A fault arc detection method based on RLC arc model and convolutional neural network

Abstract

The application relates to the technical field of electrical fault detection, and discloses a fault arc detection method based on an RLC arc model and a convolutional neural network, which effectively suppresses strong background noise such as BMS switch noise and electromagnetic interference and significantly improves the capture precision of weak high-frequency oscillation signals by adopting a differential high-frequency current transformer to collect arc current signals and extract high-frequency components; meanwhile, the RLC arc model is used to generate simulation data for training, thereby breaking through the bottleneck of the scarcity of real fault samples, enabling the one-dimensional convolutional neural network model to have excellent generalization ability, accurately identifying unknown scenes such as new loads and weak arcs, avoiding problems such as poor adaptability and threshold rigidity caused by the dependence of traditional methods on artificial design features, realizing fast reasoning while ensuring detection precision, effectively solving the contradiction between real-time performance and precision, and providing a highly reliable and strongly adaptive technical scheme for fault arc detection of a lithium battery string system.

Description

Technical Field

[0001] This invention relates to the field of electrical fault detection technology, and in particular to a fault arc detection method based on the RLC arc model and convolutional neural network. Background Technology

[0002] In lithium-ion battery applications, fault arcing in series operation exhibits significant characteristics: on the one hand, the probability of fault arcing is relatively low; however, once it occurs, its destructive power is extremely strong. Given this characteristic, obtaining comprehensive sample data covering all operating conditions of lithium-ion batteries through numerous real-world experiments presents enormous difficulties and challenges. Due to limitations in actual experimental conditions and the randomness of fault arcing occurrence, it is difficult to collect sufficiently diverse and representative samples within a limited number of experiments, which greatly restricts research based on real-world experimental data.

[0003] Traditional arc detection models are primarily trained on real-world data. However, these models exhibit significant limitations when faced with unfamiliar scenarios such as novel loads and weak arcs. Their generalization performance is poor, making it difficult to accurately adapt to new operating conditions and environmental factors. In practical applications, they are highly prone to false alarms or missed detections. Missed detections can lead to faults going undetected and unhandled, potentially causing more serious safety incidents; false alarms, on the other hand, result in unnecessary resource waste and equipment downtime, impacting the normal operating efficiency of lithium battery systems.

[0004] Secondly, the signal characteristics of lithium battery systems pose significant challenges to fault detection. These systems exhibit various strong background noises, such as BMS switching noise and electromagnetic interference (EMI). These noises intertwine with the fault arc signal, severely impacting signal purity. Simultaneously, the high-frequency oscillation component of the DC arc has extremely weak amplitude and short duration, making it difficult for conventional time-domain / frequency-domain analysis methods to effectively distinguish fault features from noise components when processing such signals. The inability to accurately separate fault features leads to a significant reduction in feature extraction accuracy, consequently affecting the accuracy and reliability of subsequent fault detection.

[0005] In addition, existing arc detection methods have significant shortcomings in terms of scenario adaptability. Currently, most arc detection methods are designed for AC power systems, relying on the unique power frequency cycle and load type of AC power. Lithium-ion battery systems, however, use DC power, which differs significantly from AC power systems in signal characteristics and operating rules. Therefore, existing AC arc detection methods cannot be directly applied to DC scenarios involving lithium-ion batteries. Even some detection methods specifically designed for lithium-ion batteries require manual feature design, a process highly dependent on expert experience and knowledge. Differences in the experience and understanding of different experts result in features lacking universality and failing to adapt to various lithium-ion battery application scenarios. Furthermore, these methods lack flexibility in threshold setting, making it difficult to adaptively adjust according to actual conditions, further limiting their effectiveness in practical applications.

[0006] Finally, the trade-off between real-time performance and detection accuracy is particularly prominent in lithium-ion battery arc fault detection. While high-precision deep learning models can provide more accurate detection results, these models typically have high computational complexity. When deployed on embedded or edge devices, the limited computing resources of these devices significantly slow down the model's inference speed, failing to meet the stringent real-time requirements of lithium-ion battery thermal runaway early warning. Detection delays can lead to missed opportunities for optimal fault handling, resulting in accidents. Conversely, using lightweight models to improve inference speed can meet real-time requirements to some extent, but it sacrifices the model's feature extraction capabilities, leading to decreased detection accuracy and an inability to accurately identify arc faults, similarly posing a threat to the safe operation of lithium-ion battery systems.

[0007] Therefore, improvements to existing technologies are necessary.

[0008] The above information is provided as background information only to aid in understanding the present invention, and does not constitute an assertion or admission that any of the above content can be used as prior art relative to the present invention. Summary of the Invention

[0009] This invention provides a fault arc detection method based on the RLC arc model and convolutional neural network to solve the problems existing in the prior art.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] In a first aspect, the present invention provides a fault arc detection method based on an RLC arc model and a convolutional neural network, the method comprising:

[0012] A differential high-frequency current transformer is used to acquire the arc current signal, and the high-frequency component in the arc current signal is extracted.

[0013] The high-frequency components are input into a one-dimensional convolutional neural network model to obtain the fault arc detection result; the one-dimensional convolutional neural network model is obtained by training with simulation data generated by the RLC arc model.

[0014] Furthermore, in the fault arc detection method based on the RLC arc model and convolutional neural network, the method further includes:

[0015] Construct a one-dimensional convolutional neural network model;

[0016] Simulation data is generated using an RLC arc model.

[0017] The simulation data is input into the one-dimensional convolutional neural network model, and the one-dimensional convolutional neural network model is trained to obtain the trained one-dimensional convolutional neural network model.

[0018] Furthermore, in the fault arc detection method based on the RLC arc model and convolutional neural network, the RLC arc model includes a high-resistance arc resistor. Low resistance arc resistance Arc inductance Inter-electrode capacitance ;

[0019] The high-resistance arc resistor With the first switch The switch connected in parallel to the main circuit after being connected in series The two ends are equivalent to the cases of arc extinction and reignition;

[0020] The low-resistance arc resistor Arc inductance With the second switch The switches connected in series and then in parallel in the main circuit The two ends are equivalent to the case of stable combustion of an electric arc;

[0021] The inter-electrode capacitor With the third switch The switch connected in parallel after being connected in series in the main circuit At both ends, the third switch is quickly switched on and off. This allows the arc current signal to contain high-frequency pulses.

[0022] Furthermore, in the fault arc detection method based on the RLC arc model and convolutional neural network, the main circuit also includes an AC power supply voltage. Equivalent resistance of power supply side lines Equivalent inductance of power supply side lines Equivalent capacitance of power supply side lines Load impedance Equivalent resistance of the load-side circuit Equivalent inductance of the load-side circuit Equivalent capacitance of the load-side circuit ;

[0023] The AC power supply voltage Equivalent inductance of power supply side lines Equivalent resistance of power supply side lines ,switch Equivalent resistance of the load-side circuit Equivalent inductance of the load-side circuit and load impedance They are connected in series to form a circuit;

[0024] The equivalent capacitance of the power supply side line One end is connected to the equivalent resistance of the power supply side line. With the switch Between the two ends, the other end is connected to the AC power supply voltage. With the load impedance between;

[0025] The equivalent capacitance of the load-side circuit One end is connected to the equivalent inductance of the load-side line. With the load impedance Between, the other end is connected to the load impedance. With the AC power supply voltage between.

[0026] Furthermore, in the fault arc detection method based on the RLC arc model and convolutional neural network, the one-dimensional convolutional neural network model includes an input layer, a hidden layer, and an output layer;

[0027] The input layer is used to receive a one-dimensional array composed of high-frequency components in the arc current signal acquired by the differential high-frequency current transformer.

[0028] The hidden layer includes a convolutional layer, a pooling layer, and a global average pooling layer, which are used for feature recognition to realize a complex nonlinear mapping from the input layer to the output layer and to fit the decision rules between the input data and the output data.

[0029] The output layer is used to output the fault arc detection results.

[0030] Furthermore, in the fault arc detection method based on the RLC arc model and convolutional neural network, each convolutional layer contains several convolutional kernels, the specific parameters of which are adaptively learned through the training process, and the output of the convolution calculation can be expressed as:

[0031] ;

[0032] Where X and Y are the input column vector and output matrix of the first convolutional layer, respectively; the input data needs to be normalized before entering the first convolutional layer, scaling the magnitude of all data to the range of [-1,1] to speed up the convergence of the network and avoid gradient vanishing or gradient explosion due to excessive numerical differences; w and b are the weight vector and bias vector, respectively; * is the discrete convolution operation; f is the nonlinear activation function, realizing the nonlinear mapping from input to output.

[0033] Furthermore, in the fault arc detection method based on the RLC arc model and convolutional neural network, the convolutional layer uses the rectified linear unit ReLU as the activation function, expressed as:

[0034] .

[0035] Furthermore, in the fault arc detection method based on the RLC arc model and convolutional neural network, the operation of the output layer is divided into two steps:

[0036] First, the one-dimensional column vector output by the global average pooling layer is mapped to two numerical values, which serve as the original output of the output layer. This mapping process depends on the weights and biases between the input neurons and each output neuron, and the calculation formula is as follows:

[0037] ;

[0038] Where y0 and y1 are the original outputs of the output layer; x is the input column vector of the input neuron; and Let b0 be the weight matrix between the input neuron and the two output neurons; b0 and b1 are the biases of the two output neurons, respectively.

[0039] Secondly, the Sigmoid function is used as the activation function to map the two values ​​mentioned above to the interval [0,1]. The probability is calculated using the following formula:

[0040] ;

[0041] Wherein, the mapping value P0 is the probability that the sample is in a normal state, labeled 0; P1 is the probability that the sample is in an arc state, labeled 1;

[0042] Next, the probability with the larger value is selected as the final fault arc detection result.

[0043] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the fault arc detection method based on the RLC arc model and convolutional neural network provided in the first aspect above.

[0044] Thirdly, the present invention provides a computer-readable storage medium having computer-executable instructions stored thereon, the computer-executable instructions being executed by a computer processor to implement the fault arc detection method based on the RLC arc model and convolutional neural network provided in the first aspect above.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] This invention provides a fault arc detection method based on an RLC arc model and a convolutional neural network. By using a differential high-frequency current transformer to collect arc current signals and extract high-frequency components, it effectively suppresses strong background noise such as BMS switching noise and electromagnetic interference, significantly improving the acquisition accuracy of weak high-frequency oscillation signals. Simultaneously, by using the RLC arc model to generate simulation data for training, it overcomes the bottleneck of scarce real fault samples, enabling the one-dimensional convolutional neural network model to possess excellent generalization ability. This allows it to accurately identify unknown scenarios such as novel loads and weak arcs, avoiding the problems of poor adaptability and rigid thresholds caused by traditional methods relying on manually designed features. While ensuring detection accuracy, it achieves rapid inference, effectively resolving the contradiction between real-time performance and accuracy. This provides a highly reliable and adaptable technical solution for fault arc detection in lithium battery series systems.

[0047] The present invention has other features and advantages, which will be apparent from or will be set forth in detail in the accompanying drawings and the following detailed description, which together serve to explain the particular principles of the invention. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart illustrating a fault arc detection method based on an RLC arc model and a convolutional neural network, provided in Embodiment 1 of the present invention.

[0050] Figure 2This is an experimental schematic diagram of the method for acquiring high-frequency components in arc current signals using a differential high-frequency current transformer, as provided in Embodiment 1 of the present invention.

[0051] Figure 3 This is an experimental result diagram provided in Embodiment 1 of the present invention;

[0052] Figure 4 This is a circuit topology diagram of the RLC arc model provided in Embodiment 1 of the present invention;

[0053] Figure 5 This is the equivalent circuit diagram of each stage provided in Embodiment 1 of the present invention;

[0054] Figure 6 This is an analysis diagram of the high-frequency path working process provided in Embodiment 1 of the present invention;

[0055] Figure 7 This is a simulation circuit diagram of the impedance-type arc model provided in Embodiment 1 of the present invention;

[0056] Figure 8 This is a schematic diagram of the arc control module provided in Embodiment 1 of the present invention;

[0057] Figure 9 These are high-frequency feature maps of the experimental and simulation results provided in Embodiment 1 of the present invention;

[0058] Figure 10 This is a schematic diagram of the structure of a computer device provided in Embodiment 2 of the present invention. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Example 1

[0061] Please refer to Figure 1 This is a flowchart illustrating a fault arc detection method based on an RLC arc model and a convolutional neural network, provided in Embodiment 1 of the present invention. This method can be implemented in software and / or hardware. Specifically, the method includes the following steps:

[0062] S101. A differential high-frequency current transformer is used to collect the arc current signal and extract the high-frequency component from the arc current signal.

[0063] It should be noted that this invention uses a differential high-frequency current transformer to collect the high-frequency component of the arc current signal and applies it to fault arc detection. The experimental principle is as follows: Figure 2 As shown. Because the two conductors carrying reverse current (live wire and neutral wire) weaken the magnetic flux in the iron core, the equivalent magnetic flux density in the iron core is much lower than the saturation magnetic flux density, significantly reducing the risk of low-frequency saturation; at the same time, the differential high-frequency current transformer has a higher gain for high-frequency signals than for low-frequency signals. Based on the International Electrotechnical Commission (IEC) standard 62606-2017, this invention customizes an arc generator composed of cylindrical copper electrodes and conical carbon electrodes. An air gap is generated by moving the electrodes to ignite the arc. Different types of loads are connected using load connectors, and the output voltage waveform is observed using an oscilloscope. The line current is collected using a 1Ω sampling resistor and compared with the output of the differential high-frequency current transformer, verifying that the differential high-frequency current transformer signal is more suitable for fault arc detection than a direct current signal.

[0064] With the same type of load, the output voltage waveform was observed using an oscilloscope, the line current was collected using a 1Ω sampling resistor, and compared with the output of a differential high-frequency current transformer. This verified that the signal from the differential high-frequency current transformer is more suitable for detecting series arc faults than the direct current signal.

[0065] Incandescent lamp, vacuum cleaner, and fluorescent lamp were connected to the experimental circuit respectively, and the experimental results are as follows. Figure 3 As shown. Due to the different characteristics of the three types of loads, Figure 3 The current waveforms under arc conditions in (a), (b), and (c) of section 3 exhibit different characteristics, which complicates arc fault detection. When more new loads are connected to the circuit, new current characteristics may emerge, causing the classification of established detection methods to fail. However, after processing by a differential high-frequency current transformer, its output under arc conditions consistently exhibits high-frequency pulses of varying amplitudes (low-frequency characteristics are filtered out). Under normal conditions, the outputs corresponding to various loads are close to zero, only slightly reduced due to unavoidable small-amplitude Gaussian white noise. Therefore, the high-frequency component of the arc current is a common characteristic of arc faults.

[0066] The generation mechanism of high-frequency components of electric arc current:

[0067] Under normal circumstances, the air between two separated conductors acts as an insulating medium. When the potential difference between the conductors exceeds a certain threshold, the air medium breaks down, and a high-temperature conductor forms in the air gap, resulting in a series arc fault. During the air medium breakdown process, thermoelectric emission and field emission cause a large number of free electrons from the cathode surface to enter the air gap, thereby igniting the arc. The mechanisms of these two electron emission methods are different. Increased current density at the contact point leads to high temperature, which in turn generates thermoelectric emission; simultaneously, the extremely small distance between the two conductors creates a strong electric field, which in turn induces field emission.

[0068] The strong electric field between the protrusions on the conductor surface and the arc column easily promotes the formation of new field emission points. However, the heat generated by the arc vaporizes the protrusions, causing the field emission points to disappear. The repeated disappearance and regeneration of emission points continuously provide a large number of electrons to the arc, maintaining its combustion. The change in the position of the old and new emission points causes the arc root to move. The continuously moving arc root provides a continuous flow of electrons to the arc column, generating a power frequency component in the arc current. Conversely, the formation and disappearance of emission points provide intermittent electron pulses to the arc column, generating a high frequency component in the arc current.

[0069] S102. Input the high-frequency components into a one-dimensional convolutional neural network model to obtain the fault arc detection result; the one-dimensional convolutional neural network model is obtained by training with simulation data generated by the RLC arc model.

[0070] In one embodiment of this invention, the method further includes:

[0071] Construct a one-dimensional convolutional neural network model;

[0072] Simulation data is generated using an RLC arc model.

[0073] The simulation data is input into the one-dimensional convolutional neural network model, and the one-dimensional convolutional neural network model is trained to obtain the trained one-dimensional convolutional neural network model.

[0074] In one embodiment of this example, such as Figure 4 As shown, the RLC arc model includes a high-resistance arc resistor. Low resistance arc resistance Arc inductance Inter-electrode capacitance ;

[0075] The high-resistance arc resistor With the first switch The switch connected in parallel to the main circuit after being connected in series The two ends are equivalent to the cases of arc extinction and reignition;

[0076] The low-resistance arc resistor Arc inductance With the second switch The switches connected in series and then in parallel in the main circuit The two ends are equivalent to the case of stable combustion of an electric arc;

[0077] The inter-electrode capacitor With the third switch The switch connected in parallel after being connected in series in the main circuit At both ends, the third switch is quickly switched on and off. This allows the arc current signal to contain high-frequency pulses.

[0078] In one embodiment of this invention, the main circuit further includes an AC power supply voltage. Equivalent resistance of power supply side lines Equivalent inductance of power supply side lines Equivalent capacitance of power supply side lines Load impedance Equivalent resistance of the load-side circuit Equivalent inductance of the load-side circuit Equivalent capacitance of the load-side circuit ;

[0079] The AC power supply voltage Equivalent inductance of power supply side lines Equivalent resistance of power supply side lines ,switch Equivalent resistance of the load-side circuit Equivalent inductance of the load-side circuit and load impedance They are connected in series to form a circuit;

[0080] The equivalent capacitance of the power supply side line One end is connected to the equivalent resistance of the power supply side line. With the switch Between the two ends, the other end is connected to the AC power supply voltage. With the load impedance between;

[0081] The equivalent capacitance of the load-side circuit One end is connected to the equivalent inductance of the load-side line. With the load impedance Between, the other end is connected to the load impedance. With the AC power supply voltage between.

[0082] It should be noted that, as Figure 4 As shown, through an ideal switch To achieve on / off control of the main circuit and the arc branch: AC power supply voltage; These are the equivalent parameters of the power supply side lines; These are the equivalent parameters of the load-side lines; The load impedance is given. The resistor-inductor branch is the low-frequency path for the arc current, while the capacitor branch is the high-frequency path for the arc current.

[0083] When the circuit is working normally, the switch closure, When disconnected, the line current is primarily determined by the load impedance. In the event of an arc fault, the switch... The circuit remains disconnected, and the arc begins to periodically extinguish and reignite. The specific process is as follows: 1) High-frequency path Fast on / off ( Figure 5 (a) Figure 5 (b) of the above means that the arc current contains high-frequency pulses; 2) When the arc is extinguished and reignited, the arc resistance is extremely large and the arc inductance is negligible, which is equivalent to a switch. disconnect, closure( Figure 5 (c)); 3) When the electric arc is burning stably, the arc resistance decreases, which is equivalent to a switch. disconnect, closure( Figure 5 (d)). Among them, the on / off state of the high-frequency path is random, while the state change of the low-frequency path depends on the combustion state of the electric arc.

[0084] The operation of a high-frequency path can be equivalent to the zero-input response of a second-order circuit. Figure 6 In (a)), the operational circuit after its Laplace transform is as follows: Figure 6 As shown in (b) above. Assume the instantaneous voltage on the power supply side is... The instantaneous voltage on the load side is When the voltage difference between the two is sufficient to cause field-induced breakdown at the high-frequency current emission point, and Energy exchange will occur, triggering a rapid oscillation process of the arc current.

[0085] The following equation can be obtained using the loop current method:

[0086] ……(1);

[0087] Further derivation yields the frequency domain solution for the current:

[0088] ……(2);

[0089] in, ;

[0090] In reality, high-frequency arc current pulses exhibit damped oscillations, and the circuit is in an underdamped state. Therefore, the time-domain solution of the current can be expressed as:

[0091] ...(3);

[0092] The high-frequency characteristics of the arc current are related to the arc voltage, arc parameters, and line parameters, and mainly depend on the high-frequency charging and discharging process of the ground capacitance through the arc capacitance and line impedance.

[0093] Simulation verification of the electric arc model:

[0094] To verify the effectiveness of the proposed arc model, a detailed simulation model was built in SIMULINK. Figure 7 , Figure 8 ). Figure 7 The circuit is divided into low-frequency and high-frequency paths, used to generate the low-frequency and high-frequency components of the arc current, respectively. The arc controller and... Figure 8 The correlation (a) in the diagram is used to determine whether the circuit is in a normal or arcing state: In the normal state, switch K is closed, K4 is open, the low-frequency path is short-circuited, and the high-frequency path is disconnected; when a fault occurs, switch K is open, K4 is closed, and both the low-frequency and high-frequency paths operate simultaneously. For the low-frequency path, the arc resistance R... a The resistance value is determined by the state of the electric arc ( Figure 8 (c)); For the high-frequency path, the corresponding switch K3 is randomly switched on and off at high speed. In the simulation, the random switching of K3 is achieved by setting the signal generation probability of the binary sequence generator. Figure 8 (b)). To improve the simulation effect, this invention adds Gaussian white noise to the simulation signal, and finally superimposes the high-frequency component, low-frequency component and white noise component to obtain the line current i.

[0095] This invention selects resistive loads with operating currents of 1A, 2A, and 5A to verify the effectiveness of the arc model. For ease of comparison, the simulated output current is converted into a voltage signal via a sensor. The simulation results and actual experimental results are as follows: Figure 9 As shown. Under normal conditions, the high-frequency signal is close to zero; under arc conditions, a large number of high-frequency spikes appear in the signal, and as the load resistance decreases, the high-frequency components of the arc decrease slightly. Figure 9 The simulation waveform in (b) retains Figure 9 The fault characteristics of the actual load waveform in (a) verify the effectiveness of the proposed arc model. By traversing the initial phase angle of the AC power source, the frequency of the Bernoulli random sequence, and the load type, sufficient high-frequency arc current data can be generated for training the neural network.

[0096] In one embodiment of this invention, the one-dimensional convolutional neural network model includes an input layer, a hidden layer, and an output layer;

[0097] The input layer is used to receive a one-dimensional array composed of high-frequency components in the arc current signal acquired by the differential high-frequency current transformer.

[0098] The hidden layer includes a convolutional layer, a pooling layer, and a global average pooling layer, which are used for feature recognition to realize a complex nonlinear mapping from the input layer to the output layer and to fit the decision rules between the input data and the output data.

[0099] The output layer is used to output the fault arc detection results.

[0100] In one embodiment of this invention, each convolutional layer contains several convolutional kernels, the specific parameters of which are adaptively learned through the training process, and the output of the convolution calculation can be expressed as:

[0101] ;

[0102] Where X and Y are the input column vector and output matrix of the first convolutional layer, respectively; the input data needs to be normalized before entering the first convolutional layer, scaling the magnitude of all data to the range of [-1,1] to speed up the convergence of the network and avoid gradient vanishing or gradient explosion due to excessive numerical differences; w and b are the weight vector and bias vector, respectively; * is the discrete convolution operation; f is the nonlinear activation function, realizing the nonlinear mapping from input to output.

[0103] It should be noted that convolutional layers use different convolutional kernels (filters) to act on the original sequence to extract sequence features, and the features extracted by different convolutional kernels are different.

[0104] In one embodiment of this invention, the convolutional layer uses Rectified Linear Unit (ReLU) as the activation function, expressed as:

[0105] .

[0106] It should be noted that pooling layers do not contain trainable parameters; they calculate the average or maximum value of a data segment based on a set stride, with max pooling being the most widely used. Pooling layers essentially purify the input data, simplifying the output information of convolutional layers while preserving signal characteristics. After processing by pooling layers, the data needs to be input into several fully connected layers for classification. However, the large number of parameters in fully connected layers can easily lead to excessive model complexity. This invention uses global average pooling layers instead of fully connected layers, significantly reducing the number of parameters in the model.

[0107] In one embodiment of this invention, the operation of the output layer is divided into two steps:

[0108] First, the one-dimensional column vector output by the global average pooling layer is mapped to two numerical values, which serve as the original output of the output layer. This mapping process depends on the weights and biases between the input neurons and each output neuron, and the calculation formula is as follows:

[0109] ;

[0110] Where y0 and y1 are the original outputs of the output layer; x is the input column vector of the input neuron; and Let b0 be the weight matrix between the input neuron and the two output neurons; b0 and b1 are the biases of the two output neurons, respectively.

[0111] Secondly, the Sigmoid function is used as the activation function to map the two values ​​mentioned above to the interval [0,1]. The probability is calculated using the following formula:

[0112] ;

[0113] Wherein, the mapping value P0 is the probability that the sample is in a normal state, labeled 0; P1 is the probability that the sample is in an arc state, labeled 1;

[0114] Next, the probability with the larger value is selected as the final fault arc detection result.

[0115] Although this invention uses terms such as differential high-frequency current transformer and high-frequency component frequently, the possibility of using other terms is not excluded. These terms are used merely for the convenience of describing and explaining the essence of this invention; interpreting them as any additional limitation would contradict the spirit of this invention.

[0116] Example 2

[0117] Figure 10 This is a schematic diagram of the structure of a computer device provided in Embodiment 2 of the present invention. Figure 10 A block diagram of an exemplary computer device 12 suitable for implementing embodiments of the present invention is shown. Figure 10 The computer device 12 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0118] like Figure 10 As shown, the computer device 12 is represented in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0119] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0120] Computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0121] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 10 Not shown; usually referred to as a "hard drive"). Although Figure 10 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0122] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of the present invention.

[0123] Computer device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with the computer device 12, and / or with any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 22. Furthermore, computer device 12 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 20. As shown, network adapter 20 communicates with other modules of computer device 12 via bus 18. It should be understood that, although... Figure 10 As not shown, it can be used in conjunction with computer device 12 with other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0124] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the fault arc detection method based on the RLC arc model and convolutional neural network provided in the embodiments of the present invention.

[0125] Example 3

[0126] Embodiment 3 of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the fault arc detection method based on the RLC arc model and convolutional neural network provided in all embodiments of the present invention.

[0127] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0128] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0129] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0130] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0131] Finally, it should be noted that although the above embodiments have been described in the description and drawings of this invention, this should not limit the scope of patent protection of this invention. Any technical solutions that are based on the essential concept of this invention, utilize the content described in the description and drawings of this invention to make equivalent structural or procedural substitutions or modifications, as well as the direct or indirect application of the technical solutions of the above embodiments to other related technical fields, are all included within the scope of patent protection of this invention.

Claims

1. A fault arc detection method based on an RLC arc model and a convolutional neural network, characterized in that, The method includes: A differential high-frequency current transformer is used to acquire the arc current signal, and the high-frequency component in the arc current signal is extracted. The high-frequency components are input into a one-dimensional convolutional neural network model to obtain the fault arc detection result; the one-dimensional convolutional neural network model is obtained by training with simulation data generated by the RLC arc model.

2. The fault arc detection method based on the RLC arc model and convolutional neural network according to claim 1, characterized in that, The method further includes: Construct a one-dimensional convolutional neural network model; Simulation data is generated using an RLC arc model. The simulation data is input into the one-dimensional convolutional neural network model, and the one-dimensional convolutional neural network model is trained to obtain the trained one-dimensional convolutional neural network model.

3. The fault arc detection method based on the RLC arc model and convolutional neural network according to claim 1, characterized in that, The RLC arc model includes a high-resistance arc resistor. Low resistance arc resistance Arc inductance Inter-electrode capacitance ; The high-resistance arc resistor With the first switch The switch connected in parallel to the main circuit after being connected in series The two ends are equivalent to the cases of arc extinction and reignition; The low-resistance arc resistor Arc inductance With the second switch The switches connected in series and then in parallel in the main circuit The two ends are equivalent to the case of stable combustion of an electric arc; The inter-electrode capacitor With the third switch The switch connected in parallel after being connected in series in the main circuit At both ends, the third switch is quickly switched on and off. This allows the arc current signal to contain high-frequency pulses.

4. The fault arc detection method based on the RLC arc model and convolutional neural network according to claim 3, characterized in that, The main circuit also includes an AC power supply voltage. Equivalent resistance of power supply side lines Equivalent inductance of power supply side lines Equivalent capacitance of power supply side lines Load impedance Equivalent resistance of the load-side circuit Equivalent inductance of the load-side circuit Equivalent capacitance of the load-side circuit ; The AC power supply voltage Equivalent inductance of power supply side lines Equivalent resistance of power supply side lines ,switch Equivalent resistance of the load-side circuit Equivalent inductance of the load-side circuit and load impedance They are connected in series to form a circuit; The equivalent capacitance of the power supply side line One end is connected to the equivalent resistance of the power supply side line. With the switch Between the two ends, the other end is connected to the AC power supply voltage. With the load impedance between; The equivalent capacitance of the load-side circuit One end is connected to the equivalent inductance of the load-side line. With the load impedance Between, the other end is connected to the load impedance. With the AC power supply voltage between.

5. The fault arc detection method based on the RLC arc model and convolutional neural network according to claim 1, characterized in that, The one-dimensional convolutional neural network model includes an input layer, a hidden layer, and an output layer; The input layer is used to receive a one-dimensional array composed of high-frequency components in the arc current signal acquired by the differential high-frequency current transformer. The hidden layer includes a convolutional layer, a pooling layer, and a global average pooling layer, which are used for feature recognition to realize a complex nonlinear mapping from the input layer to the output layer and to fit the decision rules between the input data and the output data. The output layer is used to output the fault arc detection results.

6. The fault arc detection method based on the RLC arc model and convolutional neural network according to claim 5, characterized in that, Each convolutional layer contains several convolutional kernels, the specific parameters of which are adaptively learned through the training process. The output of the convolution calculation can be expressed as: ； Where X and Y are the input column vector and output matrix of the first convolutional layer, respectively; the input data needs to be normalized before entering the first convolutional layer, scaling the magnitude of all data to the range of [-1,1] to speed up the convergence of the network and avoid gradient vanishing or gradient explosion due to excessive numerical differences; w and b are the weight vector and bias vector, respectively; * is the discrete convolution operation; f is the nonlinear activation function, realizing the nonlinear mapping from input to output.

7. The fault arc detection method based on the RLC arc model and convolutional neural network according to claim 6, characterized in that, The convolutional layer uses the rectified linear unit ReLU as the activation function, expressed as: 。 8. The fault arc detection method based on the RLC arc model and convolutional neural network according to claim 5, characterized in that, The operation of the output layer is divided into two steps: First, the one-dimensional column vector output by the global average pooling layer is mapped to two numerical values, which serve as the original output of the output layer. This mapping process depends on the weights and biases between the input neurons and each output neuron, and the calculation formula is as follows: ； Where y0 and y1 are the original outputs of the output layer; x is the input column vector of the input neuron; and Let b0 be the weight matrix between the input neuron and the two output neurons; b0 and b1 are the biases of the two output neurons, respectively. Secondly, the Sigmoid function is used as the activation function to map the two values ​​mentioned above to the interval [0,1]. The probability is calculated using the following formula: ； Wherein, the mapping value P0 is the probability that the sample is in a normal state, labeled 0; P1 is the probability that the sample is in an arc state, labeled 1; Next, the probability with the larger value is selected as the final fault arc detection result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the fault arc detection method based on the RLC arc model and convolutional neural network as described in any one of claims 1-8.

10. A computer-readable storage medium having computer-executable instructions stored thereon, characterized in that, The computer-executable instructions are executed by a computer processor to implement the fault arc detection method based on the RLC arc model and convolutional neural network as described in any one of claims 1-8.

Images