A Temporary Ground Wire Location Detection Method for Quantum Fuzzy Decomposition Graph Neural Networks

Abstract

The present invention proposes a method for locating and detecting temporary grounding wires of a quantum fuzzy decomposition graph neural network. This method combines deep graph neural networks, fuzzy logic, quantum computing, and empirical mode decomposition to calculate the positions of temporary grounding wires in distribution lines. First, the temporary grounding wire location detection device inputs a 200Hz three-phase sine signal into the line to be measured, samples the signal, and performs empirical mode decomposition on the sampled signal. Then, a fault graph training set is established to train the deep graph neural network, and a line node grounding confidence level matrix is predicted. It is input into the simulated grounding location quantum transmission line, fuzzified, and the correlation coefficient is calculated. Finally, the position of the grounding wire in the line is calculated. The proposed quantum fuzzy decomposition graph neural network method can solve the problem of the long time-consuming for manual search of grounding wires, realize the function of locating and detecting temporary grounding wires of the line, optimize the renovation and maintenance work of the distribution network, and improve the efficiency of grounding wire inspection in the power system.

Description

Technical Field

[0001] The present invention belongs to the fields of power systems and artificial intelligence, relates to a quantum fuzzy decomposition graph neural network, and is applicable to the positioning detection of temporary grounding wires in power systems. Background Art

[0002] Hanging a grounding wire is an essential step in the renovation and maintenance of the distribution network. It is necessary to remove the originally artificially set grounding wire after the maintenance work is completed and before the power is restored. In recent years, the Chinese power grid has been attaching importance to strengthening the quality education and technical training of power workers, and continuously improving the management and organizational systems, which has reduced the occurrence frequency of the accident of closing the switch with the grounding wire on the transmission line to a certain extent, but such accidents still cannot be completely eliminated. The development of the power grid urgently needs a technology to solve such incidents, so as to ensure the safe operation of the power system and the personal safety of power workers. The existing temporary grounding wire positioning detection methods, such as the grounding detection device composed of a series resonance amplification circuit and a transmitting antenna, this method detects and analyzes fault signals under abnormal operation of the distribution network, so as to realize the investigation of grounding faults. The disadvantage is that most grounding faults will cause the distribution line to trip, and valuable fault pulse signals cannot be detected by using the above method under the power-off state of the line.

[0003] Usually, the impedance ranging method is used to determine the distance of the temporary grounding wire connection point and then manually inspect and remove the grounding wire. Even when using a high-precision electrical signal acquisition chip, due to the inevitable impedance fluctuations of the long-distance transmission line and the influence of the residual waves in the line, the grounding position cannot be accurately located, and the operating personnel still spend most of the time in the process of manually searching for the grounding wire connection point, which is not conducive to the rapid removal of the grounding wire. This also causes losses of manpower and financial resources brought by the long-term power outage of the line and reduces the operation quality of the power grid.

[0004] Therefore, a temporary grounding wire positioning detection method based on a quantum fuzzy decomposition graph neural network is proposed to solve the problem of the long time-consuming for manually searching for the grounding wire. Summary of the Invention

[0005] The present invention proposes a temporary grounding wire positioning detection method based on a quantum fuzzy decomposition graph neural network, which is executed by a temporary grounding wire positioning detection device:

[0006] The temporary grounding wire positioning device uses a "12V rechargeable battery" to supply power to the "SPWM signal generation module" and the "ARM chip" respectively after boosting and bucking. The "SPWM signal generation module" is composed of an IPM chip with the model of PSS10S92E6-AG, and the model of the "ARM chip" is ARM920T;

[0007] To generate a strong enough current signal in the circuit under test, the Raspberry Pi generates a PWM signal to control the "SPWM signal generation module" to generate a three-phase sine signal. After passing through "LCL filtering" and "current-limiting resistors", a three-phase sine signal with a specific frequency of 200 Hz is output. The output signal is fed back to the "Raspberry Pi" in real time to adjust the PWM signal, avoiding the influence of residual waves in the circuit under test and having the ability to resist interference. The model of the "Raspberry Pi" used is 3B+;

[0008] The temporary grounding wire positioning detection device measures the voltage signal and current signal on the circuit under test. The voltage signal and current signal in the circuit under test are converted into weak voltage and current signals through "isolation transformers" and "anti-aliasing filtering", and then conditioned and amplified through a "signal amplification circuit", and finally input into an "ADC analog-to-digital converter" to obtain a sampling signal and transmit it to the "Raspberry Pi". The "signal amplification circuit" uses the integrated differential operational amplifier INA129, and the "ADC analog-to-digital converter" uses the AD7606 chip. The AD7606 chip is conveniently combined with the "Raspberry Pi" to achieve accurate signal acquisition;

[0009] The "Raspberry Pi" performs digital signal processing on the sampling signal and then transmits the data to the "ARM chip". The quantum fuzzy decomposition graph neural network method in the "ARM chip" accurately calculates the position of the grounding wire based on the data and finally outputs the position of the grounding wire to the display screen.

[0010] The quantum fuzzy decomposition graph neural network method combines deep graph neural networks, fuzzy logic, quantum computing, and empirical mode decomposition to calculate the position of the temporary grounding wire in the distribution line, with the advantages of high accuracy and less time consumption in grounding wire positioning; the steps in the use process are as follows:

[0011] Step S1-1: Number the nodes of the circuit under test. Let i represent the i-th node, where i ≤ n and n is the total number of nodes in the circuit under test;

[0012] Step S1-2: Select any node on the circuit under test to hang the grounding wire and record the serial number i of the node where the grounding wire is hung; e ;

[0013] Step S1-3: The "SPWM signal generation module" generates a 200 Hz three-phase sine signal and inputs it into the circuit under test. Measure the voltage signal at each node of the circuit under test, and obtain the sampling signal after ADC analog-to-digital conversion. Denote the sampling signal of the j-th phase voltage at the i-th node of the circuit under test as y i,j (t), where j = 1 represents the A-phase voltage, j = 2 represents the B-phase voltage, j = 3 represents the C-phase voltage, and t = 0.0001, 0.0002,..., 1 s, that is, 50 points are sampled in each signal period;

[0014] Step S1-2: Decompose the sampled signal y i,j (t) n1 times until a signal that meets the Hilbert condition is obtained That is, the first mode function spectral component v i,j,1 (t); The specific process of the k-th decomposition includes:

[0015] Use the cubic spline method to solve the maximum envelope line L i,j,k-1 (t) and the minimum envelope line L max,i,j,k (t) of the sampled signal y min,i,j,k (t); When k = 1, y i,j,k-1 (t) = y i,j (t);

[0016] Calculate the envelope mean value:

[0017]

[0018] Subtract the envelope mean value h i,j,k-1 (t) from y i,j,k (t) to obtain the intermediate signal y i,j,k (t), that is

[0019] y i,j,k (t) = y i,j,k-1 (t) - h i,j,k (t), t = 0.0001, 0.0002,..., 1s (2)

[0020] Step S1-3: Subtract the first mode function spectral component v i,j (t) from y i,j,1 (t) to obtain the signal d without high-frequency components i,j,1 (t), that is

[0021] d i,j,1 (t) = y i,j (t) - v i,j,1 (t), t = 0.0001, 0.0002,..., 1s (3)

[0022] Step S1-4: Decompose d i,j,1 (t) n2 times according to the decomposition process in Step S1-2 until the absolute value of the mode function spectral component is smaller than the preset value ε to end the decomposition. Therefore, finally, n2 mode function spectral components of the sampled signal y i,j (t) are obtained

[0023] Step S2-1: Extract the voltage peak value U of each mode function spectral component v i,j (t) of the voltage sampled signal y i,j,k (t)m,i,j,k and the moment t corresponding to the first voltage peak m,i,j,k ;

[0024] Step S2-1: Obtain a matrix V representing the voltages of all nodes in the line under test, i.e.:

[0025]

[0026] where U m,i,j,k is the voltage peak of the k-th mode function spectral component of the j-th phase voltage sampling signal at the i-th node of the line under test, and U m,1,1,1 is the voltage peak of the first mode function spectral component of the first phase voltage sampling signal at the first node of the line under test, and U m,1,1,2 is the voltage peak of the second mode function spectral component of the first phase voltage sampling signal at the first node of the line under test, and U m,1,j,k is the voltage peak of the k-th mode function spectral component of the j-th phase voltage sampling signal at the first node of the line under test, is the voltage peak of the (n2 - 1)-th mode function spectral component of the third phase voltage sampling signal at the first node of the line under test, is the voltage peak of the n2-th mode function spectral component of the third phase voltage sampling signal at the first node of the line under test, and U m,i,1,1 is the voltage peak of the first mode function spectral component of the first phase voltage sampling signal at the i-th node of the line under test, and U m,i,1,2 is the voltage peak of the second mode function spectral component of the first phase voltage sampling signal at the i-th node of the line under test, is the voltage peak of the (n2 - 1)-th mode function spectral component of the third phase voltage sampling signal at the i-th node of the line under test, is the voltage peak of the n2-th mode function spectral component of the third phase voltage sampling signal at the i-th node of the line under test, and U m,n,1,1 is the voltage peak of the first mode function spectral component of the first phase voltage sampling signal at the n-th node of the line under test, and U m,n,1,2 is the voltage peak of the second mode function spectral component of the first phase voltage sampling signal at the n-th node of the line under test, and U m,n,j,k is the voltage peak of the k-th mode function spectral component of the j-th phase voltage sampling signal at the n-th node of the line under test, is the voltage peak of the (n2 - 1)-th mode function spectral component of the third phase voltage sampling signal at the n-th node of the line under test, is the voltage peak of the n2-th mode function spectral component of the third phase voltage sampling signal at the n-th node of the line under test;

[0027] Step S2-2: Traverse all nodes of the circuit under test and their two-dimensional connection relationships to obtain the topological matrix A of the circuit under test, that is:

[0028]

[0029] When there is a direct branch between node i1 and node i2,

[0030] When there is no direct branch between node i1 and node i2, where a 1,1 is the element in the first row and first column, a 1,2 is the element in the first row and second column, is the element in the first row and third column, a 1,n is the element in the first row and, a 2,1 , a 2,2 is the element in the second row and second column, is the element in the second row and i2 column, a 2,n is the element in the second row and n column, is the element in the i1 row and first column, is the element in the i1 row and second column, is the element in the i1 row and i2 column, is the element in the i1 row and n column, a n,1 is the element in the n row and first column, a n,2 is the element in the n row and second column, a n,3 is the element in the n row and third column, a n,n is the element in the n row and n column;

[0031] Step S2-3: Repeat sampling N R times to establish a fault graph training set; The training set consists of the topological matrix A, N R node voltage matrices V of the circuit under test, and N R node numbers i at the grounding wire connection points e ; The topological matrix A and N R node voltage matrices V of the circuit under test are used as the input of the deep graph neural network; N R node numbers i at the grounding wire connection points e are used as the output of the deep graph neural network; Train the deep graph neural network model;

[0032] The deep graph neural network consists of 3 convolutional layers and 1 fully connected layer;

[0033] The output of the convolutional layer is shown in the following formula:

[0034]

[0035] A′ = A + I (7)

[0036] Wherein, H l and A are the inputs of the l-th convolutional layer; H l+1 is the output of the l-th convolutional layer; when l = 1, H l is the node voltage matrix V of the line to be measured; I is an identity matrix of size n×n; A′ is the adjacency matrix after adding the identity matrix to A; D′ is the degree matrix; W l is the parameter matrix of dimension 3n2×3n2 to be trained for the l-th convolutional layer; σ is the convolution kernel function;

[0037] The degree matrix D′ is a diagonal matrix, and its main diagonal elements are the sum of the elements in each row of the matrix A′, that is:

[0038]

[0039] Wherein, A ij ′ is the element in the i-th row and j-th column of the adjacency matrix A′;

[0040] The output of the fully connected layer is shown as follows:

[0041] Y FC = σ(W FC X FC b) (9)

[0042] Wherein, X FC is the input of the fully connected layer, that is, the output H4 of the 3rd convolutional layer; Y FC is the output of the fully connected layer; W FC is the parameter matrix to be trained for the fully connected layer, with a dimension of 1×n; b is a preset weight matrix, with a dimension of 3n2×1;

[0043] Step S3-1: According to the methods in steps S1 and S2, perform N c times of sampling on the line to be measured to obtain N c node voltage matrices V of the line to be measured t , use the trained deep graph neural network to predict the nodes at the grounding wire connection points of the line to be measured. After all the prediction results are generated, count the number of times each node appears as a prediction result, and N i is the number of times the i-th node appears as a prediction result;

[0044] Step S3-2: Calculate the probability level a i that the i-th node has a grounding wire, that is:

[0045]

[0046] Obtain the grounding confidence level matrix T FT , that is, represent the probability level that each node of the line has a grounding wire:

[0047] T FT = (a1,a2,…,a i ,…,a n ) T (11)

[0048] In the formula, (a1,a2,…,a i ,…,a n ) T represents the transpose of the vector (a1,a2,…,a i ,…,a n );

[0049] Step S4-1: Select the number of qubits n q to satisfy the condition The confidence level matrix T of the nodes of the circuit to be measured grounded FT After quantum state encoding, an initial quantum state vector with the number of bits n1 is obtained, that is

[0050]

[0051] In the formula, |i〉 is the ground state represented in decimal, and ρ is the normalization constant;

[0052] Step S4-2: Input the initial quantum state vector into the simulated grounded positioning quantum circuit:

[0053] The simulated grounded positioning quantum circuit is designed to obtain a quantum state grounding range matrix R with the dimension of n×1 FT ; The simulated grounded positioning quantum circuit consists of quantum gates Y gate, Z gate and measurement devices M1,M2,…,M n-1 ,M n ; The matrix forms of the Y gate and the Z gate are respectively:

[0054]

[0055]

[0056] In the formula, θ is the preset quantum rotation angle;

[0057] The measurement devices M1,M2,…,M n-1 ,M n Measure the entangled quantum state in the Z-axis direction to obtain the probability p i of the quantum ground state |i>, that is:

[0058]

[0059] In the formula, Y + (θ) and Z +$(\theta)$ is the unitary matrix of $Y(\theta)$ and $Z(\theta)$ respectively; is the initial vector of the quantum state The initial vector in the opposite direction; $\langle i|$ is the quantum ground state in the opposite direction of $|i\rangle$;

[0060] Obtain the quantum state grounding range matrix $R$ FT , that is:

[0061]

[0062] Step S5-1: Fuzzify $p$ i :

[0063] Describe $p$ according to the fuzzy subsets $\{SS, S, M, B, BB\}$ i where $SS$ means very little grounding possibility, $S$ means little grounding possibility, $M$ means medium grounding possibility, $B$ means high grounding possibility, and $BB$ means very high grounding possibility;

[0064] Calculate the membership degree $\mu$ i of $p$ SS belonging to $SS$: i

[0065]

[0066] Calculate the membership degree $\mu$ i of $p$ S belonging to $S$: i

[0067]

[0068] Calculate the membership degree $\mu$ i of $p$ M belonging to $M$: i

[0069]

[0070] Calculate the membership degree $\mu$ i of $p$ B belonging to $B$: i

[0071]

[0072] Calculate the membership degree $\mu$ i of $p$ BB belonging to $BB$: i

[0073]

[0074] Step S5-2: Determine the grounding location correlation coefficient $K$ of the $i$-th nodei :

[0075] K i = 0.5 μ SS (p i ) + 0.75 μ S (p i ) + μ M (p i ) + 1.25 μ B (p i ) + 1.5 μ BB (p i ) i = 1, 2, 3, …, n; (22)

[0076] Step S6-1: Extract the maximum likelihood node i of the quantum state grounding range matrix P , which is the node of the line under test corresponding to the maximum value of p i , that is:

[0077]

[0078] Step S6-2: According to the method in Step S1, sample the current signal successively on all direct branches connected to the maximum likelihood node i P , and obtain the peak values I m,g,j,k of each harmonic component of the sampled current and the corresponding time t g,j,k of the first peak, where g represents the g-th branch connected to the maximum likelihood node, g ≤ n3, and n3 is the number of direct branches connected to the maximum likelihood node i P ;

[0079] Step S6-3: Calculate the distance from the grounding wire position on the g-th direct branch to the maximum likelihood node i P , that is:

[0080]

[0081] In the formula, and are respectively the voltage peak value of the k-th mode function spectral component of the j-th phase voltage sampling signal at the maximum likelihood node i P and the time corresponding to the first peak;

[0082] Step S6-4: Output the maximum likelihood node i P and the distances P from the grounding wire positions on a total of n3 direct branches to the maximum likelihood node i , that is, n3 results of the grounding wire hanging positions in the line under test.

[0083] The present invention has the following advantages and effects compared with the prior art:

[0084] (1) The present invention designs a temporary grounding wire positioning and detection device. After inputting a 200Hz three-phase sine signal into the line to be measured, signal sampling is performed at each node, avoiding the influence of residual waves remaining in the line to be measured and having anti-interference performance.

[0085] (2) Based on the prediction results of the depth graph neural network, the present invention uses fuzzy logic to perform non-linear correlation calculations, realizing the complementary advantages of linear analysis and non-linear analysis and improving the accuracy of grounding wire positioning and detection.

[0086] (3) The present invention uses quantum computing to globally calculate the grounding probability levels of all nodes in the line to be measured, accelerating the process of temporary grounding wire positioning and detection of the line, reducing the time-consuming for finding the grounding wire, and optimizing the work of distribution network transformation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Figure 1 is the hardware framework diagram of the temporary grounding wire positioning and detection device of the method of the present invention.

[0088] Figure 2 is the flow chart of the quantum fuzzy decomposition graph neural network method of the method of the present invention.

[0089] Figure 3 is the schematic diagram of the simulated grounding positioning quantum circuit of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0090] A temporary grounding wire positioning and detection method based on a quantum fuzzy decomposition graph neural network proposed by the present invention is described in detail as follows in conjunction with the accompanying drawings:

[0091] Figure 1 is the hardware framework diagram of the temporary grounding wire positioning and detection device of the method of the present invention.

[0092] The detection device uses a "12V rechargeable battery" to supply power to the "SPWM signal generation module" and the "ARM chip" respectively after boosting and bucking. The "SPWM signal generation module" is composed of an IPM chip of model PSS10S92E6-AG, and the model of the "ARM chip" is ARM920T;

[0093] To form a sufficiently strong current signal in the line to be measured, the Raspberry Pi generates a PWM signal to control the "SPWM signal generation module" to generate a three-phase sine signal. After passing through "LCL filtering" and "current limiting resistors", a three-phase sine signal of 200Hz with a specific frequency is output. The output signal is fed back to the "Raspberry Pi" in real time to adjust the PWM signal, avoiding the influence of residual waves remaining in the line to be measured and having anti-interference ability. The model of the "Raspberry Pi" used is 3B+;

[0094] The device measures the voltage signal and current signal on the line to be measured, converts the voltage signal and current signal in the line to be measured into weak voltage signals and weak current signals through an "isolation transformer" and "anti-aliasing filtering", then adjusts and amplifies them through a "signal amplification circuit", and finally inputs them into an "ADC analog-to-digital converter" to obtain sampling signals and transmit them to a "Raspberry Pi". The "signal amplification circuit" uses an integrated differential operational amplifier INA129, and the "ADC analog-to-digital converter" uses an AD7606 chip. The AD7606 chip is conveniently combined with the "Raspberry Pi" to achieve precise signal acquisition;

[0095] The "Raspberry Pi" performs digital signal processing on the sampling signals and then transmits the data to an "ARM chip". The quantum fuzzy decomposition graph neural network method in the "ARM chip" accurately calculates the position of the grounding wire according to the data and finally outputs the position of the grounding wire to the display screen.

[0096] Figure 2 It is the flow chart of the quantum fuzzy decomposition graph neural network method of the method of the present invention.

[0097] The following are the specific steps of the quantum fuzzy decomposition graph neural network method of the method of the present invention:

[0098] S1: Input a 200Hz three-phase sine signal to the line to be measured, perform signal sampling, and perform empirical mode decomposition on the sampling signals;

[0099] S2: Establish a fault graph training set and train a deep graph neural network;

[0100] S3: Use the deep graph neural network for prediction to obtain a line node grounding confidence level matrix;

[0101] S4: Construct a quantum state grounding range matrix through a simulated grounding positioning quantum transmission line;

[0102] S5: Perform fuzzification processing and calculate the correlation coefficient;

[0103] S6: Calculate the position of the grounding wire on the line.

[0104] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied to other related technical fields, shall be similarly included in the patent protection scope of the present invention.

Claims

1. A method for detecting the location of a temporary grounding wire of a quantum fuzzy decomposition graph neural network, characterized in that, Performed by the temporary grounding wire positioning detection device: The Raspberry Pi generates a PWM signal to control the "SPWM signal generation module" to generate three-phase sine signals. After passing through "LCL filtering" and "current-limiting resistors", a three-phase sine signal with a specific frequency of 200 Hz is output, and the output signal is fed back to the "Raspberry Pi" in real time to adjust the PWM signal; The voltage signal and current signal in the line to be measured are converted into weak voltage and current signals through "isolation transformers" and "anti-aliasing filtering", and then conditioned and amplified through a "signal amplification circuit" and input into an "ADC analog-to-digital converter" to obtain sampling signals and transmit them to the "Raspberry Pi"; The "Raspberry Pi" processes the sampling signals through digital signal processing and then transmits the data to the "ARM chip". The quantum fuzzy decomposition graph neural network method in the "ARM chip" accurately calculates the position of the grounding wire based on the data; The quantum fuzzy decomposition graph neural network method combines deep graph neural networks, fuzzy logic, quantum computing, and empirical mode decomposition to calculate the position of the temporary grounding wire in the distribution line; The steps during use are as follows: Step S1: Input a 200 Hz three-phase sine signal into the line to be measured, perform signal sampling, and perform empirical mode decomposition on the sampling signals; Step S2: Establish a fault graph training set and train the deep graph neural network; Step S3-1: According to the methods in Steps S1 and S2, sample the line to be measured times to obtain node voltage matrices of the line to be measured . Using the trained deep graph neural network, predict the nodes at the grounding wire connection points of the line to be measured. After all the prediction results are generated, count the number of times each node appears as a prediction result, is the number of times the th node appears as a prediction result; Step S3-2: Calculate the probability level that the th node has a grounding wire : , obtain a grounding confidence level matrix , that is, to characterize the probability level of the existence of grounding wires at each node of the line: In the formula, represents the transpose of the vector . Step S4-1: Select the number of qubits to meet the condition , and ground the confidence level matrix of the circuit nodes to be measured . After quantum state encoding, the initial quantum state vector with the number of bits is obtained : wherein, is the ground state in decimal representation, is the normalization constant; Step S4-2: Input the initial quantum state vector into the simulated ground positioning quantum circuit: The simulated ground positioning quantum circuit aims to obtain a quantum state ground range matrix with a dimension of ; The simulated ground positioning quantum circuit consists of quantum gates , gates, gates and a measurement device . The matrix forms of the Y gate and the Z gate are respectively: In the formula, is a preset quantum rotation angle; Measuring device Measure the entangled quantum state in the Z-axis direction to obtain the probability of the quantum probability , that is: In the formula, and are unitary matrices of and respectively; is the initial vector of the quantum state in the opposite direction; is in the opposite direction of the quantum ground state; Obtain the quantum state ground range matrix : Step S5-1: Fuzzify : According to the fuzzy subset {SS, S, M, B, BB}, it is described that SS indicates a very low possibility of grounding, S indicates a low possibility of grounding, M indicates a medium level of grounding possibility, B indicates a high possibility of grounding, and BB indicates a very high possibility of grounding; Calculation Membership degree belonging to SS : Calculation Membership degree belonging to S : Calculation Membership degree belonging to M : Calculation Membership degree belonging to B : Calculation Membership degree belonging to BB : Step S5-2: Determine the grounding positioning correlation coefficient of the th node : ； Step S6-1: Extract the maximum likelihood node of the quantum state ground range matrix is the node of the circuit under test corresponding to the maximum value : Step S6-2: According to the method in Step S1, sample the current signal successively on all the direct branches connected to the maximum likelihood node and obtain the peak values of the harmonic components of the sampled current and the moment corresponding to the first peak value , denotes the th branch connected to the maximum likelihood node, , is the maximum likelihood node the number of direct branches connected to; Step S6-3: Calculate the distance from the grounding wire position on the th direct branch to the maximum likelihood node :​ ， Wherein, and are respectively the voltage peak value of the -th modal function spectral component of the -phase voltage sampling signal at the maximum likelihood node and the time corresponding to the first peak value; Step S6-4: Output the maximum likelihood node and the total distance from the grounding wire position on each of the direct branches to the maximum likelihood node , , ……, , that is, the possible results of the grounding wire positions in the line under test.

2. The temporary grounding wire positioning detection method of a quantum fuzzy decomposition graph neural network according to claim 1, wherein The specific steps of inputting a 200 Hz three-phase sine signal into the line to be measured, performing signal sampling, and performing empirical mode decomposition in Step S1 are as follows: Step S1-1: Number the nodes of the circuit under test, indicating the th node, , where is the total number of nodes of the circuit under test; Step S1-2: Hang a grounding wire at any node of the circuit under test, and record the serial number of the node where the grounding wire is hung ; Step S1-3: The "SPWM signal generation module" generates a three-phase sine signal of 200 Hz and inputs it to the circuit under test. The voltage signal is measured at each node of the circuit under test, and the sampling signal is obtained after ADC analog-to-digital conversion. Denote the sampling signal of the th node of the th phase voltage of the circuit under test as , represents the voltage of phase A, represents the voltage of phase B, represents the voltage of phase C, , that is, 50 points are sampled in each signal period; Step S1-4: For the sampled signal Perform times of decomposition until a signal that meets the Hilbert condition is obtained, that is, the first mode function spectral component ; The specific process of the th decomposition includes: The sampling signal is solved by using the cubic spline method maximum envelope and minimum envelope ; when then ; Calculate the envelope mean: ， Subtract the envelope mean value from it to obtain an intermediate signal , that is ， Step S1-5: Subtract from the first modal function spectral component to obtain a high-frequency component-free , that is ， Step S1-6: For After the decomposition process in Step S1-4 for times of decomposition until the absolute value of the modal function spectral component is smaller than the preset value to end the decomposition. Therefore, finally, the modal function spectral components of the sampling signal are obtained.

3. The temporary grounding wire positioning detection method of a quantum fuzzy decomposition graph neural network according to claim 1, characterized in that The establishment of the fault diagram training set in step S2 and the training of the depth graph neural network are specifically as follows: Step S2-1: Extract the spectral components of each modal function of the voltage sampling signal of the voltage peak and the moment corresponding to the first voltage peak ; Step S2-1: Obtain a matrix representing the voltages of all nodes of the circuit under test, i.e.: , namely: Wherein, is the voltage peak value of the -th modal function spectral component of the -phase voltage sampling signal at the -th node of the line to be measured, is the voltage peak value of the first modal function spectral component of the first-phase voltage sampling signal at the first node of the line to be measured, is the voltage peak value of the second modal function spectral component of the first-phase voltage sampling signal at the first node of the line to be measured, is the voltage peak value of the -th modal function spectral component of the -phase voltage sampling signal at the first node of the line to be measured, is the voltage peak value of the -th modal function spectral component of the third-phase voltage sampling signal at the first node of the line to be measured, is the voltage peak value of the -th modal function spectral component of the third-phase voltage sampling signal at the first node of the line to be measured, is the voltage peak value of the first modal function spectral component of the first-phase voltage sampling signal at the -th node of the line to be measured, is the voltage peak value of the second modal function spectral component of the first-phase voltage sampling signal at the -th node of the line to be measured, is the voltage peak value of the -th modal function spectral component of the third-phase voltage sampling signal at the -th node of the line to be measured, is the voltage peak value of the -th modal function spectral component of the third-phase voltage sampling signal at the -th node of the line to be measured, is the voltage peak value of the first modal function spectral component of the first-phase voltage sampling signal at the -th node of the line to be measured, is the voltage peak value of the second modal function spectral component of the first-phase voltage sampling signal at the -th node of the line to be measured, is the voltage peak value of the -th modal function spectral component of the -phase voltage sampling signal at the -th node of the line to be measured, is the voltage peak value of the -th modal function spectral component of the third-phase voltage sampling signal at the -th node of the line to be measured, is the voltage peak value of the The voltage peak of the th modal function spectral component of the third-phase voltage sampling signal at a node; Step S2-2: Traverse all nodes of the circuit under test and their two-dimensional connection relationships to obtain the topological matrix of the circuit under test , that is: When there is a direct branch between node and node , ; When there is no direct branch between node and node , ; In the formula, is the element in the first row and the first column, is the element in the first row and the second column, is the element in the first row and the third column, is the element in the first row and the, , is the element in the second row and the second column, is the element in the second row and the column, is the element in the second row and the column, is the element in the row and the first column, is the element in the row and the second column, is the element in the row and the column, is the element in the row and the column, is the element in the row and the column, is the element in the row and the second column, is the element in the row and the third column, is the element in the row and the column; Step S2-3: Repeat times of sampling, and then establish a fault graph training set; The training set consists of a topological matrix , node voltage matrices of the lines to be measured and node numbers at the grounding wire connection points ; The topological matrix and node voltage matrices of the lines to be measured are used as the input of the deep graph neural network; node numbers at the grounding wire connection points are used as the output of the deep graph neural network; Train the deep graph neural network model; The deep graph neural network consists of 3 convolutional layers and 1 fully connected layer; The output of the convolutional layer is: ， Wherein, and are the inputs of the -th convolutional layer; is the output of the -th convolutional layer; when , is the node voltage matrix of the line to be measured ; is an identity matrix of size ; is the adjacency matrix after plus the identity matrix; is the degree matrix; is the parameter matrix with the dimension to be trained of the -th convolutional layer being ; is the convolution kernel function; Degree matrix is a diagonal matrix whose main diagonal elements are the sum of the elements in each row of the matrix : In the formula, is the adjacency matrix of the th row and th column element; The output of the fully connected layer is: Wherein, is the input of the fully connected layer, i.e., the output of the 3rd convolutional layer ; is the output of the fully connected layer; is the parameter matrix to be trained by the fully connected layer, with a dimension of ; is the pre-set weight matrix, with a dimension of

Images