wave and specific harmonics, calculate the amplitude and phase angle of the zero-sequence current and negative-sequence current, and calculate the rate of change within a preset time window to form the steady-state characteristics. For the electric / magnetic field signal, calculate its effective value, pulse count, and the ratio of peak value to average value within a specific frequency band, as characteristic quantities reflecting the insulation state and switching action; For the temperature signal, calculate its real-time value, historical average value, and linear fitting slope within a preset time period to constitute the non-electrical feature; After normalizing all extracted transient features, steady-state features, electric / magnetic field features, and non-electrical features, they are combined in a preset order to form the multidimensional feature vector used for model input.

4. A fault location sensor for a pole-mounted circuit breaker according to claim 2, characterized in that, In step S2, the preset fault diagnosis model is a multi-threshold logic tree model; The multi-threshold logic tree model has a hierarchical judgment structure: First, determine whether the amplitude of the traveling wave front in the transient feature exceeds the first current threshold, and whether the time difference between the arrival time of the wave front in the transient feature and the zero-crossing point of the power frequency is less than a preset time window. If so, proceed to the short circuit / ground fault discrimination branch: further determine whether the zero-sequence current in the steady-state characteristics exceeds the second current threshold. If so, it is judged as a ground fault; otherwise, it is judged as a phase-to-phase short circuit fault. If not, proceed to the interference event discrimination branch: determine whether the amplitude of the traveling wave front in the transient feature exceeds the third current threshold but the effective value of the power frequency current in the steady-state feature does not increase significantly, and whether the pulse count in the electric / magnetic field feature increases sharply in a very short time. If so, it is judged as a lightning strike event; otherwise, combine the current change rate and waveform harmonic content in the steady-state feature to determine whether it is a load impact event. For all discrimination branches, the temperature change rate in the non-electrical characteristics is used for auxiliary verification and alarm level adjustment.

5. A fault location sensor for a pole-mounted circuit breaker according to claim 2, characterized in that, In step S2, the preset fault diagnosis model is a trained machine learning classification model; The machine learning classification model is a support vector machine, random forest, or one-dimensional convolutional neural network model. Before deployment, the model was trained using historical multi-physical quantity sample data containing various known fault types and normal operating conditions to establish a mapping relationship from the multi-dimensional feature vector to event type classification. The processing module is also configured to receive model parameter update files trained based on new field data from the background system during operation, so as to realize online iterative optimization of the fault diagnosis model.

6. A fault location sensor for a pole-mounted circuit breaker according to claim 2, characterized in that, In step S3, the triggering of the data recording and communication strategy corresponding to the fault level includes: If the judgment result is a short circuit fault or a ground fault, and the confidence level is higher than the first threshold, then the full waveform recording of the fault is initiated, and the structured fault information containing detailed waveform data, fault type, timestamp and feature values ​​is immediately uploaded through the communication module. If the judgment result is a lightning strike event, or the confidence level is lower than the first threshold but higher than the second threshold, a simplified event log is recorded and uploaded to a non-real-time channel via the communication module or temporarily stored and then uploaded periodically. If the judgment result is a load shock or a normal event, or the confidence level is lower than the second threshold, a fault alarm will not be triggered.

7. A fault location sensor for a pole-mounted circuit breaker according to claim 1, characterized in that, The sensor module includes: High-frequency current sensors, using Rogowski coils or high-frequency current transformers, have a bandwidth of not less than 1MHz and a sampling rate of not less than 10MHz. They are used to capture transient traveling wave signals generated by faults to achieve fault location. The power frequency current sensor uses a high-precision current transformer with an accuracy class of no less than 0.5S. It is used to collect steady-state current signals for fault severity and type analysis. Electric field sensor and / or magnetic field sensor, using a combination of a three-dimensional magnetic field sensor and an electric field induction element, is used for contactless monitoring of the opening and closing status of circuit breakers and detection of partial discharge signals; In addition, infrared temperature sensors or contact digital temperature sensors, whose temperature probes are oriented toward or attached to circuit breaker contacts or cable joints, are used to monitor the temperature rise at connection points and provide early warning of overheating faults.

8. A fault location sensor for a pole-mounted circuit breaker according to claim 1, characterized in that, The communication module includes: The wireless communication unit integrates a cellular network module that supports long-range wireless communication and a low-power wireless module that supports local self-organizing networks; The processing module is configured to: when a fault is diagnosed, control the cellular network module to upload the structured fault information and waveform data to a remote backend system; at the same time, control the low-power wireless module to communicate with other fault location sensors installed on the same power distribution line in real time or at regular intervals, and exchange fault characteristics and time information to collaboratively achieve a preliminary judgment of the fault section.

9. A fault location sensor for a pole-mounted circuit breaker according to claim 1, characterized in that, Also includes: The high-precision clock module uses a timing chip that supports satellite synchronization; The high-precision clock module provides a unified and synchronized time reference at the microsecond level or higher for the multi-channel synchronous sampling of the sensor module, the event records generated by the processing module, and the collaborative positioning of the communication module and adjacent sensors.

10. A fault location sensor for a pole-mounted circuit breaker according to claim 1, characterized in that, The power supply module includes: The current transformer self-powered unit includes an energy-collecting CT and a high-efficiency power management circuit, which is used to extract power from the power frequency current of the measured conductor and charge the backup energy storage unit. Backup energy storage units use supercapacitors or high-temperature lithium batteries; The power supply module is configured to: supply power to the self-powered unit and charge the backup energy storage unit when the line is operating normally; and automatically and seamlessly switch to power supply from the backup energy storage unit when the line loses power due to a fault, ensuring that the sensor continues to work and complete data upload during the critical period after the fault.

Images