Arc fault detection by mixed signal machine learning and neural network

Through the combination of broadband current sensor and microcontroller, combined with digital filter and spectral density analysis, the arc fault prediction is used to use machine learning models to solve the problem of misjudgment and misjudgment of arc fault detection in the existing technology, and the complexity is achieved Efficient identification and accurate detection in circuit environments.

CN118974866APending Publication Date: 2024-11-15HUBBELL INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202380032404.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2022-02-07
Filing Date
2023-02-07
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing arc fault detection technology has problems of misjudgment and misjudgment when detecting arc faults, and it is difficult to efficiently identify arc faults in complex circuit environments.

Method used

A broadband current sensor and microcontroller are used to calculate arc characteristics through digital filters and spectral density analysis, and predict and identify arc faults based on machine learning models. The probability of arc faults is determined by using inference algorithms and machine learning models to activate interrupts. Device.

Benefits of technology

It improves the accuracy and efficiency of arc fault detection, reduces misjudgment and misjudgment, and can effectively identify arc faults in complex circuit environments and activate the interrupt device in a timely manner.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118974866A_ABST
    Figure CN118974866A_ABST
Patent Text Reader

Abstract

A circuit interrupting device, the circuit interrupting device comprising: a line terminal; a broadband current sensor configured to measure a current flowing through the line terminal; a zero-cross detection circuit configured to measure a voltage and a frequency of the line terminal; and a microcontroller including an electronic processor. The microcontroller is configured to apply a digital filter to a line current measurement signal received from the broadband current sensor and to estimate a spectral density of the filtered line current measurement signal. The microcontroller is further configured to calculate an arc characteristic based on the spectral density; determining whether an arc fault is present within the circuit interruption device based on a comparison of the arc characteristic to one or more thresholds; and activating the interruption device when there is an arc fault.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 307,392, filed on February 7, 2022, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure relates generally to switching electrical devices. More particularly, the present disclosure relates to circuit interrupting devices, such as arc fault circuit interrupter (AFCI) devices, that change from a "reset" or latched state to a "tripped" or unlatched state when one or more conditions are detected. Summary of the invention

[0004] A first aspect of the present disclosure provides a circuit interruption device, the circuit interruption device comprising: a line terminal; a broadband current sensor configured to measure a current flowing through the line terminal; a zero-crossing detection circuit configured to measure a voltage and a frequency of the line terminal; and a microcontroller including an electronic processor. The microcontroller is configured to apply a digital filter to a line current measurement signal received from the broadband current sensor, and to estimate a spectral density of the filtered line current measurement signal. The microcontroller is also configured to calculate an arc characteristic based on the spectral density; determine whether an arc fault exists within the circuit interruption device based on a comparison of the arc characteristic with one or more thresholds; and activate the interruption device when an arc fault exists.

[0005] Another aspect of the present disclosure provides a method for detecting the presence of an arc fault occurring within a circuit including a line terminal. The method includes: measuring a current flowing through the line terminal with a broadband current sensor; applying a digital filter to a line current measurement signal received from the broadband current sensor by a microcontroller including an electronic processor; and estimating, by the microcontroller, a spectral density of the filtered line current measurement signal. The method also includes: calculating, by the microcontroller, an arc characteristic based on the spectral density; determining, by the microcontroller, whether an arc fault exists within the circuit based on a comparison of the arc characteristic with one or more threshold values; and activating, by the microcontroller, an interrupt device when an arc fault exists.

[0006] Another aspect of the present disclosure provides a system comprising a circuit interrupting device and an external device. The circuit interrupting device comprises: a line terminal; a broadband current sensor configured to measure a current flowing through the line terminal; and a communication circuit configured to wirelessly transmit a signal comprising a current measurement taken by the broadband current sensor. The external device comprises a second communication circuit configured to wirelessly receive the signal; and a controller comprising an electronic processor. The controller is configured to: estimate a spectral density of the current measurement; calculate at least one arc characteristic associated with the spectral density; develop a machine learning model based on the spectral density and the at least one arc characteristic; and deploy the machine learning module to the circuit interrupting device.

[0007] Another aspect of the present disclosure provides a method for deploying a machine learning model to a circuit interrupting device. The method includes: receiving, by a communication circuit, a first set of line current measurements taken when an arc fault is present; receiving, by the communication circuit, a second set of line current measurements taken when an arc fault is not present; estimating, by a controller including an electronic processor, a first spectral density of the first set of line current measurements; and estimating, by the controller, a second spectral density of the second set of line current measurements. The method also includes: calculating, by the controller, a first set of arc features associated with the first spectral density; calculating, by the controller, a second set of non-arc features associated with the second spectral density; executing, by the controller, a training algorithm to create a machine learning model based on the first set of arc features and the second set of non-arc features; and deploying the machine learning model to the circuit interrupting device.

[0008] Another aspect of the present disclosure provides a circuit interruption device, the circuit interruption device comprising: a line terminal; a broadband current sensor configured to measure a current flowing through the line terminal; and a microcontroller including an electronic processor. The microcontroller is configured to estimate a spectral density of line current measurements taken by the broadband current sensor; calculate an arc signature based on the spectral density; and calculate a probability of an arc fault occurring based on the arc signature and the spectral density by executing an inference algorithm. The microcontroller is also configured to set a first flag when the probability exceeds a threshold within a first amount of time; and activate the interruption device when the first flag is set.

[0009] Another aspect of the present disclosure provides a method for detecting the presence of an arc fault occurring within a circuit including a line terminal. The method includes: measuring a current flowing through the line terminal with a broadband current sensor; estimating, by a microcontroller, a spectral density of a line current measurement signal generated by the broadband current sensor; and calculating, by the microcontroller, an arc signature based on the spectral density. The method also includes: calculating a probability of an arc fault occurring based on the arc signature and the spectral density by executing an inference algorithm; setting a first flag when the probability exceeds a threshold within a first amount of time; and activating an interrupt device when the first flag is set.

[0010] Other aspects of the application will become apparent by consideration of the detailed description and accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1A A perspective view of an AFCI device or AFCI receptacle is shown according to some embodiments.

[0012] Figure 1B According to some embodiments Figure 1A A perspective view of an AFCI outlet with the front cover removed to expose the main printed circuit board (PCB).

[0013] Figure 2 According to some embodiments Figure 1B A three-dimensional diagram of the main PCB.

[0014] Figure 3 According to some embodiments Figure 1A A perspective view of an AFCI receptacle with the front cover removed to reveal the secondary PCB.

[0015] Figure 4 Some embodiments of the present application are shown Figure 3 A three-dimensional view of the secondary PCB.

[0016] Figure 5A and Figure 5B A perspective view of an AFCI receptacle is shown in accordance with some embodiments.

[0017] Figure 6 Some embodiments of the present application are shown Figure 1A Block diagram of the AFCI socket control system.

[0018] Figure 7 is a flow chart illustrating a process or operation for detecting an arc fault in accordance with some embodiments.

[0019] Figure 8 Synchronized Short Time Fourier Transform windows are shown in accordance with some embodiments.

[0020] Fig. 9 An exemplary spectral density of line current measurements according to some embodiments is shown.

[0021] Fig.10 A schematic diagram of a machine learning model according to some embodiments of the present application is shown.

[0022] Fig.11 is a flow chart illustrating a process or operation for detecting an arc fault in accordance with some embodiments.

[0023] Fig.12 is a flow chart illustrating a process or operation for aggregating outputs of an arc fault detection method according to some embodiments.

[0024] Fig.13 is a flow chart illustrating a process or operation for combining outputs from more than one arc fault detection method in accordance with some embodiments.

[0025] Fig.14 According to some embodiments Figure 6 Block diagram of the control system of the external devices shown.

[0026] Fig.15 is a flow chart illustrating a process or operation for creating a machine learning model according to some embodiments.

[0027] Fig.16 is a flow chart illustrating a process or operation for detecting an arc fault in accordance with some embodiments. DETAILED DESCRIPTION

[0028] Before explaining any embodiments of the present application in detail, it should be understood that the present application is not limited in its application to the details of construction and arrangement of components set forth in the following description or shown in the drawings. The present application can be practiced or implemented in various ways.

[0029] Figure 1AA front view of an AFCI device or AFCI outlet 100 according to some embodiments of the present disclosure is shown. The AFCI outlet 100 includes a front cover 105 having a double-gang outlet panel 110 with a phase or live opening 115, a neutral opening 120, and a ground opening 125. The panel 110 also has an opening 130 (accommodating a RESET button 135), an adjacent opening 140 (accommodating a TEST button 145), and one or more additional openings 150. In some embodiments, the one or more additional openings 150 accommodate indicators, such as, but not limited to, light emitting diodes (LEDs) of various colors. In some embodiments, the one or more additional openings 150 accommodate bright LEDs used, for example, as night lights. In some embodiments, the one or more additional openings 150 accommodate photoconductive photocells, for example, for controlling night light LEDs. In some embodiments, the one or more additional openings 150 provide access to fixing screws for adjusting photocell devices or buzzers according to this embodiment and other embodiments. AFCI outlet 100 also includes a rear cover (not shown or listed) secured to front cover 105 by a plurality of fasteners (not shown or listed) and ground yoke / bridge assembly 155. Ground yoke / bridge assembly 155 includes standard mounting ears 160 protruding from both ends of outlet 100.

[0030] Figures 1B to 3 A perspective view of an AFCI outlet 100 is shown in which the front cover 105, the back cover, and other components have been removed to reveal a main printed circuit board (PCB) or main board 200, according to some embodiments. In some embodiments, the main board 200 provides controls and physical support for most of the working components included in the AFCI outlet 100. For example, Figure 1B As shown, the top surface 205 of the main plate 200 provides support for the solenoid or interrupter device 210. The top surface 205 also supports the cantilevered phase and neutral line contact arms 220, 225 and the phase and neutral load contact arms 230, 235 ( Figure 3 ). The respective distal ends of the line contact arms 220, 225 support the phase line and neutral line contacts 240, 245. Similarly, the respective distal ends of the load contact arms 230, 235 support the phase and neutral load contacts 250, 255 ( Figure 3 ). The resiliency of the cantilevered line contact arms 220, 225 biases or separates the line contacts 240, 245 away from or from the load contacts 250, 255. The load contact arms 230, 235 extend from a movable contact carriage 260 constructed of an insulating material.

[0031] like Figure 2As shown, the main board 200 also includes a bottom surface 305 that provides physical and operational support for many of the control electronics included in the AFCI outlet 100. For example, the bottom surface 305 supports a controller 310 and one or more slots or interfaces 315 for accommodating connections. The controller 310 is an integrated circuit device, such as a Microchip microcontroller. However, in other embodiments, the controller 310 is implemented as another type of processor-based control device. The controller 310 (including a memory and an electronic processor) can be configured to control various operations of the AFCI outlet 100. For example, in some embodiments, the controller 310 is configured to detect the occurrence of an arc fault. In some embodiments, the main board 200 includes additional communication interfaces CP1-CP5, which can be referred to as compliance pins.

[0032] In some embodiments, the AFCI outlet 100 includes one or more additional PCBs that provide physical and operational support for one or more additional control electronics included in the AFCI outlet 100. For example, Figure 3 and Figure 4 As shown, the AFCI outlet 100 may include a second PCB or secondary board 400 for detecting the occurrence of an arc fault. The secondary board 400 may be connected to the main board 200, for example, via one or more pins 505. In some embodiments, the pins 505 include one or more serial communication pins for transmitting data signals between the main board 200 and the secondary board 400. For example, the pins 505 may include one or more serial peripheral interface (SPI) pins configured to link the communication between the main board 200 and the secondary board 400. In some embodiments, the pins 505 are accommodated by the interface 315 of the main board 200. In some embodiments, the pins 505 are connected to the interfaces CP1-CP5 of the main board 200. In addition, a first hole 510 and a second hole 515 are formed in the secondary board 400 and are arranged to accommodate the line contact arms 220, 225, respectively. In some embodiments, the first hole 510 is configured to receive the phase line contact arm 220, and the second hole 515 is configured to accommodate the neutral line contact arm 225.

[0033] like Figure 4 As shown, the secondary board 400 may be implemented as an AFCI module that includes one or more circuit components for detecting the presence of an arc fault within the AFCI outlet 100 and / or the circuit to which the AFCI outlet 100 is connected. Figure 4As shown, the secondary board 400 includes one or more broadband current sensors, such as a first coil 520 and a second coil 525. In some embodiments, the first coil 520 and the second coil 525 are embedded in the secondary board 400 and include a first coil hole and a second coil hole, respectively. In some embodiments, the first coil hole is configured to accommodate the phase line contact arm 220, and the second coil hole is configured to accommodate the neutral line contact arm 225. In some embodiments, the first coil 520 and the second coil 525 are implemented as Rogowski coils. In some embodiments, the first coil 520 and the second coil 525 are implemented as other types of coils.

[0034] In some embodiments, the secondary board 400 further includes an arc fault detection circuit 530 that includes one or more electrical components for detecting the occurrence of an arc fault. For example, the arc fault detection circuit 530 may include a bandpass filter, an analog-to-digital converter (ADC), an integrator, a gain stage or scaling module, a resonator, and / or a time domain correlator. In some embodiments, the secondary board 400 does not include its own controller. In such embodiments, the controller 310 is configured to determine whether an arc fault exists based on data received from the arc fault detection circuit 530. In some embodiments, the arc fault detection circuit 530 includes its own controller that is configured to control the operation of the components included in the arc fault detection circuit 530. In such embodiments, the arc fault detection circuit 530 is configured to perform a variety of functions related to detecting the occurrence of an arc fault. The controller included in the arc fault detection circuit 530 may be implemented as the same type of controller as the controller 310. However, the controller included in the arc fault detection circuit 530 may be implemented as some other type of microcontroller.

[0035] In some embodiments, AFCI outlet 100 does not include secondary board 400. In such embodiments, first coil 520 and second coil 525 and arc fault detection circuit 530 are supported by primary board 200. Thus, in such embodiments, controller 310 is configured to determine whether an arc fault exists.

[0036] although Figure 3 1 is shown as including terminal screws configured to receive power from an external power source and / or provide power to additional downstream devices, but in some embodiments, the AFCI outlet 100 includes different terminal connection components configured to receive power from an external power source and / or provide power to additional downstream devices. For example, FIG. 5A to FIG. 5BAn embodiment of an AFCI outlet 100 is shown that includes a line terminal connector and a load terminal connector coupled to the AFCI outlet 100 by a snap-fit ​​connection, such as the outlet described in U.S. Patent Application Publication No. 2021 / 0226389, entitled "GROUND FAULT CIRCUIT INTERRUPTERS AND CONNECTORS FOR USE WITH THE SAME," published on July 22, 2021, the entire contents of which are incorporated herein by reference. It should be understood that in some embodiments, the AFCI outlet 100 is implemented as an outlet type not explicitly described herein.

[0037] Figure 6 is a block diagram illustrating a control system (or circuit) 600 of an AFCI outlet 100. In the illustrated embodiment, the control system 600 includes a controller 310 supported by a primary board. However, it should be understood that in some embodiments, the control system 600 is implemented with a controller supported by a secondary board 400. That is, in embodiments where the secondary board 400 includes its own controller, the controller supported by the secondary board 400 can be configured to perform the processes described herein with respect to the controller 310. Similarly, the control system 600 of the illustrated embodiment includes a first coil 520, a second coil 525, and an arc fault monitoring circuit 530 supported by the secondary board 400. However, it should be understood that in embodiments where the AFCI outlet 100 does not include a secondary board 400, the first coil 520, the second coil 525, and the arc fault detection circuit 530 are supported by the primary board 200.

[0038] like Figure 6 As shown, the controller 310 is electrically and / or communicatively connected to various modules or components of the AFCI outlet 100. For example, the controller 310 is connected to the circuit interrupting device 210, the first coil 520, the second coil 525, the arc fault detection circuit 530, the zero crossing detection circuit 602, the power supply circuit 605, and the communication circuit 610.

[0039] In some embodiments, the controller 310 includes a plurality of electrical and electronic components that provide power, operational control, and protection to the components and modules within the controller 310 and / or the AFCI outlet 100. For example, the controller 310 includes, among other things, an electronic processor 615 (e.g., a microprocessor or another suitable programmable device) and a memory 620. In some embodiments, the controller 310 also includes an arc fault detection circuit 530 and a zero crossing detection circuit 602. That is, in some embodiments, the arc fault detection circuit 530 is integrated within the controller 310.

[0040] The memory 620 includes, for example, a program storage area and a data storage area. The program storage area and the data storage area may include a combination of different types of memory, such as a read-only memory (ROM) and a random access memory (RAM). Various non-transitory computer-readable media may be used, such as magnetic memory, optical memory, physical memory, or electronic memory. The electronic processor 615 is communicatively coupled to the memory 620 and executes software instructions stored in the memory 620 or stored in another non-transitory computer-readable medium, such as another memory or an optical disk. The instructions may include instructions that, when executed by the processor 615, cause the control system 600 to implement any of the various arc fault detection actions described herein. The software may include one or more applications, program data, filters, rules, one or more program modules, and other executable instructions.

[0041] In some embodiments, memory 620 stores a machine learning model to be implemented by electronic processor 615. In some embodiments, the machine learning model is executed by electronic processor 615 to cause control system 600 to detect arc faults within AFCI outlet 100 and / or the circuit to which AFCI outlet 100 is connected. More particularly, the machine learning model may be executed by electronic processor 615 to cause controller 310 and / or arc detection circuit 530 to detect arc faults within AFCI outlet 100 and / or the circuit to which AFCI outlet 100 is connected. The machine learning model may be implemented as, for example, a neural network, a fuzzy logic model, a convolutional network, or other such model trained to detect arc faults as detailed herein.

[0042] The phase line contact arm or terminal 220 and the neutral line contact arm or terminal 225 are configured to receive line power from a power source. The first coil 520 and the second coil 525 are arranged to monitor the current flowing through the phase line terminal 220 and the neutral line terminal 225, respectively. As described above, the phase line terminal support contact 240 and the neutral line terminal support contact 245 are selectively connected to the load contacts 250, 255 supported by the phase load terminal 230 and the neutral load terminal 235 via the interrupt device 210. The phase line load terminal 230 and the neutral line load terminal 235 are configured to power an external load connected to the socket 625 of the AFCI outlet 100. The zero crossing detection circuit 602 is configured to measure the line voltage and frequency of the line terminals 220, 225 and / or the load terminals 230, 235.

[0043] In some embodiments, the first coil 520 and the second coil 525 are arranged to monitor current flowing through the phase load terminal 230 and the neutral load terminal 235. In some embodiments, the first coil 520 and the second coil 525 are arranged to monitor current flowing through the phase line terminal and the neutral line terminal, respectively, and the third coil and the fourth coil (not shown) are arranged to monitor current flowing through the phase load terminal 230 and the neutral load terminal 235, respectively. The current measurements taken by the first coil 520 and the second coil 525 are provided to the arc fault detection circuit 530 and / or the controller 310. In some embodiments, the arc fault detection circuit 520 also includes a zero crossing detection circuit 602. That is, in some embodiments, the zero crossing detection circuit 602 is integrated within the arc fault detection circuit 530.

[0044] The power circuit 605 is configured to convert line power to a nominal power for use by the controller 310. For example, the power circuit 605 may include a rectifier configured to rectify the line power to a nominal power for powering the controller 310. In some embodiments, the power circuit 605 rectifies alternating current (AC) power to a nominal direct current (DC) power. In some embodiments, the power circuit 605 includes one or more additional conversion circuits for converting line power to one or more additional power levels for use by the control system 600.

[0045] The communication circuit 610 is configured to provide communication between the AFCI outlet 100 and one or more external devices (e.g., other outlets, electrical devices, external computers, smart phones, tablets, etc.). For example, the communication circuit 610 is configured to provide communication between the AFCI outlet 100 and the external device 635. In the illustrated embodiment, the external device 635 is shown as a laptop computer including an electronic processor and a memory. However, it should be understood that the external device 635 can be implemented as one or more of the examples mentioned above.

[0046] In such embodiments, the AFCI outlet 100 communicates with one or more external devices via a network using, for example, the transceiver 630. The network is, for example, a wide area network (WAN) (e.g., the Internet, a TCP / IP-based network, a cellular network such as, for example, a Global System for Mobile Communications [GSM] network, a General Packet Radio Service [GPRS] network, a Code Division Multiple Access [CDMA] network, an Evolution Data Optimized [EV-DO] network, an Enhanced Data Rates for GSM Evolution [EDGE] network, a 3GSM network, a 4GSM network, a Digital Enhanced Cordless Telecommunications [DECT] network, a Digital AMPS [IS-136 / TDMA] network, or an Integrated Digital Enhanced Network [iDEN], etc.). In other embodiments, the network is, for example, a Local Area Network (LAN), a Neighborhood Area Network (NAN), a Home Area Network (HAN), or a Personal Area Network (PAN) using any of a variety of communication protocols such as Wi-Fi, Bluetooth, ZigBee, etc. In yet another embodiment, the network includes one or more of a wide area network (WAN), a local area network (LAN), a neighborhood area network (NAN), a home area network (HAN), or a personal area network (PAN). In some embodiments, the communication circuit 610 communicates with the external device 635 using a wired connection.

[0047] In some embodiments, the transceiver 630 is configured to enable wireless communication between the AFCI outlet 100 and the external device 635 using the wireless communication link 640. In other embodiments, the AFCI outlet 100 includes separate transmitting and receiving components, such as a transmitter and a receiver, instead of the transceiver 630. In operation, the controller 310 is configured to control the communication circuit 610 to transmit and receive data to and from the AFCI outlet 100.

[0048] In some embodiments, the control system 600 also includes an oscilloscope 645. The oscilloscope 645 is configured to measure current, voltage, frequency, and / or other electrical characteristics of the AFCI outlet 100. In the illustrated embodiment, the oscilloscope 645 is shown as measuring line-side current and voltage; however, it should be understood that the oscilloscope 645 can also be configured to measure load-side characteristics of the AFCI outlet 100. As shown, the oscilloscope 645 is configured to provide current measurements and voltage measurements to the external device 635 via a communication link 650, which can be implemented as a wireless or wired connection.

[0049] During operation of the AFCI outlet 100, the control system 600 can be in a standby mode or an operating mode. When in the standby mode, the interrupt device 210 electrically disconnects the line terminals 220, 225 from the load terminals 230, 235. Thus, in the standby mode, power is not provided to the outlet 625. When in the operating mode, the interrupt device 220 electrically connects the line terminals 220, 225 to the load terminals 230, 235. Thus, in the operating mode, power is provided to the outlet 625, and thus to the external load electrically connected to the outlet 625.

[0050] When power is provided to the outlet 625, the control system 600 is configured to monitor the occurrence of arc faults within the AFCI outlet 100 and / or the circuit to which the AFCI outlet 100 is connected. In particular, the arc fault detection circuit 530 and / or the controller 310 are configured to perform one or more arc fault detection processing techniques to determine whether an arc fault exists. For example, when determining whether an arc fault exists, the arc fault detection circuit 530 and / or the controller 310 can be configured to extract and analyze measurements indicating the volatility, power, and frequency content of the external load from the load terminals 230, 235. In addition, for another example, the arc fault detection circuit 530 and / or the controller 310 are configured to employ statistical and spectral analysis to analyze the line and / or load current measurements recorded by the first coil 520 and the second coil 525. The arc fault detection circuit 530 and / or the controller 310 analyzes the line and load terminal current, voltage, and / or frequency in the digital domain to reduce noise. As described below, the controller 310 may also be configured to determine if an arc fault exists by applying analysis of line and / or load electrical characteristics to a deep learning / machine learning model.

[0051] Figure 7 A process or operation 700 for detecting the presence of an arc fault within an AFCI outlet 100 and / or a circuit to which the AFCI outlet 100 is connected is shown in accordance with some embodiments. Although operation 700 is described as being performed in part by controller 310, operation 700 may also be performed by a controller included in arc fault detection circuit 530, and / or a combination of controller 310, a controller included in arc fault detection circuit 530, and one or more other components included in control system 600. It should be understood that the order of the steps disclosed in operation 700 may vary. For example, although shown as occurring in a serial order, in other embodiments, the disclosed steps may be performed in a parallel order. Furthermore, in some embodiments, additional steps may be added to the process.

[0052] At block 705, a broadband current sensor, such as the first coil 520 and the second coil 525, measures the current flowing through the line terminals 220, 225 (block 705). As described above, in some embodiments, the first coil 520 and the second coil 525 instead measure the current flowing through the load terminals 230, 235. In some embodiments, the first coil 520 and the second coil 525 measure the current flowing through the line terminals 220, 225, and the third coil and other coils measure the current flowing through the load terminals 230, 235. The current measurements are provided to the arc fault detection circuit 530 and / or the controller 310 for digitization and further processing and analysis.

[0053] At block 710, the zero-crossing detection circuit 602 measures the voltage and frequency of the line terminals 220, 225. In some embodiments, the zero-crossing detection circuit 602 instead measures the voltage and frequency of the load terminals 230, 235. In some embodiments, the zero-crossing detection circuit 602 measures the voltage and frequency of both the line terminals 220, 225 and the load terminals 230, 235. As described above, the zero-crossing detection circuit 602 may be implemented as a separate circuit integrated within the arc fault detection circuit 530 or integrated within the controller 310. The voltage and frequency measurements are provided to the arc fault detection circuit 530 and / or the controller 310 for further processing and analysis.

[0054] At block 715, a digital filter is applied to the line current measurements taken by the first coil 520 and the second coil 525. That is, the line current measurements are digitized and filtered. In some embodiments, the arc fault detection circuit 530 applies a digital filter to the line current measurement signal, while in other embodiments, the controller 310 applies a digital filter to the line current measurement signal. In some embodiments, the controller 310 is also configured to implement multi-rate analysis (e.g., downsampling / decimation and upsampling / interpolation) to optimize the digital filter applied to the line current measurement signal. For example, the controller 310 can be configured to phase lock an analog-to-digital conversion (ADC) sampling rate with the line frequency measured by the zero crossing detection circuit 602 so that the current samples used to calculate the spectral density of the current are synchronized with the line voltage. In some embodiments, the arc fault detection circuit 530 includes a logic component capable of phase locking the ADC sampling rate, such as a separate controller.

[0055] At block 720, the controller 310 is configured to estimate or calculate the spectral density of the line current measurement signal (block 720). In some embodiments, the controller 310 is configured to calculate the spectral density of the line current using the Welch method. In other embodiments, the controller 310 is configured to calculate the spectral density of the line current using a short-time Fourier transform (STFT). In some embodiments, a controller included in the arc fault detection circuit 530 calculates the spectral density of the line current.

[0056] As described above, the controller 310 synchronizes the ADC sampling rate of the line current with the measured line voltage while calculating the spectral density of the line current. Therefore, in an embodiment where the controller 310 uses STFT to calculate the spectral density of the line current, the line voltage zero crossing is located at the center of the Fourier transform window. For example, Figure 8 The STFT window of voltage-time is shown. As shown, the fast Fourier transform (FFT) window used to calculate the STFT is centered on the zero crossing of the line voltage. An FFT is calculated once every half AC (AC) cycle (e.g., approximately every 8.33ms in a 60Hz system), and a particular FFT window consists of one or more AC cycles that overlap with the previous FFT window. Since the FFT assumes a periodic signal, symmetry is preserved and spectral leakage is minimized by synchronizing and aligning the voltage zero crossing with the center of the FFT window.

[0057] In some embodiments, when using the STFT or Welch method to estimate the line current spectral density, the controller 310 periodically calculates 'n' frequency bin values. That is, the controller 310 may be configured to calculate 'n' frequency bin values ​​for each time interval 'm'. Fig. 9 An example of how the STFT can be used to estimate the spectral density of the line current is shown. As shown, at each time value or point along the time axis, a value V is calculated for time 'm' and frequency bin 'n' m,n For example, at time 0, the value V is calculated 0,0 、V 0,1 、V 0,2 ……V 0,n Similarly, at time 1, the value V is calculated 1,0 、V 1,1 、V 1,2 ……V 1,n , and at time m, calculate the value V m,0 、V m,1 、V m,2 ……V m,n .

[0058] At block 725, the controller 310 is configured to analyze the spectral density of the line current using statistical (e.g., standard deviation) analysis and spectral (e.g., spectral coefficient) analysis. The analysis of the spectral density of the line current may include estimating or calculating one or more arc characteristics or features of the line current spectral density.

[0059] In some embodiments, the controller 310 is configured to estimate the line current fluctuation by calculating the standard deviation of the line current spectral density magnitude, and thereby estimate the load power fluctuation. Fig. 9 , the controller 310 may be configured to calculate a set of values ​​V m,n The standard deviation of V can be used to determine the volatility of the amplitude for a specific frequency bin. m-x,0 To V m,0 The standard deviation of is calculated as , where 'x' is the size of the sampling history (eg, typically between 4 and 32). This standard deviation process can be repeated for frequency bins 1 to 'n' (eg, taking V m-x,n To V m,n Therefore, the controller 310 can be configured to quantify the load power fluctuation by estimating the standard deviation of the line current spectral density amplitude.

[0060] In some embodiments, the controller 310 is configured to calculate a ratio or total percentage of the amplitude at a particular frequency bin 'n' to the weighted sum of the amplitudes across the frequency bins. In other words, the controller 310 may be configured to determine a percentage 'x%' or ratio of the frequency bin amplitude attributed to a particular frequency bin value. For example, referring to Fig. 9 , the controller 310 can be configured to pass V 15,1 Divide by V 15,0 To V 15,7 The sum of V 15,1 ratio.

[0061] In some embodiments, the controller 310 is configured to implement different methods of quantizing volatility by using cepstral coefficients. In such embodiments, the cepstral coefficients can be calculated by taking a discrete cosine transform of the line current STFT directly or using a Mel frequency scale (MFCC) conversion. Combining or converting the line current spectrum to a Mel frequency scale or some other scale reduces the amount of data required for processing by the controller 310.

[0062] At block 730, these arc characteristics (e.g., volatility, standard deviation, frequency bin amplitude ratio, and cepstral coefficients) are used by the controller 310 to predict the presence of an arc fault. In some embodiments, the controller 310 is configured to compare one or more of these arc characteristics with one or more thresholds. For example, the controller 310 may be configured to compare one or more of the calculated cepstral coefficients, the spectral density amplitude indicating load power volatility, the standard deviation of the line current spectral density, and / or the frequency bin amplitude ratio with one or more corresponding thresholds. Based on the comparison of the arc characteristics with the one or more thresholds, the controller 310 is configured to determine whether an arc fault exists (block 730). In some embodiments, the controller 310 is configured to determine that an arc fault exists when at least one of the cepstral coefficients, the spectral density amplitude volatility, the standard deviation of the line current spectral density, and / or the frequency bin amplitude ratio exceeds a corresponding threshold. In some embodiments, the controller 310 determines that an arc fault exists when two or more of the cepstral coefficients, the spectral density amplitude volatility, the standard deviation of the line current spectral density, and / or the frequency bin amplitude ratio exceed a corresponding threshold. In some embodiments, the controller included in the arc fault detection circuit 530 is configured to compare the arc characteristics to one or more thresholds. When the controller 310 and / or the arc fault detection circuit 530 determines that an arc fault exists (e.g., predicts that the probability of an arc fault existing exceeds one or more thresholds), the controller 310 activates the interrupt device 210 to separate the line terminals 220, 225 from the load terminals 230, 235 (block 735).

[0063] In some embodiments, determining whether an arc fault exists includes predicting a probability of the presence of an arc fault based on the calculated arc characteristics. For example, in some embodiments, the controller 310 is configured to estimate the probability of an arc fault occurring based on a comparison of the arc characteristics with one or more thresholds. In such embodiments, the controller 310 is configured to determine that an arc fault is occurring when it is determined that the probability of an arc fault occurring exceeds a threshold value (e.g., 90%).

[0064] In some embodiments, the controller 310 is configured to execute a machine learning model when predicting the presence of an arc fault based on the determined arc characteristics of the line current. Fig.10 The general structure of the machine learning model 1000 is shown. The machine learning model 1000 is depicted as a neural network; however, the machine learning model can be implemented as any one or more of a deep learning algorithm, a neural network, a support vector machine, and a long short-term memory.

[0065] As shown, N arc features or characteristics of the line current spectral density (such as MFCC, frequency bin amplitude, volatility / standard deviation of frequency bin amplitude, and ratio of frequency bin amplitude to sum) can be provided as input 1005 to the machine learning model 1000. The machine learning model 1000 also includes multiple hidden layers 1010, each hidden layer 1010 including an independent number of neurons. For example, the machine learning model 1000 may include 2 to 4 hidden layers 1010, each layer having 2 to 256 nodes. The hidden layer 1010 is configured to generate M outputs 1015 based on one or more weights, biases, and / or thresholds associated with the arc characteristics provided as input 1005. As will be described in more detail later, the weights, biases, and / or thresholds used to calculate the outputs 1015 are generated using a supervised training algorithm executed by an external device (such as external device 635). The outputs 1015 represent the probability of a binary classification or a multi-class classification of the load current, including arc faults, normal operation, and / or other classifications of the load. That is, output 1015 indicates whether the load current is or may be experiencing an arc fault state or a normal operating state. In some embodiments, output 1015 includes the probability of an arc fault existing in the AFCI outlet 100 and / or the circuit to which the AFCI outlet 100 is connected. In such embodiments, controller 310 is configured to determine that an arc fault exists when the probability exceeds a probability threshold. In some embodiments, the probability threshold is a configurable value determined by a user. In other embodiments, the probability threshold is determined during the creation of the machine learning model 1000.

[0066] In some embodiments, the controller 310 is configured to implement more than one arc fault detection method when determining whether an arc fault exists within the AFCI outlet 100 and / or the circuit to which the AFCI outlet 100 is connected. For example, the controller 310 can be configured to implement a machine learning probability-based arc fault detection method, such as a method similar to the method described above with respect to operation 700. In some embodiments, the controller 310 is also configured to implement a correlation-based arc fault detection method, such as the arc fault detection method described in U.S. Patent Application Publication No. 2020 / 0036183, entitled “SYSTEM AND METHOD FOR DISCERNING ARCING INELECTRICAL WIRING,” published on January 30, 2020, the entire contents of which are incorporated herein by reference. In some embodiments, the controller 310 is configured to implement the correlation-based arc fault detection method described in U.S. patent application publication No. 2020 / 0264234, entitled “APPARATUSES AND METHODS FOR PASSIVE FAULT MONITORING OF CURRENTSENSING DEVICES IN PROTECTIVE CIRCUIT INTERRUPTERS,” published on August 20, 2020, the entire contents of which are incorporated herein by reference.

[0067] Fig.11 A process (or operation) 1100 for detecting the presence of an arc fault within an AFCI outlet 100 and / or a circuit to which the AFCI outlet 100 is connected by using two or more different arc fault detection methods is shown in accordance with some embodiments. Although operation 1100 is described as being performed in part by controller 310, operation 1100 may also be performed by a controller included in arc fault detection circuit 530, and / or a combination of controller 310, controller included in arc fault detection circuit 530, and one or more other components included in control system 600. It should be understood that the order of the steps disclosed in operation 1100 may vary. For example, although shown as occurring in a serial order, in other embodiments, the disclosed steps may be performed in a parallel order. Furthermore, in some embodiments, additional steps may be added to the process.

[0068] At block 1105, the voltage and / or frequency of the line terminals 220, 225 are measured. In some embodiments, the zero-crossing detection circuit 602 measures the voltage and frequency on the line side. In some embodiments, the zero-crossing detection circuit 602 measures the voltage and frequency on the load side instead. In some embodiments, the zero-crossing detection circuit 602 measures the voltage and frequency on the line side and the voltage and frequency on the load side. As described above, the zero-crossing detection circuit 602 can be implemented as a separate circuit integrated within the arc fault detection circuit 530 or integrated within the controller 310.

[0069] At block 1110, an analog-to-digital converter (ADC) sampling clock included in the controller 310 is phase locked to the line voltage frequency. In some embodiments, the ADC sampling clock is a component in the arc fault detection circuit 530. In some embodiments, the ADC sampling clock is a separate component included in the control system 600.

[0070] At block 1115, a broadband current sensor, such as the first coil 520 and the second coil 525, measures the current flowing through the line terminals 220, 225. As described above, in some embodiments, the first coil and the second coil 520, 525 measure the current flowing through the load terminals 230, 235. In some embodiments, the first coil and the second coil 520, 525 measure the current flowing through the line terminals 220, 225, and the third coil and other coils measure the current flowing through the load terminals 230, 235.

[0071] At block 1120, two or more buffers included in the memory 620 are used to store line and / or load current sample data until a zero crossing edge (1 / 2 AC cycle) is detected. Thus, due to the phase-locked ADC sampling clock, the current measurement is fixed between the rising and falling edges of the zero crossing detector.

[0072] At block 1125, the controller 310 determines whether a zero crossing is detected. When a zero crossing has not been detected, the process 1100 returns to block 1120 where the current sampling data is buffered. However, if a zero crossing is detected at block 1125, the operation 1100 proceeds to block 1130.

[0073] At block 1130, the controller 310 applies a digital filter to the buffered current data. That is, the line current measurement is digitized and filtered. In some embodiments, the arc fault detection circuit 530 applies a digital filter to the line current measurement signal. In some embodiments, the controller 310 is also configured to implement multi-rate analysis (e.g., downsampling / decimation and upsampling / interpolation) to optimize the digital filter applied to the line current measurement signal.

[0074] At block 1135, the current data from one 1 / 2 AC cycle is appended to the current data from the previous 1 / 2 AC cycle. The appended current data is then provided as input to both the machine learning (ML) or inference based arc fault detection method (blocks 1140 to 1160) and the correlation based arc fault detection method (blocks 1165 to 1170).

[0075] At block 1140, the ML-based arc fault detection method begins with the controller 310 estimating the spectral density of the line current. As described above with respect to operation 700, the controller 310 may be configured to use the Welch method or short-time Fourier transform for calculating the spectral density of the line current. In some embodiments, the controller included in the arc fault detection circuit 530 calculates the spectral density of the line current.

[0076] At block 1145, the controller 310 is configured to convert and combine the frequency bins of the line current spectral density using the Mel frequency conversion scale or some other conversion scale. As described above with respect to operation 700, combining or converting the line current spectral density to the Mel frequency scale or some other scale reduces the amount of data that needs to be processed by the controller 310. In some embodiments, the controller included in the arc fault detection circuit 530 is configured to estimate the spectral density of the line current.

[0077] At block 1150, the controller 310 calculates one or more arc features or characteristics associated with the spectral density of the line current. As described above with respect to operation 700, these arc features include, but are not limited to, cepstral coefficients, spectral density magnitudes indicating load power fluctuations, standard deviations of the line current spectral density, and frequency bin magnitude ratios. These arc features may be calculated using the same or similar methods as described above with respect to operation 700.

[0078] At block 1155, the controller 310 is configured to infer one or more arc characteristics of the line current spectral density and output an arc probability. That is, the controller 310 executes a machine learning model, such as the one described above with respect to Fig.10The machine learning model 1000 described. One or more arc characteristics are provided as input 1005 to the machine learning model 1000, which produces an output 1015. In some embodiments, the output 1015 includes one or more probabilities indicating whether an arc is occurring within the AFCI outlet 100 and / or the circuit to which the AFCI outlet 100 is connected. In some embodiments, the output 1015 includes a single probability value indicating the probability of whether an arc is occurring. In some embodiments, the output 1015 includes more than one probability value indicating whether an arc is occurring. In some embodiments, the controller included in the arc fault detection circuit 530 infers one or more arc characteristics of the line current spectral density.

[0079] At block 1160, the controller 310 is configured to aggregate the arc probability values ​​calculated by inference. More specifically, the controller 310 is configured to set one or more arc fault flags and override flags based on the probability output 1015 of the machine learning model 1000. The arc fault flags and override flags (which will be described in more detail below) include an arc fault flag, an arc fault override flag, and a no arc override flag. In some embodiments, these flags are temporarily stored in the memory 620. In some embodiments, these flags are stored in a log that is stored in the memory 620. In some embodiments, a controller included in the arc fault detection circuit 530 sets these flags.

[0080] In some embodiments, the controller 310 is configured to set an arc fault flag when the controller 310 determines that the probability of an arc fault occurring exceeds a first percentage threshold value within at least a first amount of time during the time window. For example, the controller 310 may be configured to set an arc fault flag when the probability of an arc fault occurring exceeds 70% within at least half a 1s time window. In some embodiments, the controller 310 is further configured to set an arc fault override flag when the probability of an arc fault occurring exceeds a second threshold value within at least a second amount of time during the time window. For example, the controller 310 may be configured to set an arc fault override flag when the probability of an arc fault occurring exceeds 80% within a 1s time window of at least 0.9s. In some embodiments, the first threshold and the second threshold have the same value. In other embodiments, the first threshold and the second threshold have different values. In some embodiments, the controller 310 is further configured to set a no arc override flag when the probability of an arc fault occurring is less than a third threshold value within at least a third amount of time during the time window. For example, the controller 310 may be configured to set the arc fault flag when the probability of an arc fault occurring is less than 30% within a 1 s time window of at least 0.5 s.

[0081] Fig.12An exemplary process (or operation) 1200 for aggregating arc probability values ​​calculated at block 1155 is shown in accordance with some embodiments. Although operation 1200 is described as being performed in part by controller 310, operation 1200 may also be performed by a controller included in arc fault detection circuit 530, and / or a combination of controller 310, controller included in arc fault detection circuit 530, and one or more other components included in control system 600. It should be understood that the order of the steps disclosed in operation 1200 may vary. For example, although shown as occurring in a serial order, in other embodiments, the disclosed steps may be performed in a parallel order. Furthermore, in some embodiments, additional steps may be added to the process.

[0082] At block 1205, arc fault probability is calculated every 1 / 2 AC cycle using inference (eg, neural network 1000). For example, in a 60 Hz system, arc fault probability is calculated every 8.3 ms.

[0083] At block 1210, the controller 310 determines whether the arc fault probability is greater than an arc fault presence threshold. As will be described in more detail below, in some embodiments, the arc fault presence threshold is determined during creation of a machine learning model, such as the machine learning model 1000. In some embodiments, the arc fault presence threshold is a configurable value determined by a user.

[0084] If the arc fault probability is greater than the arc fault presence threshold, the controller 310 is configured to add an arc fault presence flag to a moving window buffer (box 1215). The moving window buffer is configured to contain or store 'n' flags, where 'n' is a number greater than zero. That is, the moving window buffer stores 'n' flags associated with the previous or most recent 'n' calculated arc probabilities and line current data samples. In some embodiments, the moving window buffer stores 32 flags. In such embodiments, for a 60 Hz system, the moving window buffer contains approximately 266.6 ms of line current data (e.g., 32*8.3 ms). In some embodiments, the moving window buffer stores more than 32 flags. In other embodiments, the moving window buffer stores less than 32 flags. In some embodiments, the moving buffer window is stored in a memory 620 of the controller 310. In some embodiments, the moving buffer window is included in a machine learning model 1000 executed by the controller 310. In some embodiments, the moving buffer window is a component of the control system 600.

[0085] At block 1220, the controller 310 is configured to count the number of arc fault presence flags saved in the moving buffer window. At block 1225, the controller 310 determines whether the count of arc fault presence flags saved in the window exceeds an arc fault count threshold. In some embodiments, the arc fault count threshold is determined during creation of the machine learning model 1000, which will be described in more detail later. In some embodiments, the arc fault count threshold is a configurable value determined by a user.

[0086] If the controller 310 determines that the count is not greater than the arc fault count threshold, the operation 1200 proceeds to block 1245. However, if the controller 310 determines that the count is greater than the arc fault count threshold, the controller 310 is configured to set the arc fault flag (block 1230). In some embodiments, the arc fault flag is a binary value stored in the memory 620. In some embodiments, the arc fault flag is stored in a register of the memory 620. In some embodiments, when the arc fault flag is set, the value of the arc fault flag is '1' or 'TRUE'. Likewise, when the count is not greater than the arc fault count threshold and the arc fault flag is not set, the value of the arc fault flag is '0' or 'FALSE'.

[0087] At block 1235, the controller 310 determines whether the count of arc fault presence flags saved in the moving buffer window exceeds an arc fault override count threshold. The value of the arc fault override count threshold is greater than the value of the arc fault count threshold. In some embodiments, the value of the arc fault override count threshold exceeds the value of the arc fault count threshold by a configurable amount. In some embodiments, the arc fault override count threshold is determined during creation of the machine learning model 1000, which will be described in more detail later. In some embodiments, the arc fault override count threshold is a configurable value determined by a user.

[0088] If the controller 310 determines that the count is not greater than the arc fault override count threshold, the operation 1200 proceeds to block 1245. However, if the controller 310 determines that the count is greater than the arc fault override count threshold, the controller 310 is configured to set the arc fault override flag (block 1240). In some embodiments, the arc fault override flag is a binary value stored in the memory 620. In some embodiments, the arc fault flag is stored in a register of the memory 620. In some embodiments, when the arc fault override flag is set, the value of the arc fault override flag is '1' or 'TRUE'. Likewise, when the count is not greater than the arc fault override count threshold and the arc fault override flag is not set, the value of the arc fault override flag is '0' or 'FALSE'.

[0089] At block 1245, the controller 310 determines whether the arc fault probability is less than a no arc override threshold. As will be described in more detail below, in some embodiments, the no arc override threshold is determined during creation of the machine learning model 1000. In some embodiments, the no arc override threshold is a configurable value determined by a user.

[0090] If the arc fault probability is less than the no arc fault presence threshold, the controller 310 is configured to add a no arc fault presence flag to the no arc moving window buffer (block 1250). In some embodiments, the no arc moving window buffer is the same as the moving window buffer. In other embodiments, the no arc moving window buffer is separate from the moving window buffer. In some embodiments, the no arc moving window buffer is configured to contain or save 'n' flags, where 'n' is a number greater than zero. In some embodiments, the no arc moving window buffer saves 32 flags. In some embodiments, the moving window buffer saves more than 32 flags. In other embodiments, the no arc moving window buffer saves less than 32 flags. In some embodiments, the no arc moving buffer window is stored in the memory 620 of the controller 310. In some embodiments, the no arc moving buffer window is included in a machine learning model executed by the controller 310. In some embodiments, the no arc moving buffer window is a component of the control system 600.

[0091] At block 1255, the controller 310 determines whether the arc fault probability is greater than a no arc clearing threshold. As will be described in more detail below, in some embodiments, the no arc clearing threshold is determined during creation of the machine learning model 1000. In some embodiments, the no arc clearing threshold is a configurable value determined by a user.

[0092] If the arc fault probability is greater than the no arc clearing threshold, the controller 310 is configured to clear or reset the no arc moving window buffer (block 1260). Thus, when the arc fault probability is greater than the no arc clearing threshold, the controller 310 removes any no arc fault presence flag from the no arc moving window buffer.

[0093] At block 1265, the controller 310 is configured to count the number of no arc fault presence flags saved in the no arc moving buffer window. At block 1270, the controller 310 determines whether the count of no arc fault presence flags saved in the window exceeds a no arc override count threshold. In some embodiments, the no arc override count threshold is determined during creation of the machine learning model 1000, which will be described in more detail later. In some embodiments, the no arc override count threshold is a configurable value determined by a user.

[0094] If the controller 310 determines that the no arc presence flag count is greater than the no arc override count threshold, the controller 310 is configured to set the no arc override flag (block 1275). In some embodiments, the no arc override flag is a binary value stored in the memory 620. In some embodiments, the no arc override flag is stored in a register of the memory 620. In some embodiments, when the no arc override flag is set, the value of the no arc override flag is '1' or 'TRUE'. Similarly, when the no arc fault presence flag count is less than or equal to the no arc override count threshold and the no arc override flag is not set, the value of the no arc override flag is '0' or 'FALSE'.

[0095] It should be understood that the no arc override flag and the arc fault flag are mutually exclusive. That is, the no arc override flag and the arc fault cannot be set at the same time. For example, the clearing / resetting of the no arc moving window buffer that occurs in block 1260 ensures that the no arc override flag is not set when the arc fault flag has been set.

[0096] Referring again to operation 1100, the filtered and additional line current buffer data (block 1135) is also provided as an input to a second arc fault detection method (block 1165) included in operation 1100. As described above, the second arc fault detection is a correlation-based arc fault detection method performed in the time domain. For example, the second arc fault detection method is implemented as one or more correlation-based arc fault detection methods described in U.S. Patent Application Publication No. 2020 / 0036183, entitled “SYSTEM AND METHOD FOR DISCERNING ARCING INELECTRICAL WIRING,” published on January 30, 2020, the entire contents of which are incorporated herein by reference and partially reproduced below. In some embodiments, the second arc fault detection method is implemented as one or more correlation-based arc fault detection methods described in U.S. patent application publication No. 2020 / 0264234, entitled “APPARATUSES AND METHODS FOR PASSIVE FAULT MONITORING OF CURRENT SENSING DEVICES IN PROTECTIVE CIRCUIT INTERRUPTERS,” published on August 20, 2020, the entire contents of which are incorporated herein by reference and partially reproduced below.

[0097] At block 1165, the controller 310 performs or implements a correlation-based arc fault detection method using the current data provided from block 1135. In some embodiments, the controller included in the arc fault detection circuit 530 performs a correlation-based arc fault detection method.

[0098] At block 1170, the controller 310 is configured to output the results of the correlation-based arc fault detection method. In some embodiments, the output includes the status of a correlation arc fault flag and / or the status of a correlation arc override flag. Similar to the machine learning (ML) arc flag or ML flag described above, the correlation arc fault flag and the correlation arc override flag have a value equal to '1' or 'TRUE' when they are set by the controller 310. In some embodiments, the output includes the cause of the arc fault, which can be detected based on the load current magnitude.

[0099] At block 1175, the controller 310 is configured to combine the output of the ML-based arc fault detection method (blocks 1140-1160) with the output of the correlation-based arc fault detection method (blocks 1165-1170). In some embodiments, combining the output of the ML-based arc fault detection method with the output of the correlation-based arc fault detection method includes developing a relationship between the outputs of the two different arc fault detection methods. In such embodiments, the controller 310 may be configured to provide the outputs of the two arc fault detection methods (e.g., values ​​of the arc fault flag) to one or more of a set of rules, weights, biases, thresholds, and / or truth tables that define the relationship between the ML-based arc fault detection method and the correlation-based arc fault detection method.

[0100] Fig.13 An exemplary process or operation 1300 for combining the output of an ML-based arc fault detection method with the output of a correlation-based arc fault detection method according to some embodiments is shown. Although operation 1300 is described as being performed in part by controller 310, operation 1300 may also be performed by a controller included in arc fault detection circuit 530, and / or a combination of controller 310, controller included in arc fault detection circuit 530, and one or more other components included in control system 600. It should be understood that the order of the steps disclosed in operation 1300 may vary. For example, although shown as occurring in a serial order, in other embodiments, the disclosed steps may be performed in a parallel order. Furthermore, in some embodiments, additional steps may be added to the process.

[0101] At block 1305, the controller 310 provides the output of the correlation-based arc fault detection method to the input of the decision tree or truth table 1302. In some embodiments, the output of the correlation-based arc fault detection method includes the value or state (e.g., 'TRUE' or 'FALSE') of the correlation arc fault flag and the correlation arc override flag. In some embodiments, additional outputs of the correlation-based arc fault detection method are provided to the truth table 1302.

[0102] At block 1310, the controller 310 provides the output of the ML-based arc fault detection method to the truth table 1302. In some embodiments, the output of the ML-based arc fault detection method includes a value or state (e.g., 'TRUE' or 'FALSE'), an ML arc fault flag, an ML arc fault override flag, and an ML no arc override flag. The ML arc fault flag, the ML arc fault override flag, and the ML no arc override flag are the same as the arc fault flag, the arc fault override flag, and the no arc override flag described above with respect to operation 1200, respectively. The identifier 'ML' is used to distinguish the output of the correlation-based method from the ML-based arc fault detection method. In some embodiments, additional outputs of the ML-based arc fault detection method are provided to the truth table 1302.

[0103] At block 1315, the controller 310 is configured to calculate a probability of an arc fault based on the window buffer counts from the correlation-based arc fault detection method. The probability may be referred to hereinafter as the correlation arc fault count probability. In some embodiments, the controller 310 calculates the probability by dividing the number of counts by a correlation probability threshold. In some embodiments, the correlation probability threshold is determined during creation or training of the machine learning model 1000. In some embodiments, the correlation probability threshold is a configurable value determined by a user.

[0104] At block 1320, the controller 310 is configured to calculate an ML arc fault count probability based on the counts of arc fault presence flags saved in the moving window buffer. In some embodiments, the controller 310 calculates the probability by dividing the number of counts by an ML count probability threshold. In some embodiments, the ML count probability threshold is determined during creation or training of the machine learning model 1000. In some embodiments, the machine learning count probability threshold is a configurable value determined by a user.

[0105] At block 1325, the controller 310 is configured to determine whether the sum of the ML count probability and the correlation arc fault count probability is greater than a combined arc fault probability threshold. In some embodiments, the combined arc fault threshold is determined during creation or training of the machine learning model 1000. In some embodiments, the combined arc fault probability threshold is a configurable value determined by a user. If the sum of the probabilities exceeds the combined arc fault probability threshold, the controller 310 sets the combined arc fault flag (e.g., the combined arc fault flag is 'TRUE') and provides the state of the combined arc fault flag to the truth table 1302 (block 1330). If the sum of the probabilities does not exceed the combined arc fault probability threshold, the controller 310 does not set the combined arc fault flag (e.g., the combined arc fault flag is 'FALSE') and provides the state of the combined arc fault flag to the truth table 1302 (block 1335).

[0106] Referring again to operation 1100, at block 1180, the controller 310 is configured to determine whether an arc or arc fault exists within the AFCI outlet 100 and / or the circuit to which the AFCI outlet 100 is connected. In particular, the controller 310 is configured to determine whether an arc fault exists based on one or more of the combined ML-based arc fault detection output and the correlation-based arc fault detection output.

[0107] In some embodiments, the controller 310 determines whether an arc exists based on a relationship or rule set that defines the truth table 1302. Fig.13 As shown, in such embodiments, the controller 310 is configured to determine that an arc exists (e.g., the system arc fault is TRUE) when at least one of the following six conditions is met: (1) the combined arc fault flag is TRUE and the remaining flags are FALSE,

[0108] (2) The correlation arc fault flag and the ML arc fault flag are both TRUE, and the remaining flags are FALSE,

[0109] (3) The correlation arc fault flag, correlation arc override flag, and ML no arc override flag are all TRUE, and the remaining flags are FALSE,

[0110] (4) The correlation arc fault flag and the correlation arc override flag are both TRUE, and the remaining flags are FALSE,

[0111] (5) the dependent arc fault flag is TRUE and the remaining flags are FALSE, and

[0112] (6) The ML Arc Fault Flag and the ML Arc Override Flag are both TRUE, and the remaining flags are all FALSE.

[0113] like Fig.13 As shown, some entries in the truth table 1302 are missing (indicated by an 'X'). The missing entries are the result of mutually exclusive conditions existing and / or conditions existing that render the value of a particular flag meaningless. For example, in the first row of the truth table 1302, the data entries for the ML No Arc Override Flag, the Dependency Arc Override Flag, and the ML Arc Override Flag are all marked with an 'X' because the Combined Arc Fault Flag, the Dependency Arc Fault Flag, and the ML Arc Fault Flag are all FALSE.

[0114] In some embodiments, the controller 1310 is configured to determine that an arc exists when one or more of the following conditions are met. In some embodiments, the controller 1310 determines that an arc exists when the ML arc fault flag is set. In some embodiments, the controller 1310 determines that an arc exists when the ML arc fault flag is set and the correlation arc fault flag is set. In some embodiments, the controller 1310 determines that an arc exists when at least one of the correlation arc override flags is set and the ML arc override flag is set. In some embodiments, the controller determines whether an arc exists based on the arc probability calculated in block 1155. In such embodiments, the controller 1310 may be configured to determine that an arc exists when the arc probability exceeds a threshold, such as an arc fault existence threshold.

[0115] Referring again to operation 1100, if the controller determines at block 1180 that an arc is present, the controller 310 is configured to activate the interruption device 210 to separate the line terminals 220, 225 from the load terminals 230, 235 (block 1185). If the controller 310 determines at block 1180 that an arc is not present, operation 1100 returns to block 1105.

[0116] As described above, the controller 310 is configured to execute the machine learning model 1000 for detecting the presence of an arc fault. In some embodiments, the machine learning model 1000 is created using an external computing device, such as the external device 635. In such embodiments, the machine learning model 1000 is stored in the memory 620 of the controller 310 after creation. However, in other embodiments, the machine learning model 1000 is created by the controller 310.

[0117] Fig.14is a block diagram of a control system 1400 of an external device 635 for creating a machine learning model 1000. As described above, the external device 1400 may be implemented as one or more of a computer terminal, a desktop computer, a laptop computer, a smartphone, a tablet computer, a server, or any other electronic device including a memory and an electronic processor capable of training a machine learning model.

[0118] As shown, the control system 1400 of the external device 1400 includes a controller 1405. The controller 1405 is electrically and / or communicatively connected to various modules or electronic components of the external device 635. For example, the controller 1405 is connected to a power circuit 1410, a communication circuit 1415, and a user interface 1420.

[0119] The controller 1405 includes a number of electrical and electronic components that provide power, operational control, and protection to components and modules within the controller 1405 and / or external devices. For example, the controller includes an electronic processor 1425 and memory 1430, among other aspects.

[0120] For example, memory 1430 includes a program storage area and a data storage area. The program storage area and the data storage area may include a combination of different types of memory, such as a read-only memory (ROM) and a random access memory (RAM). Various non-transitory computer-readable media may be used, such as magnetic memory, optical memory, physical memory, or electronic memory. Electronic processor 1425 is communicatively coupled to memory 1430 and executes software instructions stored in memory 1430 or stored in another non-transitory computer-readable medium, such as another memory or optical disk. Software may include one or more applications, program data, filters, rules, one or more program modules, and other executable instructions. For example, software includes one or more programs for training a machine learning model.

[0121] The power circuit 1410 is configured to supply power to the controller 1405 and / or other components of the external device 635. In some embodiments, the power circuit 1410 receives power from a power source (e.g., a battery) and provides regulated power to the controller 1405 and / or other components of the external device 635. In some configurations, the power circuit 1410 includes a DC-DC converter, an AC-DC converter, a DC-AC converter, and / or an AC-AC converter. In some embodiments, the power circuit 1410 receives power from an AC power source (e.g., an AC power outlet).

[0122] The communication circuit 1415 enables the external device 635 to communicate with the communication circuit 610 and the oscilloscope 645 of the AFCI outlet 100. In some embodiments, the communication circuit 1415 communicates wirelessly with the communication circuit 610 and / or the oscilloscope 645 of the AFCI outlet 100. In some embodiments, the communication circuit 1415 uses a wired connection to communicate with the AFCI outlet 100 and / or the oscilloscope 645. In some embodiments, the communication circuit 1415 is capable of wireless and wired communication with the communication circuit 610 and the oscilloscope 645 of the AFCI outlet 100. In some embodiments, the communication circuit 1415 includes, for example, a transceiver that includes an antenna and / or is coupled to an antenna. In some embodiments, the communication circuit 1415 includes a port configured to receive a wired connection.

[0123] In some embodiments, the communication circuit 1415 communicates with the communication circuit 610 and / or the oscilloscope 645 of the AFCI outlet 100 using a network. The network is, for example, a wide area network (WAN) (e.g., the Internet, a TCP / IP-based network, a cellular network such as, for example, a global system for mobile communications [GSM] network, a general packet radio service [GPRS] network, a code division multiple access [CDMA] network, an evolution data optimized [EV-DO] network, an enhanced data rate for GSM evolution [EDGE] network, a 3GSM network, a 4GSM network, a digital enhanced cordless telecommunications [DECT] network, a digital AMPS [IS-136 / TDMA] network, or an integrated digital enhanced network [iDEN], etc.). In other embodiments, the network is, for example, a local area network (LAN), a neighborhood area network (NAN), a home area network (HAN), or a personal area network (PAN) using any of a variety of communication protocols (such as Wi-Fi, Bluetooth, ZigBee, etc.). In yet another embodiment, the network includes one or more of a wide area network (WAN), a local area network (LAN), a neighborhood area network (NAN), a home area network (HAN), or a personal area network (PAN). In some embodiments, the communication circuit 1415 communicates using a wired connection.

[0124] The user interface 1420 is configured to receive input from a user and / or output information about the AFCI socket 100 and / or the machine learning model 1000 to the user. The user interface 1420 includes a display (e.g., a primary display, a secondary display, etc.) and / or an input device (e.g., a keyboard, a touch screen display, a plurality of knobs, dials, switches, buttons, etc.). The display can be, for example, a liquid crystal display ("LCD"), a light emitting diode ("LED") display, an organic LED ("OLED") display, an electroluminescent display ("ELD"), a surface conduction electron emitter display ("SED"), a field emission display ("FED"), a thin film transistor ("TFT") LCD, etc.

[0125] Fig.15 A process or operation 1500 for creating an arc fault detection machine learning model such as machine learning model 1000 is shown. Although operation 1500 is described as being performed in part by controller 1405 of external device 635, operation 1500 may also be performed by controller 310, a controller included in arc fault detection circuit 530, and / or a combination of external device 635, controller 310, a controller included in arc fault detection circuit 530, and one or more other components included in control system 600. It should be understood that the order of the steps disclosed in operation 1500 may vary. For example, although shown as occurring in a serial order, in other embodiments, the disclosed steps may be performed in a parallel order. Furthermore, in some embodiments, additional steps may be added to the process.

[0126] At block 1505, when an arc condition is known to exist (e.g., during the occurrence of an arc fault), the AFCI outlet 100 measures the current, voltage, and / or frequency on the line side. In some embodiments, when an arc condition is known to exist, the AFCI outlet 100 instead measures the current, voltage, and frequency on the load side. In some embodiments, when an arc condition is known to exist, the AFCI outlet 100 measures the current, voltage, and / or frequency on both the line side and the load side simultaneously. In some embodiments, the electrical characteristics are measured by the arc fault detection circuit 530 and / or the zero crossing detection circuit 602 included in the AFCI outlet 100. In some embodiments, the arc condition is created by adjusting one or more components in the AFCI outlet 100.

[0127] At block 1510, the AFCI outlet 100 measures the current, voltage, and / or frequency on the line side when it is known that no arc is occurring (e.g., when there is no arc fault). In some embodiments, when it is known that no arc condition exists, the AFCI outlet 100 instead measures the current, voltage, and frequency on the load side. In some embodiments, when it is known that no arc condition exists, the AFCI outlet 100 measures the current, voltage, and / or frequency on both the line side and the load side. In some embodiments, the electrical characteristics are measured by the arc fault detection circuit 530 and / or the zero crossing detection circuit 602 included in the AFCI outlet 100.

[0128] At block 1515, when an arc condition is known to exist (e.g., during the occurrence of an arc fault), the oscilloscope 645 measures the current, voltage, and / or frequency on the line side. In some embodiments, when an arc condition is known to exist, the oscilloscope 645 instead measures the current, voltage, and frequency on the load side. In some embodiments, when an arc condition is known to exist, the oscilloscope 645 measures the current, voltage, and / or frequency on both the line side and the load side simultaneously. In some embodiments, the arc condition is created by adjusting one or more components in the AFCI outlet 100.

[0129] At block 1520, the oscilloscope 645 measures the current, voltage, and / or frequency on the line side when it is known that no arcing is occurring (e.g., when no arc fault is present). In some embodiments, when it is known that no arcing condition exists, the oscilloscope 645 instead measures the current, voltage, and frequency on the load side. In some embodiments, when it is known that no arcing condition exists, the oscilloscope 645 measures the current, voltage, and / or frequency on both the line side and the load side simultaneously.

[0130] At block 1525, the external device 635 receives current, voltage, and / or frequency measurements from the AFCI outlet 100 and the oscilloscope 645 via the communication circuit 1415. That is, the AFCI outlet 100 transmits the current, voltage, and / or frequency measurements taken when the arc is present (block 1505) and the current, voltage, and / or frequency measurements taken when the arc is not present (block 1510) to the external device 635 via the communication circuit 610. In some embodiments, the AFCI outlet 100 transmits the measurements to the external device 635 wirelessly. In other embodiments, the AFCI outlet 100 transmits the measurements to the external device 635 using a wired connection. Similarly, the oscilloscope 645 transmits the current, voltage, and / or frequency measurements taken when the arc is present (block 1515) and the current, voltage, and / or frequency measurements taken when the arc is not present (block 1520) to the external device 635 via the communication circuit 610. In some embodiments, the oscilloscope 645 transmits the measurements wirelessly to the external device 635. In other embodiments, the oscilloscope 645 transmits the measurements to the external device 635 using a wired connection.

[0131] Although operation 1500 includes measuring the electrical characteristics of arc current and non-arc current with both AFCI outlet 100 and oscilloscope 645, it should be understood that in some embodiments, only one device is used to measure the electrical characteristics of arc current and non-arc current. For example, in some embodiments, only oscilloscope 645 is used to measure the current, voltage and / or frequency during arc and non-arc conditions. Similarly, in some embodiments, only AFCI outlet 100 is used to measure the current, voltage and / or frequency during arc and non-arc conditions.

[0132] At block 1530, the controller 1405 of the external device 635 applies a transform and / or filter to the current, voltage, and / or frequency measurements received from the oscilloscope 645. Thus, the oscilloscope measurements are adjusted to be similar to measurements taken by an AFCI outlet.

[0133] At block 1535, the controller 1405 combines the current, voltage, and / or frequency measurements taken by the oscilloscope 645 with the measurements taken by the AFCI outlet 100. In particular, the controller 1405 combines the measurements taken by the oscilloscope 645 during an arc condition with the measurements taken by the AFCI outlet 100 during an arc condition. Similarly, the controller 1405 combines the measurements taken by the oscilloscope 645 when no arc condition exists with the measurements taken by the AFCI outlet 100 when no arc condition exists. During the following calculation steps, the arc fault presence measurements or data are saved separately from the no arc fault presence data.

[0134] At block 1540, the controller 1405 estimates the spectral density of the line current of the current measurements taken during the arc condition and the current measurements taken during the non-arcing condition. As described above with respect to operations 700 and 1100, the controller 1405 may be configured to use Welch's method or a short-time Fourier transform for calculating the current spectral density.

[0135] At block 1545, the controller 1405 is configured to convert and combine the frequency bins of the arcing and non-arcing current spectral densities using the Mel frequency conversion scale or some other conversion scale. As described above with respect to operations 700 and 1100, combining or converting the current spectral densities to the Mel frequency scale or some other scale reduces the amount of data required for processing by the controller 1405.

[0136] At block 1550, the controller 1405 calculates one or more arc features or characteristics associated with the spectral density of the arc current and the non-arcing current. As described above with respect to operations 700 and 1100, these arc features include, but are not limited to, cepstral coefficients, spectral density magnitudes indicating load power fluctuations, standard deviations of line current spectral densities, and frequency bin magnitude ratios. These arc features may be calculated using the same or similar methods as described above with respect to operations 700 and 1100.

[0137] At block 1555 , the controller 1405 is configured to separate both arc signatures of arc current and arc signatures of non-arcing current into training, testing, and validation data sets.

[0138] At block 1560, the controller 1405 executes a supervised training algorithm to train the machine learning model 1000. Training the machine learning model 1000 includes developing one or more thresholds, weights, and / or biases based on the arc current and non-arcing current data sets. For example, one or more thresholds (e.g., arc fault presence threshold, arc fault count threshold, arc fault override threshold, no arc override threshold, no arc clear threshold, etc.) during training, testing, validation, or creation of the machine learning model 1000 as described above with respect to operations 1100 and 1200.

[0139] After the machine learning model 1000 is trained or created, the controller 1405 deploys the machine learning model 1000 to the AFCI outlet 100 (block 1565). In some embodiments, the controller 1405 deploys the machine learning model 1000 to the AFCI outlet 100 wirelessly using the communication circuit 1415. In some embodiments, the controller 1405 deploys the machine learning model 1000 to the AFCI outlet 100 using a wired connection. In some embodiments, upon deployment of the machine learning model 1000, the controller 310 of the AFCI outlet 100 stores the machine learning model 1000 in the memory 620.

[0140] In some embodiments, the controller 1405 of the external device 635 is configured to update the machine learning model 1000. In such embodiments, the controller 1405 of the external device 635 is configured to repeat operation 1500 so that an updated version of the machine learning model 1000 is deployed. That is, the AFCI outlet 100 is deployed with an updated version of the machine learning model 1000, which includes updated values ​​of one or more thresholds, weights, and / or biases for detecting the presence of an arc fault. In some embodiments, the controller 1405 is configured to repeat operation 1500 periodically (e.g., daily, weekly, monthly, etc.). In some embodiments, the controller 1405 is configured to repeat operation 1500 on demand.

[0141] Fig.16 A machine learning based process or operation 1600 is shown for detecting the presence of an arc fault within an AFCI outlet 100 and / or a circuit to which the AFCI outlet 100 is connected, similar to the operation 1100 described above, according to some embodiments. Fig.16As shown, operation 1600 includes many of the same steps as operation 1100; however, operation 1600 does not include any steps related to the correlation-based arc fault detection method. That is, operation 1600 includes all of the steps from operation 1100 except any steps related to the correlation-based arc fault detection method. For example, operation 1600 includes every step from operation 1100 except performing the correlation-based arc fault detection method (block 1165), outputting the results of the correlation-based arc fault detection method (block 1170), and combining the correlation-based arc fault detection output with the ML-based arc fault detection output (block 1175).

[0142] In addition, since no correlation-based arc fault detection output is generated during operation 1600, the controller 310 is configured to determine whether an arc is present based on the aggregated arc probability output by the ML-based arc fault detection method, rather than any output from the correlation-based arc fault detection method (block 1165). Therefore, at block 1165, the controller 310 may be configured to determine whether an arc is present based on one or more states of the ML arc fault flag, the ML arc fault override flag, and the ML no arc override flag. For example, in some embodiments, the controller 1310 determines that an arc is present when the ML arc fault flag is set. In some embodiments, the controller 1310 determines that an arc is present when the ML arc fault flag is set and / or the ML arc fault override flag is set. In some embodiments, the controller 310 is configured to determine that an arc is not present when the ML no arc override flag is set. In some embodiments, the controller 310 is configured to determine that an arc is present when the probability output at block 1655 exceeds a threshold. In some embodiments, the controller 310 is configured to determine that an arc is not present when the probability output at block 1655 is less than a threshold.

[0143] Although operation 1600 is described as being performed in part by controller 310, operation 1600 may also be performed by a controller included in arc fault detection circuit 530, and / or a combination of controller 310, a controller included in arc fault detection circuit 530, and one or more other components included in control system 600. It should be understood that the order of the steps disclosed in operation 1600 may vary. For example, although shown as occurring in a serial order, in other embodiments, the disclosed steps may be performed in a parallel order. Furthermore, in some embodiments, additional steps may be added to the process.

Claims

1. A circuit interrupting device, comprising: Line terminals; a broadband current sensor configured to measure a current flowing through the line terminal; a zero-crossing detection circuit configured to measure a voltage and a frequency at the line terminal; as well as A microcontroller comprising an electronic processor, the microcontroller being configured to: applying a digital filter to a line current measurement signal received from the broadband current sensor; estimating a spectral density of the filtered line current measurement signal; calculating arc characteristics based on the spectral density; determining whether an arc fault exists within the circuit interrupting device based on a comparison of the arc characteristic to one or more threshold values; as well as The interrupting device is activated when an arc fault exists.

2. The circuit interrupting device according to claim 1, wherein: The threshold is determined by a machine learning model stored in a memory of the microcontroller.

3. The circuit interrupting device according to claim 1, wherein: The characteristic is one selected from the group consisting of: amplitude, standard deviation characteristic, cepstral coefficient characteristic, and frequency bin amplitude ratio.

4. The circuit interrupting device according to claim 1, wherein: The microcontroller is configured to estimate the spectral density by applying a short-time Fourier transform to the filtered line current measurement signal.

5. The circuit interrupting device of claim 1, wherein: The microcontroller is also configured to synchronize an analog-to-digital conversion (ADC) rate of the line current measurement signal with a frequency measured by the zero-crossing detection circuit.

6. The circuit interrupting device of claim 1, further comprising: a first printed circuit board configured to support the microcontroller; as well as A second printed circuit board is configured to support the broadband current sensor and a second microcontroller.

7. The circuit interrupting device of claim 1, further comprising: A first printed circuit board is configured to support the microcontroller and the broadband current sensor.

8. The circuit interrupting device of claim 1, wherein: The broadband current sensor is a Rogowski coil embedded in or connected to the printed circuit board of the circuit interruption device.

9. The circuit interrupting device of claim 1, wherein: The microcontroller is further configured to calculate a probability of an arc fault occurring based on the arc characteristics.

10. The circuit interrupting device of claim 1, wherein: The controller is further configured to calculate a standard deviation of the magnitude of the spectral density, wherein the standard deviation indicates fluctuations in power.

11. A method of detecting the presence of an arc fault occurring in an electrical circuit including a line terminal, the method comprising: measuring the current flowing through the line terminals with a broadband current sensor; applying, by a microcontroller including an electronic processor, a digital filter to a line current measurement signal received from the broadband current sensor; estimating, by the microcontroller, a spectral density of the filtered line current measurement signal; calculating, by the microcontroller, arc characteristics based on the spectral density; determining, by the microcontroller, whether an arc fault exists within the circuit based on a comparison of the arc characteristic with one or more threshold values; as well as An interrupt device is activated by the microcontroller when an arc fault exists.

12. The method according to claim 11, further comprising: The probability of an arc fault occurring is determined by the microcontroller.

13. The method according to claim 11, wherein: The characteristic is one selected from the group consisting of: amplitude, standard deviation characteristic, cepstral coefficient characteristic, and frequency bin amplitude ratio.

14. The method according to claim 11, wherein: Estimating the spectral density of the filtered line current measurement signal includes applying a short-time Fourier transform to the filtered line current measurement signal.

15. The method according to claim 11, further comprising: An analog-to-digital conversion (ADC) rate of the line current measurement signal is synchronized by the microcontroller with a frequency measured by a zero-crossing detection circuit included in the circuit interruption device.

16. The method according to claim 11, wherein: The microcontroller is supported by a first printed circuit board and the broadband current sensor is supported by a second printed circuit board.

17. The method according to claim 11, wherein: The microcontroller and the broadband current sensor are supported by a first printed circuit board.

18. The method according to claim 11, wherein: The broadband current sensor is a Rogowski coil.

19. The method according to claim 11, further comprising: A machine learning model configured to estimate a probability of an arc fault occurring based on the spectral density is executed by the microcontroller.

20. The method according to claim 11, further comprising: A standard deviation of the magnitude of the spectral density is calculated, wherein the standard deviation indicates fluctuations in power.

21. A system, comprising: A circuit interrupting device, the circuit interrupting device comprising: Line terminals; a broadband current sensor configured to measure current flowing through the line terminals; and communications circuitry configured to wirelessly transmit a signal including current measurements taken by the broadband current sensor; and An external device, the external device comprising: a second communication circuit configured to wirelessly receive signals; and A controller comprising an electronic processor, the controller being configured to: estimating a spectral density of the current measurements; calculating at least one arc characteristic associated with the spectral density; developing a machine learning model based on the spectral density and the at least one arc characteristic; and The machine learning module is deployed to the circuit interruption device.

22. A method of deploying a machine learning model to a circuit interruption device, the method comprising: receiving, by the communication circuit, a first set of line current measurements taken when an arc fault is present; receiving, by the communication circuit, a second set of line current measurements taken when an arc fault is not present; estimating, by a controller including an electronic processor, a first spectral density of the first set of line current measurements; estimating, by the controller, a second spectral density of a second set of line current measurements; calculating, by the controller, a first set of arc characteristics associated with the first spectral density; calculating, by the controller, a second set of non-arcing features associated with the second spectral density; executing, by the controller, a training algorithm to create the machine learning model based on the first set of arcing features and the second set of non-arcing features; and The machine learning model is deployed to the circuit interrupting device.

23. A circuit interrupting device, the circuit interrupting device comprising: Line terminals; a broadband current sensor configured to measure a current flowing through the line terminal; A microcontroller comprising an electronic processor, the microcontroller being configured to: estimating a spectral density of line current measurements taken by the broadband current sensor; calculating arc characteristics based on the spectral density; Calculating the probability of an arc fault occurring based on the arc characteristics and the spectral density by executing an inference algorithm; setting a first flag when the probability exceeds a threshold for a first amount of time; as well as An interrupt device is activated when the first flag is set.

24. A method of detecting the presence of an arc fault occurring in an electrical circuit including a line terminal, the method comprising: measuring the current flowing through the line terminals with a broadband current sensor; estimating, by the microcontroller, a spectral density of a line current measurement signal generated by the broadband current sensor; calculating, by the microcontroller, arc characteristics based on spectral density; Calculating the probability of an arc fault occurring based on the arc characteristics and the spectral density by executing an inference algorithm; setting a first flag when the probability exceeds a threshold for a first amount of time; as well as An interrupt device is activated when the first flag is set.