Arc fault detection method and apparatus, computer readable storage medium and system

Abstract

This application discloses an arc fault detection method, apparatus, computer-readable storage medium, and system. The method includes constructing a stochastic dynamic system model of the arc fault in the circuit to be detected, the stochastic dynamic system model including the state equation and observation equation of the arc fault; acquiring sampled data of the circuit to be detected in the current observation period, extracting features from the sampled data to obtain a multidimensional observation vector of the current observation period that matches the observation equation; based on the stochastic dynamic system model and the multidimensional observation vector of the current observation period, using a Bayesian filter as a framework, performing state estimation through a continuous-discrete extended Kalman filter to obtain the posterior state vector of the current observation period; and determining whether a fault arc exists in the current observation period through a preset fault decision scheme. This application provides a novel implementation path for high-precision, high-reliability, and low-cost AC series, parallel, and grounding arc fault detection.

Description

Technical Field

[0001] This invention relates to the field of electric arc detection technology in power systems, and in particular to an electric arc fault detection method and device, computer-readable storage medium and system. Background Technology

[0002] Electrical fires are one of the major disasters threatening public safety, and arc faults are the primary cause of electrical fires. Arc faults are usually caused by insulation damage, loose connections, conductor breakage, or internal defects in electrical circuits. When an arc fault occurs, it produces a continuous high-temperature discharge, which can easily ignite surrounding flammable materials.

[0003] With the construction of new power systems and the widespread use of diversified electrical equipment, the current waveforms in power distribution networks have become more complex, with a significant increase in nonlinear and impulsive loads. This makes it easier for abnormal signals generated by arc faults to be masked by current changes in normal loads, resulting in accurate diagnosis of arc faults becoming a major technical bottleneck restricting the effective prevention of electrical fires. Especially in critical areas, arc faults are difficult to detect due to their concealment and randomness, yet the safety requirements are extremely high. To address these issues, the industry has developed various arc fault detection (AFD) methods. However, AC arc current signals have characteristics such as nonlinearity, non-stationarity, and randomness. Traditional overcurrent protection devices based on fixed thresholds (such as circuit breakers and fuses) cannot effectively identify arc faults in conjunction with normal transient processes such as motor starting and switch switching, posing a serious risk of misjudgment or missed detection.

[0004] Therefore, developing a solution that can accurately, quickly, and reliably extract and identify fault arc characteristics from complex load currents is of great practical significance and economic value for improving electrical safety and preventing electrical fires. Summary of the Invention

[0005] In view of the shortcomings of the prior art described above, the purpose of this application is to provide an arc fault detection method, device and system, computer-readable storage medium and terminal to solve the problem that the existing fault arc detection schemes cannot effectively identify fault arcs and normal transient arcs, and there is a serious risk of misjudgment or omission.

[0006] The first aspect of this application provides an arc fault detection method, including:

[0007] A stochastic dynamic system model of an arc fault in a circuit to be detected is constructed, wherein the stochastic dynamic system model includes the state equation and observation equation of the arc fault; Acquire the sampling data of the circuit under test in the current observation period, and perform feature extraction on the sampling data to obtain a multi-dimensional observation vector of the current observation period that matches the observation equation; Based on the stochastic dynamic system model and the multidimensional observation vector of the current observation period, using a Bayesian filter as a framework, state estimation is performed through a continuous-discrete extended Kalman filter to obtain the posterior state vector of the current observation period. Based on the posterior state vector of the current observation period, a preset fault decision scheme is used to determine whether a fault arc exists in the current observation period. If it exists, a trip command is sent to the protection actuator so that the protection actuator executes the trip command to control the circuit under test to disconnect the power. Otherwise, the sampling data of the circuit under test in the next observation period is reacquired, and a corresponding judgment is made on whether a fault arc exists.

[0008] In some embodiments of this application, based on the stochastic dynamic system model and the multidimensional observation vector of the current observation period, using a Bayesian filter as a framework, state estimation is performed through a continuous-discrete extended Kalman filter to obtain the posterior state vector of the current observation period, including: Based on the posterior state vector and posterior probability density of the circuit under test in the previous observation period, the prior state vector and prior probability density of the current observation period are obtained through a nonlinear Bayesian filter. The prior state vector of the current observation period is mapped through the observation equation in the stochastic dynamic system model to obtain the multidimensional observation prediction vector of the current observation period. Based on the multidimensional observation vector and the multidimensional observation prediction vector of the current observation period, the posterior state vector and the posterior probability density of the current observation period are obtained through a continuous-discrete extended Kalman filter.

[0009] In some embodiments of this application, the sampling data includes high-frequency current data, power frequency current data, and power frequency voltage data synchronously acquired by a multi-channel analog converter.

[0010] In some embodiments of this application, feature extraction of the sampled data includes: Feature extraction is performed on the high-frequency current data, the power frequency current data, and the power frequency voltage data using at least one of the following methods: frequency domain feature extraction, time domain feature extraction, and time-frequency domain feature extraction.

[0011] In some embodiments of this application, the time-domain features include at least one of rising edge steepness and falling edge steepness; the frequency-domain features include at least one of preset frequency band energy distribution, high-frequency noise spectrum features, current zero-crossing distortion rate, and high-frequency pulse count; the time-frequency domain features include at least one of short-time Fourier transform features, wavelet transform features, and time-frequency distribution entropy.

[0012] In some embodiments of this application, determining whether a fault arc exists in the current observation period based on the posterior state vector of the current observation period using a preset fault decision scheme includes: Based on the state characteristics corresponding to the high-frequency current data in the posterior state vector of the current observation period, a pre-set fault decision scheme is used to determine whether a fault arc exists in the current observation period.

[0013] In some embodiments of this application, determining whether a fault arc exists in the current observation period based on the state characteristics corresponding to the high-frequency current data in the posterior state vector of the current observation period and through a preset fault decision scheme includes: The arc energy state estimate and current distortion state estimate corresponding to the high-frequency current data are obtained from the posterior state vector of the current observation period. It is determined whether the standard state estimate obtained based on the arc energy state estimate of the current observation period is greater than the first preset threshold and whether the current distortion state estimate of the current observation period is greater than the second preset threshold. If so, it is determined that there is a fault arc in the circuit to be detected; otherwise, it is determined that the circuit to be detected is normal. The standard state estimate obtained based on the arc energy state estimate of the current observation period is three times the standard deviation of the arc energy state estimates of all observation periods within a preset time period, where the preset time period includes the current observation period.

[0014] A second aspect of this application provides an arc fault detection device, including a model building module, a feature extraction module, a posterior state vector acquisition module, and a fault decision module. The model building module is used to build a stochastic dynamic system model of the arc fault in the circuit to be detected. The stochastic dynamic system model includes the state equation and observation equation of the arc fault. The feature extraction module is used to acquire the sampling data of the current observation period of the circuit to be detected, and to perform feature extraction on the sampling data to obtain a multi-dimensional observation vector of the current observation period that matches the observation equation. The posterior state vector acquisition module is used to obtain the posterior state vector of the current observation period based on the stochastic dynamic system model and the multidimensional observation vector of the current observation period, using a Bayesian filter as a framework and a continuous-discrete extended Kalman filter for state estimation. The fault decision module is used to determine whether there is a fault arc in the current observation period based on the posterior state vector of the current observation period and through a preset fault decision scheme. If there is, a trip command is sent to the protection actuator so that the protection actuator executes the trip command to control the circuit under test to disconnect the power. Otherwise, the sampling data of the circuit under test in the next observation period is re-acquired, and the corresponding judgment on whether there is a fault arc is made.

[0015] A third aspect of this application provides an arc fault detection system, including a cloud service platform, an arc fault detection device, and a protection actuator that are interconnected, wherein the arc fault detection device is the aforementioned arc fault detection device. The cloud service platform is used to receive sampling data sent by the arc fault detection device and tripping status feedback data sent by the protection actuator.

[0016] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described above.

[0017] Compared with the prior art, one or more embodiments of the above solutions may have the following advantages or beneficial effects: The arc fault detection method provided in this invention improves detection accuracy while meeting real-time requirements by extracting and fusing multi-dimensional features from multi-channel synchronously acquired sampling data. This provides multi-dimensional analysis data for subsequent algorithms, enhancing the basis for judgment. A stochastic dynamic system model is introduced, and historical information is dynamically and recursively fused with current multi-dimensional observation data using Bayesian filters and continuous-discrete extended Kalman filters to obtain the optimal fault state estimate. This results in diagnostic results with "memory" and "smoothness," enabling accurate identification of fault arcs in the circuit under test and effectively suppressing transient noise interference. Furthermore, the data processing not only outputs the fault state estimate but also its probability density, providing rich and reliable scientific evidence for fault arc judgment and achieving a perfect combination of theoretical advancement and engineering practicality. By comprehensively judging the probability distribution of the posterior state vector and preset fault decision rules, the method effectively distinguishes between transient random interference and real, continuously evolving fault arcs in the circuit under test, reducing the false alarm rate of fault arc detection and improving detection accuracy.

[0018] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description, claims, and drawings. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 The diagram shown is a flowchart of the arc fault detection method described in an embodiment of this application.

[0020] Figure 2 This diagram illustrates the fault arc judgment process for a single observation cycle in the arc fault detection method described in this application embodiment.

[0021] Figure 3 The diagram shows the rate of change of current in a simulation experiment of the arc fault detection method described in this application embodiment.

[0022] Figure 4 The diagram shown illustrates the signal fluctuations in a simulation experiment of the arc fault detection method described in this application.

[0023] Figure 5 The diagram shown is a schematic diagram of envelope analysis in a simulation experiment of the arc fault detection method described in this application embodiment.

[0024] Figure 6 The diagram shown is a structural schematic of the arc fault detection device described in an embodiment of this application.

[0025] Figure 7 The diagram shown is a structural schematic of the arc fault detection system described in an embodiment of this application.

[0026] Specific element symbols: 10-Arc fault detection device, 20-Cloud service platform, 30-Protection actuator, 1-Arc fault detection system, 11-Model building module, 12-Feature extraction module, 13-Posterior state vector acquisition module, 14-Fault decision module. Detailed Implementation

[0027] To make the technical problems, technical solutions, and beneficial effects to be solved by this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of this application.

[0028] It should be noted that when a component is referred to as being "set on" another component, it can be directly on or indirectly on that other component. When a component is referred to as being "connected to" another component, it can be directly connected to or indirectly connected to that other component.

[0029] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0030] Currently, the mainstream methods for arc fault detection mainly include time-domain analysis, frequency-domain analysis, and time-frequency-domain analysis. Time-domain analysis offers advantages such as high detection speed, simple hardware implementation, and low computational resource consumption; however, it suffers from limited fault information, poor robustness to load changes, and a high rate of missed or false positives. Frequency-domain analysis algorithms are relatively mature and have some ability to distinguish loads with stable spectral characteristics, but its detection efficiency for non-sinusoidal and non-stationary arc fault signals is low, it struggles to capture transient signal characteristics, and it is susceptible to interference from devices that generate high-frequency noise, such as switching power supplies. Time-frequency-domain analysis can effectively handle non-stationary signals, has high detection efficiency, and provides rich feature information, making it the current mainstream research direction; however, it places high demands on processor performance and algorithm real-time performance, and its hardware implementation is difficult and costly.

[0031] High-precision detection (such as time-frequency domain methods) relies on complex signal processing algorithms and requires high-performance digital processing units, which dramatically increases system power consumption, size, and cost, making large-scale deployment difficult in space- and cost-constrained applications such as miniature circuit breakers (MCBs) and sockets. Meanwhile, simpler solutions (such as time-domain / frequency domain methods) are insufficient to meet the detection accuracy requirements of modern power systems under complex operating conditions.

[0032] Therefore, there is an urgent need for a fault arc detection solution that has both high-precision fault current identification capabilities, meets the detection accuracy requirements of complex operating conditions in new power systems, and can reduce system power consumption and cost.

[0033] Based on this, this application improves upon the fault arc detection methods, devices, computer-readable storage media, and systems in related technologies. The fault arc detection method of this application provides a novel implementation path for high-precision, high-reliability, and low-cost AC series, parallel, and grounding arc fault detection. Arc fault detection and identification technology plays a crucial role in AC power distribution protection of the power grid, and will also provide effective protection for DC arc fault detection in the DC field, smart cities, aerospace electrical systems, and hybrid electric new energy vehicles and electrical systems, fundamentally improving the ability to prevent electrical fires.

[0034] refer to Figure 1 As shown, the arc fault detection method of this application includes the following steps.

[0035] Step S101: Construct a stochastic dynamic system model of the arc fault in the circuit to be detected. The stochastic dynamic system model includes the state equation and observation equation of the arc fault.

[0036] The arc fault state parameters of the circuit under test (such as arc intensity, stability and type) and the load dynamic parameters of the line are set as state variables to construct a stochastic dynamic system model of the arc fault; the stochastic dynamic system model includes the state equation and observation equation of the arc fault.

[0037] Step S102: Obtain the sampling data of the current observation period of the circuit under test, and perform feature extraction on the sampling data to obtain the multi-dimensional observation vector of the current observation period that matches the observation equation.

[0038] For a single observation period, a microcontroller (MCU) with multiple analog-to-digital converters (ADCs) is needed to sample multiple signals from the circuit to be detected, obtaining sampled data containing these multiple signals. Ensuring strict timing alignment between different signals provides an accurate data foundation for subsequent time-frequency domain joint analysis and is a key factor in distinguishing real electric arcs from random noise. Subsequently, feature extraction is performed on the multiple signals in the sampled data using at least one of the following methods: time-domain feature extraction, frequency-domain feature extraction, and time-frequency domain feature extraction. Based on the extracted features, a multi-dimensional observation vector matching the observation equations in the stochastic dynamic system model is constructed.

[0039] Step S103: Based on the stochastic dynamic system model and the multidimensional observation vector of the current observation period, state estimation is performed using a continuous-discrete extended Kalman filter with Bayesian filter as the framework to obtain the posterior state vector of the current observation period.

[0040] Using a nonlinear Bayesian filter as a framework, the prior state vector and prior probability density of the current observation period are obtained based on the posterior state vector and the posterior probability density of the previous observation period. Then, a stochastic dynamic system model is used to map the prior state vector of the current observation period to obtain the multidimensional observation prediction vector of the current observation period, and the posterior state vector of the current observation period is obtained based on the multidimensional observation prediction vector of the current observation period.

[0041] Step S104: Based on the posterior state vector of the current observation period, determine whether there is a fault arc in the current observation period through a preset fault decision scheme. If there is, send a trip command to the protection actuator so that the protection actuator executes the trip command to control the circuit under test to disconnect the power. Otherwise, reacquire the sampling data of the circuit under test in the next observation period and make a corresponding judgment on whether there is a fault arc.

[0042] Based on the state characteristics corresponding to the high-frequency current data in the posterior state vector of the current observation period, a preset fault decision scheme is used to determine whether a fault arc exists in the current observation period. If it exists, a trip command is sent to the protection actuator of the circuit under test, so that the protection actuator executes the trip command and controls the circuit under test to disconnect the power, preventing fire caused by the fault arc. If it is determined to be a transient arc caused by a normal device state switch such as motor starting or switch switching, the process returns to step S102 to sample the data for the next observation period. The above steps are repeated until the circuit under test is de-energized.

[0043] The arc fault detection method in this embodiment acquires and fuses multi-dimensional features from multiple synchronous sampling data sources, improving accuracy while ensuring real-time detection. This provides multi-dimensional analysis data for subsequent algorithms, strengthening the basis for fault identification. A stochastic dynamic system model is introduced, combining a Bayesian filter and a continuous-discrete extended Kalman filter to dynamically and recursively fuse historical information with current multi-dimensional observation data to obtain the optimal fault state estimate. This gives the diagnostic results memory and smoothness, enabling accurate identification of fault arcs and effective suppression of transient noise interference. Simultaneously, by combining the posterior state vector probability distribution with preset fault decision rules, the method effectively distinguishes between transient random interference and real, continuously evolving fault arcs in the circuit, significantly reducing the false alarm rate and improving detection accuracy.

[0044] The steps S101 to S104 of the motor zero-position angle calibration method in this embodiment will be described in detail below.

[0045] Step S101: Construct a stochastic dynamic system model of the arc fault in the circuit to be detected. The stochastic dynamic system model includes the state equation and observation equation of the arc fault.

[0046] First, the arc fault state parameters and the load dynamic parameters of the circuit under test are taken as state variables, such as multiple frequency band components of the arc's high-frequency energy, the fundamental amplitude of the current, the statistical characteristics of the arc pulse (such as occurrence rate and width), and the load dynamic parameters. Variables that can be directly measured are taken as observation variables, such as the digital current and voltage signals obtained by synchronous sampling through an AFE and multiple ADCs. The observation variables are affected by both the system state and measurement noise. Then, combining the nonlinear volt-ampere characteristics of the fault arc and the dynamic equations of the circuit topology, a deterministic differential equation for the evolution of the state variables over time is derived. Since the generation and extinction of the fault arc, load changes, and high-frequency noise are not simply linearly related, the deterministic differential equation is a nonlinear model. For example, the linear approximation of the above nonlinear model under certain simplifying assumptions can be expressed as: (1) (2) in, Represents the observed variable. The simplified nonlinear system model (i.e., the state equation) is represented by equation (1), which represents the observation model (h) without noise terms. The nonlinear system model describes how the state evolves from the previous moment to the current moment (which may include the physical laws of load changes); the observation model (h) describes how to obtain the expected observation value from the current state, and is also a nonlinear model.

[0047] Simultaneously, process noise can be introduced into the above equations, such as to describe random uncertainties like arc fluctuations, load disturbances, and circuit parameter drift. Observation noise can also be introduced into the observation model to describe random errors such as sensor measurement errors and electromagnetic interference. If the model is in continuous-time form, it needs to be discretized to obtain discrete-time state and observation equations. The discretized state and observation equations constitute the stochastic dynamic system model. The stochastic dynamic system model obtained in this embodiment can be directly applied to subsequent Bayesian filters and continuous-discrete extended Kalman filters to achieve real-time estimation of the fault arc state and fault warning for the circuit under test. Furthermore, in the following description, x... k This represents the state variable (not shown in the formula).

[0048] refer to Figure 2 As shown, since arc fault detection is performed for each observation cycle of the circuit under test, the following steps are performed for each observation cycle monitored by the circuit under test.

[0049] Step S102: Obtain the sampling data of the current observation period of the circuit under test, and perform feature extraction on the sampling data to obtain the multi-dimensional observation vector of the current observation period that matches the observation equation.

[0050] Since the arc detection of the circuit under test is a real-time detection process, a data acquisition process exists for the current observation period of the circuit under test. This involves synchronously sampling multiple signals from the circuit under test using a microcontroller (MCU) with multiple analog-to-digital converters (ADCs). To facilitate multi-dimensional analysis data for subsequent data analysis and enhance the basis for judgment, this embodiment not only acquires the high-frequency component of the current (the main carrier of arc characteristics) but also simultaneously acquires power frequency current and voltage signals. Therefore, the acquired data includes high-frequency current data, power frequency current data, and power frequency voltage data. The synchronous acquisition of signals ensures strict timing alignment between different signals, providing an accurate data foundation for subsequent time-frequency domain joint analysis. Furthermore, the microcontroller can be an Arm® Cortex®-M based microcontroller. The Cortex®-M core provides sufficient computing power to run the arc detection algorithm in real time while maintaining low power consumption. Other microcontroller models are also possible; this embodiment does not impose any fixed limitations on them.

[0051] Then, feature extraction is performed on the multiple signals in the sampled data using at least one of the following methods: time-domain feature extraction, frequency-domain feature extraction, and time-frequency-domain feature extraction. In one embodiment of the present invention, the time-domain features may include at least one of rising edge steepness and falling edge steepness; the frequency-domain features may include at least one of preset frequency band energy distribution, high-frequency noise spectrum features, current zero-crossing distortion rate, and high-frequency pulse count; and the time-frequency-domain features may include at least one of short-time Fourier transform features (frequency energy distribution that varies with time), wavelet transform features (multi-scale analysis, particularly suitable for transient information), and time-frequency distribution entropy (complexity of the spectrum during quantization).

[0052] It should be noted that the time-domain feature extraction, frequency-domain feature extraction, and time-frequency-domain features mentioned above can also be other suitable features, and no fixed restrictions are imposed on them here.

[0053] After extracting the feature data from each signal source, a multi-dimensional observation vector Z for the current observation period can be constructed based on the extracted feature data, matching the observation equation in the stochastic dynamic system model. k .

[0054] In one embodiment of the present invention, after acquiring the sampling data, in order to reduce the noise interference of fixed frequency (e.g., switching power supply noise) in the sampling data, known fixed frequency interference can be filtered out by digital filtering and other processing to improve the accuracy of fault arc detection.

[0055] Step S103: Based on the stochastic dynamic system model and the multidimensional observation vector of the current observation period, state estimation is performed using a continuous-discrete extended Kalman filter with Bayesian filter as the framework to obtain the posterior state vector of the current observation period.

[0056] During the acquisition of the posterior state vector in the previous observation period, the posterior probability density of the previous observation period was also acquired simultaneously. Therefore, in the current observation period, we can first use the posterior state vector X from the previous observation period. k-1 The posterior probability density P of the previous observation period k-1 The prior state vector X for the current observation period is obtained by predicting the current observation period using a nonlinear Bayesian filter. k | k-1 And the prior probability density P of the current observation period k | k-1 This process is equivalent to "predicting" the current state based on physical laws during the intervals while waiting for new sampled data. During the acquisition of the initial state vector, the posterior state vector and posterior probability density of the previous observation period can be initialized based on the actual situation.

[0057] Then, the prior state vector X for the current observation period is obtained through the observation equation in the stochastic dynamic system model. k | k-1 A mapping is performed to obtain the multidimensional observation prediction vector Z for the current observation period, which is expected to be observed. k | k-1 Simultaneously, the Kalman filter gain K for the current observation period is obtained through a continuous-discrete extended Kalman filter. k The Kalman filter gain is a matrix that dynamically balances the confidence level of the model prediction with the confidence level of the current actual measurement. If the measurement noise is low, the gain will increase to give more confidence to the new measurement data.

[0058] After obtaining the multidimensional observation vector for the current observation period, the continuous-discrete extended Kalman filter can calculate the difference (i.e., the innovation) between the multidimensional observation vector and the multidimensional observation prediction vector for the current observation period. Then, the Kalman filter gain K for the current observation period is used. k The difference (i.e., the innovation) is fed back in the optimal way to realize the prior state vector X of the current observation period. k | k-1 The correction is then used to obtain the posterior state vector X for the current observation period. k | k At the same time, the posterior probability density P of the current observation period is obtained. k | k Then, based on the posterior state vector X of the current observation period, k | k To make the best inference of fault arc information in the circuit to be detected.

[0059] In the aforementioned state estimation process, time-domain, frequency-domain, and time-frequency-domain information acquired simultaneously from multiple sources, along with historical information from the time series, are naturally and optimally integrated, resulting in strong anti-interference capabilities and high recognition accuracy. Furthermore, the continuous-discrete extended Kalman filter is continuously optimized and operates efficiently throughout the process, meeting the real-time data requirements and achieving a good fusion of theoretical advancement and engineering practicality.

[0060] Step S104: Based on the posterior state vector of the current observation period, determine whether there is a fault arc in the current observation period through a preset fault decision scheme. If there is, send a trip command to the protection actuator so that the protection actuator executes the trip command to control the circuit under test to disconnect the power. Otherwise, reacquire the sampling data of the circuit under test in the next observation period and make a corresponding judgment on whether there is a fault arc.

[0061] Since the acquired posterior state vector of the current observation period contains state vector features corresponding to high-frequency current signals, power frequency current signals, and power frequency voltage signals, and high-frequency current is the main carrier of electric arc, this embodiment preferably uses the state features corresponding to high-frequency current data in the posterior state vector of the current observation period to determine the fault arc.

[0062] In one embodiment of the present invention, the arc energy state estimate and the current distortion state estimate corresponding to the high-frequency current data are obtained from the posterior state vector of the current observation period. Then, the corresponding standard state estimate is obtained based on the obtained arc energy state estimate, and it is determined whether the standard state estimate is greater than a first preset threshold. At the same time, it is determined whether the obtained current distortion state estimate is greater than a second preset threshold. If both are greater than the corresponding threshold, it is determined that there is a fault arc in the current circuit to be tested. If only one of them is greater than the corresponding threshold or neither is greater than the corresponding threshold, it is determined that the current circuit to be tested is normal and there is no fault arc.

[0063] The process of obtaining the corresponding standard state estimate based on the acquired arc energy state estimate includes: determining a preset time period backward from the current observation period, so that the preset time period includes multiple observation periods including the current observation period; then obtaining the arc energy state estimate corresponding to the high-frequency current data of all observation periods within the preset time period (generally, it has already been determined in the corresponding observation period and can be reused here), that is, obtaining the arc energy state estimate corresponding to N observation periods including the current observation period; calculating the standard deviation of the N arc energy state estimates; and taking 3 times the standard deviation as the standard state estimate obtained based on the arc energy state estimate obtained in the current observation period.

[0064] It should be noted that this embodiment can also determine the fault arc in the circuit under test based on other types of state estimates in the posterior state vector of the current observation period, and there are no fixed restrictions on it here.

[0065] In one embodiment of the present invention, a specific state estimate can be obtained based on the posterior state vector of the current observation period. Then, the feature state estimate is fed into a preset machine learning model to determine whether a fault arc exists in the circuit to be detected during the current observation period. It should be noted that the preset model and its learning model are models trained with massive amounts of "good arc" (normal operation of the electrical appliance) and "bad arc" (fault) data, possessing extremely strong arc classification capabilities.

[0066] It should be noted that the present invention can also set other reasonable preset fault decision schemes to determine whether there is a fault arc in the current observation period, such as a judgment scheme that combines the judgment rules with machine learning model terms, etc. This application does not impose any fixed limitations on it.

[0067] When a fault arc is detected in the circuit under test, a trip command is sent to the protection actuator to de-energize the circuit, achieving ultra-fast power-off protection and minimizing fire risk. If the circuit is found to be normal, steps S102-S104 are repeated to detect the presence of a fault arc in the next observation cycle.

[0068] To verify the effectiveness of the arc fault detection method of this application, the following simulation experiment of the arc fault detection method of this application is carried out using the waveform of a certain circuit to be verified during a time period with both fault-free and faulty arcs as an example (only the specific waveform of the signal obtained after processing by the continuous-discrete extended Kalman filter is simulated). Figure 3 This is a schematic diagram of the current change rate in a simulation experiment of the arc fault detection method described in this application embodiment; Figure 4 This is a schematic diagram of signal fluctuation in a simulation experiment of the arc fault detection method described in the embodiments of this application; Figure 5 This is a schematic diagram of envelope analysis in a simulation experiment of the arc fault detection method described in this application embodiment. (Reference) Figures 3-5 As shown, after analysis and processing by the arc fault detection method of this application, fault arcs can be identified more easily, and the accuracy rate is high.

[0069] The arc fault detection method provided in this invention can reduce electrical fires caused by arcs by more than 90%. At the same time, the data processing method in this application has strong adaptability and can reduce the amount of on-site debugging work by 80%. It is also highly compatible and has a wide range of applications, applicable to various power distribution systems such as residential, commercial, and industrial systems. It provides a high-reliability technical benchmark for arc fault detection and provides core safety detection capabilities for intelligent power distribution systems.

[0070] like Figure 6 As shown in the figure, this application embodiment also provides an arc fault detection device 10, including a model building module 11, a feature extraction module 12, a posterior state vector acquisition module 13, and a fault decision module 14.

[0071] The model building module is used to construct a stochastic dynamic system model of the arc fault in the circuit to be detected. The stochastic dynamic system model includes the state equation and observation equation of the arc fault.

[0072] The feature extraction module is used to obtain the sampling data of the circuit under test in the current observation period, and to extract features from the sampling data to obtain a multi-dimensional observation vector of the current observation period that matches the observation equation.

[0073] The posterior state vector acquisition module is used to obtain the posterior state vector of the current observation period based on the stochastic dynamic system model and the multidimensional observation vector of the current observation period. Using a Bayesian filter as a framework, it performs state estimation through a continuous-discrete extended Kalman filter.

[0074] The fault decision module is used to determine whether there is a fault arc in the current observation period based on the posterior state vector of the current observation period and through a preset fault decision scheme. If there is, a trip command is sent to the protection actuator so that the protection actuator executes the trip command to control the circuit under test to cut off the power. Otherwise, the sampling data of the circuit under test in the next observation period is re-acquired, and the corresponding judgment on whether there is a fault arc is made.

[0075] Furthermore, the posterior state vector acquisition module 13 is also specifically used for: Based on a stochastic dynamic system model and the multidimensional observation vector of the current observation period, using a Bayesian filter framework, state estimation is performed through a continuous-discrete extended Kalman filter to obtain the posterior state vector of the current observation period, including: Based on the posterior state vector and posterior probability density of the circuit under test in the previous observation period, the prior state vector and prior probability density of the current observation period are obtained through a nonlinear Bayesian filter. The prior state vector of the current observation period is mapped through the observation equation in the stochastic dynamic system model to obtain the multidimensional observation prediction vector of the current observation period. Based on the multidimensional observation vector and the multidimensional observation prediction vector of the current observation period, the posterior state vector and the posterior probability density of the current observation period are obtained through a continuous-discrete extended Kalman filter.

[0076] Furthermore, the feature extraction module 12 is specifically used for: The sampling data includes high-frequency current data, power frequency current data, and power frequency voltage data synchronously acquired by a multi-channel analog converter.

[0077] Furthermore, the feature extraction module 12 is specifically used for: Feature extraction from the sampled data includes: Feature extraction is performed on high-frequency current data, power frequency current data, and power frequency voltage data using at least one of the following methods: frequency domain feature extraction, time domain feature extraction, and time-frequency domain feature extraction.

[0078] Furthermore, the feature extraction module 12 is specifically used for: The time-domain features include at least one of rising edge steepness and falling edge steepness; the frequency-domain features include at least one of preset frequency band energy distribution, high-frequency noise spectrum features, current zero-crossing distortion rate, and high-frequency pulse count; the time-frequency domain features include at least one of short-time Fourier transform features, wavelet transform features, and time-frequency distribution entropy.

[0079] Furthermore, the fault decision module 14 is specifically used for: Based on the posterior state vector of the current observation period, the determination of whether a fault arc exists in the current observation period through a preset fault decision scheme includes: Based on the state characteristics corresponding to the high-frequency current data in the posterior state vector of the current observation period, a pre-set fault decision scheme is used to determine whether a fault arc exists in the current observation period.

[0080] Furthermore, the fault decision module 14 is specifically used for: Based on the state characteristics corresponding to the high-frequency current data in the posterior state vector of the current observation period, the determination of whether a fault arc exists in the current observation period through a preset fault decision scheme includes: The arc energy state estimate and current distortion state estimate corresponding to the high-frequency current data are obtained from the posterior state vector of the current observation period. It is determined whether the standard state estimate obtained based on the arc energy state estimate of the current observation period is greater than the first preset threshold and whether the current distortion state estimate of the current observation period is greater than the second preset threshold. If so, it is determined that there is a fault arc in the circuit to be tested; otherwise, it is determined that the circuit to be tested is normal. The standard state estimate obtained based on the arc energy state estimate of the current observation period is three times the standard deviation of the arc energy state estimate of all observation periods within the preset time period, which includes the current observation period.

[0081] This application's arc fault detection device provides a novel approach for high-precision, high-reliability, and low-cost AC arc fault detection. It can be widely used in the safety protection of intelligent low-voltage electrical appliances, new energy systems, special vehicles, and aerospace electrical systems, fundamentally improving the ability to prevent electrical fires.

[0082] like Figure 7 As shown in the figure, this application embodiment also provides an arc fault detection system 1, which includes a cloud service platform 20, an arc fault detection device 10 and a protection actuator 30 that are interconnected, wherein the arc fault detection device is the arc fault detection device 10 in the above embodiment.

[0083] The arc fault detection device 10 is mainly used to execute the steps in the above-described arc fault detection method to detect fault arcs in the circuit under test. Simultaneously, it packages multiple types of data, including at least the original sampling data, into a defined data packet. The data packet may also include diagnostic results, key state estimates, a summary of the original sampling data (such as preset frequency band energy values), timestamps, and its own health status (such as chip temperature and supply voltage). The arc fault detection device 10 can periodically or event-triggeredly send the aforementioned data packet to the cloud service platform 20 via a communication interface.

[0084] The protective actuator 30 can be a circuit breaker or relay controller containing a tripping mechanism, installed in the circuit under test. The protective actuator 30 is mainly used to receive tripping commands sent by the arc fault detection device 10 when a fault arc is detected. Upon receiving the tripping command, its tripping steps can be: immediately or after a very short configurable delay to initiate the tripping process; or illuminating a warning indicator or recording the event, but without directly tripping, continuously monitoring subsequent commands. The protective actuator 30 can also receive parameter configuration information from the cloud service platform 20 and update parameters based on the parameter configuration information; or send its own important event information to the cloud service platform 20 for information reporting. Once the arbitration logic of the protective actuator 30 decides to initiate the tripping process, it needs to drive a high-current switch (such as a MOSFET, relay, or mechanical tripping mechanism) to forcibly disconnect the circuit under test.

[0085] In one embodiment of this application, the protection actuator 30 may also feed back its own tripping status feedback data (such as "tripped", "not tripped", "coil fault" etc.) to the cloud service platform 20 after the tripping action, so that the cloud service platform 20 can make subsequent reminders or operations.

[0086] The cloud service platform 20 is mainly used to receive sampling data sent by the arc fault detection device and tripping status feedback data sent by the protection actuator. After acquiring the data, it parses all incoming data, structures it, and stores it in a database or time-series database to form a long-term "electrical fingerprint" archive.

[0087] Meanwhile, the cloud service platform 20 can also perform the following operations based on the input data. For example, it can perform deep pattern mining, that is, analyze the changes in electrical characteristics of the circuit under test before and after a fault on a longer time scale, and optimize or train more complex fault identification models. It can also perform predictive maintenance of the circuit under test: that is, analyze the long-term trends of arc characteristics and insulation condition estimates, predict the degree of line aging or the risk of loose connections, and issue maintenance warnings in advance. It can also perform multi-node correlation analysis: that is, in a network system, compare the data of different lines or nodes to locate interference sources or systemic risks.

[0088] The cloud service platform 20 can also generate web interfaces, mobile terminal or PC client views based on the input data, displaying real-time waveforms, state estimation curves, historical events, alarm lists, system topology diagrams, etc. The cloud service platform 20 can also respond to user commands and send parameter update, remote reset, firmware upgrade, and other commands to designated lower-level machines (arc fault detection device 10 and protection actuator 30).

[0089] In this embodiment, the arc fault detection system constructs an intelligent arc detection network based on the arc fault detection device 10 through multi-terminal interaction. This allows for the provision of abundant data for intelligent analysis and display, and direct drive of protection actuators to ensure real-time power-off of the circuit under test. Thus, a layered, reliable, and scalable complete arc fault protection solution is constructed.

[0090] To illustrate how the arc fault detection system in this embodiment works, a real-world scenario will be used to explain it.

[0091] Scenario: The circuit under test is an old electric kettle power cord. The insulation of the electric kettle power cord is damaged, and intermittent series arcs are generated in the circuit.

[0092] The arc fault detection device 10 has the following functions: when an abnormality is detected (current 15.3A (normal), high frequency energy +28dB, zero crossing rate +40%), local processing is performed: if an arc condition is suspected, a 500ms delay trip command is initiated; action: the yellow ARC light starts flashing rapidly and is reported to the cloud service platform 20.

[0093] The protection actuator 30 has the following functions: upon receiving a remote trip command (which can be a delayed trip command or an immediate power-off command sent by the user's App), it immediately cuts off the power supply; and the red ARC light remains constantly on, and the event is recorded.

[0094] The cloud service platform 20 receives an alarm and checks related devices: In the same circuit: refrigerator (normal), microwave oven (off); In adjacent circuits: lighting (normal), air conditioner (normal); Analysis conclusion: single point of failure, high risk; Push alarm to user's mobile app.

[0095] The user's mobile app may experience the following: receiving a push notification: "A suspected electric arc has been detected in the kitchen socket. The power will be automatically cut off in 0.3 seconds"; the user runs to the kitchen and smells a burning odor; or the user clicks "Cut off power immediately" on the app.

[0096] This embodiment of the system not only achieves ultra-fast power outage protection in a physical sense, but also transforms post-event remediation into pre-event early warning and rapid in-event response through precise diagnosis and digital management, forming a complete electrical safety closed loop and providing core safety detection capabilities for intelligent power distribution systems.

[0097] This application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps described in the above-described method embodiments.

[0098] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. A computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0099] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0100] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0101] In the embodiments provided in this application, it should be understood that the disclosed apparatus / device and method can be implemented in other ways. For example, the apparatus / device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0102] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0103] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for detecting electric arc faults, characterized in that, include: A stochastic dynamic system model of an arc fault in a circuit to be detected is constructed, wherein the stochastic dynamic system model includes the state equation and observation equation of the arc fault; Acquire the sampling data of the circuit under test in the current observation period, and perform feature extraction on the sampling data to obtain a multi-dimensional observation vector of the current observation period that matches the observation equation; Based on the stochastic dynamic system model and the multidimensional observation vector of the current observation period, using a Bayesian filter as a framework, state estimation is performed through a continuous-discrete extended Kalman filter to obtain the posterior state vector of the current observation period. Based on the posterior state vector of the current observation period, a preset fault decision scheme is used to determine whether a fault arc exists in the current observation period. If it exists, a trip command is sent to the protection actuator so that the protection actuator executes the trip command to control the circuit under test to disconnect the power. Otherwise, the sampling data of the circuit under test in the next observation period is reacquired, and a corresponding judgment is made on whether a fault arc exists.

2. The arc fault detection method according to claim 1, characterized in that, Based on the aforementioned stochastic dynamic system model and the multidimensional observation vector of the current observation period, using a Bayesian filter as a framework, state estimation is performed through a continuous-discrete extended Kalman filter to obtain the posterior state vector of the current observation period, including: Based on the posterior state vector and posterior probability density of the circuit under test in the previous observation period, the prior state vector and prior probability density of the current observation period are obtained through a nonlinear Bayesian filter. The prior state vector of the current observation period is mapped through the observation equation in the stochastic dynamic system model to obtain the multidimensional observation prediction vector of the current observation period. Based on the multidimensional observation vector and the multidimensional observation prediction vector of the current observation period, the posterior state vector and the posterior probability density of the current observation period are obtained through a continuous-discrete extended Kalman filter.

3. The arc fault detection method according to claim 1, characterized in that, The sampled data includes high-frequency current data, power frequency current data, and power frequency voltage data synchronously acquired by a multi-channel analog converter.

4. The arc fault detection method according to claim 3, characterized in that, Feature extraction of the sampled data includes: Feature extraction is performed on the high-frequency current data, the power frequency current data, and the power frequency voltage data using at least one of the following methods: frequency domain feature extraction, time domain feature extraction, and time-frequency domain feature extraction.

5. The arc fault detection method according to claim 4, characterized in that, The time-domain features include at least one of rising edge steepness and falling edge steepness; the frequency-domain features include at least one of preset frequency band energy distribution, high-frequency noise spectrum features, current zero-crossing distortion rate, and high-frequency pulse count; the time-frequency domain features include at least one of short-time Fourier transform features, wavelet transform features, and time-frequency distribution entropy.

6. The arc fault detection method according to claim 3, characterized in that, Based on the posterior state vector of the current observation period, the determination of whether a fault arc exists in the current observation period through a preset fault decision scheme includes: Based on the state characteristics corresponding to the high-frequency current data in the posterior state vector of the current observation period, a pre-set fault decision scheme is used to determine whether a fault arc exists in the current observation period.

7. The arc fault detection method according to claim 6, characterized in that, Based on the state characteristics corresponding to the high-frequency current data in the posterior state vector of the current observation period, the determination of whether a fault arc exists in the current observation period through a preset fault decision scheme includes: The arc energy state estimate and current distortion state estimate corresponding to the high-frequency current data are obtained from the posterior state vector of the current observation period. It is determined whether the standard state estimate obtained based on the arc energy state estimate of the current observation period is greater than the first preset threshold and whether the current distortion state estimate of the current observation period is greater than the second preset threshold. If so, it is determined that there is a fault arc in the circuit to be detected; otherwise, it is determined that the circuit to be detected is normal. The standard state estimate obtained based on the arc energy state estimate of the current observation period is three times the standard deviation of the arc energy state estimates of all observation periods within a preset time period, where the preset time period includes the current observation period.

8. An arc fault detection device, characterized in that, It includes a model building module, a feature extraction module, a posterior state vector acquisition module, and a fault decision module; The model building module is used to build a stochastic dynamic system model of the arc fault in the circuit to be detected. The stochastic dynamic system model includes the state equation and observation equation of the arc fault. The feature extraction module is used to acquire the sampling data of the current observation period of the circuit to be detected, and to perform feature extraction on the sampling data to obtain a multi-dimensional observation vector of the current observation period that matches the observation equation. The posterior state vector acquisition module is used to obtain the posterior state vector of the current observation period based on the stochastic dynamic system model and the multidimensional observation vector of the current observation period, using a Bayesian filter as a framework and a continuous-discrete extended Kalman filter for state estimation. The fault decision module is used to determine whether there is a fault arc in the current observation period based on the posterior state vector of the current observation period and through a preset fault decision scheme. If there is, a trip command is sent to the protection actuator so that the protection actuator executes the trip command to control the circuit under test to disconnect the power. Otherwise, the sampling data of the circuit under test in the next observation period is re-acquired, and the corresponding judgment on whether there is a fault arc is made.

9. An arc fault detection system, characterized in that, It includes a cloud service platform, an arc fault detection device, and a protection actuator that are interconnected, wherein the arc fault detection device is the arc fault detection device as described in claim 8; The cloud service platform is used to receive sampling data sent by the arc fault detection device and tripping status feedback data sent by the protection actuator.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images