CN-121978669-A - Substation personnel safety distance monitoring method and device

CN121978669ACN 121978669 ACN121978669 ACN 121978669ACN-121978669-A

Abstract

The invention provides a substation personnel safety distance monitoring method and device, which belong to the technical field of substations, the invention transmits linear frequency modulation continuous wave signals through a millimeter wave radar module and receives target echo signals, a 4 microphone array synchronously collects environment sound field signals, carries out compressed sensing reconstruction to extract distance and speed information, carries out non-negative matrix factorization to the sound field signals to extract a sound source azimuth angle and footstep sound characteristics, inputs radar characteristics and acoustic characteristics into a multipath interference suppression model, realizes coarse granularity and fine granularity progressive classification to output target real position coordinates through a layered soft maximum classifier, calculates the safety distance according to the boundary between a target position and a dangerous area of equipment, triggers LED display lamp flickering and loudspeaker voice warning when the distance is smaller than a preset threshold, and solves the technical problem of personnel position erroneous judgment caused by radar multipath interference in the substation environment.

Inventors

WANG JUNWU
SUN KEZHENG
WANG SHUAI
SUN LIANGXIAO
CAI QILIANG
YU CHAOYANG

Assignees

国网山东省电力公司青岛市即墨区供电公司

Dates

Publication Date: 20260505
Application Date: 20251219

Claims (10)

1. A transformer substation personnel safety distance monitoring method is characterized by comprising the steps of transmitting a linear frequency modulation continuous wave signal by a millimeter wave radar module and receiving a target echo signal, synchronously acquiring an environment sound field signal by a 4-microphone array, acquiring device space coordinates by a positioning module, performing undersampling reconstruction processing based on compressed sensing on the target echo signal, establishing sparse representation in a wavelet domain, solving an L1 norm minimization problem by an orthogonal matching tracking algorithm, recovering a complete echo signal to extract target distance information and target radial velocity information, performing non-negative matrix decomposition processing on the environment sound field signal, decomposing a time-frequency matrix into a sound source base matrix and an activation coefficient matrix, extracting a foot step sound feature vector and a voice feature vector, calculating a sound source azimuth angle, inputting the target distance information, the target radial velocity information, echo amplitude, the sound source azimuth angle and the foot step sound feature vector into a multipath interference suppression model, outputting a target real position coordinate and a false multipath echo mark set, triggering a high-brightness LED display lamp to flash and sending a voice warning through a 5W explosion-proof loudspeaker when the target device shortest distance value is smaller than a preset safety distance threshold, and packaging the target position coordinate, the target device and the target position azimuth angle and a time stamp to a PC end or a data package.
2. The method according to claim 1, wherein the undersampling reconstruction process based on compressed sensing, specifically, performing a deskewing process on a received target echo signal to obtain an intermediate frequency signal, performing random undersampling on the intermediate frequency signal at a rate lower than a nyquist sampling rate to obtain a measurement vector, performing wavelet transformation on the intermediate frequency signal to obtain a wavelet coefficient matrix, and constructing a product of a gaussian random observation matrix and a wavelet transformation matrix as the sensing matrix.
3. The method of claim 2, wherein the solving of the orthogonal matching pursuit algorithm adopts an iterative mode, initializes a residual vector as a measurement vector, initializes a support set as an empty set, calculates an inner product of each column of the sensing matrix and the residual vector, selects a column index with the largest absolute value of the inner product, adds the column index to the support set, and extracts a column corresponding to the support set in the sensing matrix to form a submatrix for least square fitting.
4. A method according to claim 3, further comprising verifying that the constrained equidistant constant of the perceptual matrix is less than 0.3, randomly selecting 500 sets of different column combinations from the perceptual matrix, calculating singular values for the submatrices formed by each set of column combinations, and calculating the ratio of the largest singular value to the smallest singular value of the submatrices as the condition number, prior to the undersampling reconstruction process based on compressed perceptual.
5. The method according to claim 2, wherein the wavelet coefficient matrix satisfying the sparsity condition means that the number of coefficients in the wavelet coefficient matrix above the energy threshold is less than 12%, and the optimization objective function is expressed as minimizing the L1 norm of the wavelet coefficient vector, provided that the L2 norm of the measurement vector subtracted from the product of the perceptual matrix and the wavelet coefficient vector is less than the noise energy estimate.
6. The method of claim 1, wherein the non-negative matrix factorization process specifically performs short-time fourier transform on the 4 microphone channel signals to obtain a time-frequency matrix, concatenates the time-frequency matrix magnitudes of the 4 channels into a joint time-frequency matrix, initializes a sound source base matrix and an activation coefficient matrix into random positive matrices, establishes a non-negative matrix factorization objective function, and iteratively optimizes using a multiplicative update rule.
7. The method of claim 6, wherein the non-negative matrix factorization objective function is expressed as a Frobenius norm square plus an L1 norm sparsity penalty term and a time smoothing regularization term between minimizing a joint time-frequency matrix and a product of a sound source base matrix and an activation coefficient matrix, wherein the L1 norm sparsity penalty term coefficient is 0.01 and the time smoothing regularization term coefficient is 0.05.
8. The method according to claim 6, wherein the step sound feature vector is extracted by identifying a step sound corresponding column from a sound source base matrix obtained by decomposition, extracting a row vector of a corresponding activation coefficient matrix as a time domain activation sequence, and extracting a peak interval, a zero crossing rate, and an energy envelope for the step sound time domain activation sequence as step sound feature vectors.
9. The transformer substation personnel safety distance monitoring device is characterized by comprising a millimeter wave radar module, a microphone array, an external input module, a main control chip, a storage module, a positioning module, an LED display lamp, an explosion-proof loudspeaker, a communication module, a function execution module, an external output module, a rechargeable lithium battery pack, a power supply module and a battery management system, wherein the main control chip is respectively electrically connected with the millimeter wave radar module, the microphone array, the external input module, the storage module, the positioning module, the LED display lamp, the explosion-proof loudspeaker, the communication module, the function execution module, the external output module, the rechargeable lithium battery pack, the power supply module and the battery management system, the millimeter wave radar module is used for transmitting linear frequency modulation continuous wave signals and receiving target echo signals, the microphone array is used for collecting environment sound field signals and realizing sound source localization, the external input module is used for receiving the target echo signals and the environment sound field signals and carrying out analog-digital conversion and signal preprocessing, the storage module is used for storing instruction codes, model parameters, operation data and alarm records, the positioning module is used for acquiring device space coordinates and providing position information, the high-brightness LED display lamp is used for outputting lamplight alarms with different colors and flickering modes, the anti-explosion loudspeaker is used for playing voice alarms, the communication module is used for transmitting alarm data packets and operation states to a mobile end or a PC end, the function execution module is used for driving the LED display lamp and the anti-explosion loudspeaker to execute alarm output and trigger the communication module to carry out data transmission, the power module is used for providing stable voltage for each hardware module.
10. The apparatus of claim 9, wherein the memory module stores instruction code, and the main control chip executes the instruction code to perform the method of any one of claims 1-8.

Description

Substation personnel safety distance monitoring method and device Technical Field The invention belongs to the technical field of substations, and particularly relates to a substation personnel safety distance monitoring method and device. Background The safety distance monitoring of the transformer substation personnel generally adopts millimeter wave radar technology to realize non-contact target detection, extracts distance and speed information by transmitting linear frequency modulation continuous wave signals and receiving target echo signals, and triggers an alarm when the distance between the personnel and high-voltage equipment is smaller than a safety threshold value. The traditional millimeter wave radar monitoring method is faced with the serious multipath interference problem in the complex electromagnetic environment of the transformer substation, and radar signals are reflected by a metal equipment cabinet body, a pipeline and the ground to generate false echoes, so that the system misjudges the multipath reflections as real targets or cannot accurately position the real positions of personnel. In the prior art, the multipath interference is suppressed by Doppler filtering or constant false alarm rate detection, but the methods only depend on single radar signal characteristics, fail when Doppler characteristics are not obvious under static personnel or low-speed moving scenes, and cannot distinguish physical sources of different types of multipath reflections. That is, in the prior art, there is a technical problem that the radar multipath interference in the transformer substation environment leads to misjudgment of the personnel position. Disclosure of Invention In view of the above, the invention provides a method and a device for monitoring the safety distance of a transformer substation worker, which can solve the technical problem of misjudgment of the worker position caused by radar multipath interference in the transformer substation environment in the prior art. The invention provides a substation personnel safety distance monitoring method which comprises the steps that a millimeter wave radar module transmits a linear frequency modulation continuous wave signal and receives a target echo signal, a 4-microphone array synchronously collects an environment sound field signal, a positioning module acquires device space coordinates, a sparse representation is built in a wavelet domain through undersampling reconstruction processing based on compressed sensing of the target echo signal, an L1 norm minimization problem is solved through an orthogonal matching tracking algorithm to recover a complete echo signal, target distance information and target radial velocity information are extracted, a non-negative matrix decomposition processing is conducted on the environment sound field signal, a time-frequency matrix is decomposed into a sound source base matrix and an activation coefficient matrix, a foot step sound feature vector and a voice feature vector are extracted, a sound source azimuth angle is calculated, a target distance information, a target radial velocity information, an echo amplitude, a sound source azimuth angle and a foot step sound feature vector are input into a multipath interference suppression model, a target real position coordinate and a false multipath echo mark set are calculated, a target device shortest distance value is calculated according to the target real position coordinate and a device dangerous area boundary parameter, when the shortest distance value of the target device is smaller than a preset safety distance threshold, a function execution module triggers a high-brightness LED display lamp and sends out a flash warning signal through a 5W explosion-proof loudspeaker, and the target position, the target position module and the target position information and the target position and the foot step sound feature vector are packaged to a real position data or a real time-position and a real alarm terminal. The undersampling reconstruction processing based on compressed sensing specifically comprises the steps of performing declining processing on a received target echo signal to obtain an intermediate frequency signal, performing random undersampling on the intermediate frequency signal at a rate lower than a Nyquist sampling rate to obtain a measurement vector, performing wavelet transformation on the intermediate frequency signal to obtain a wavelet coefficient matrix, and constructing a product of a Gaussian random observation matrix and a wavelet transformation matrix as a sensing matrix. The method comprises the steps of solving an orthogonal matching pursuit algorithm, initializing residual vectors as measurement vectors, initializing a support set as an empty set, calculating the inner product of each column of a sensing matrix and the residual vectors, selecting a column index with the largest absolute value of the inner product, adding the colum