CN-121980140-A - Haptic timing compensation system and method

CN121980140ACN 121980140 ACN121980140 ACN 121980140ACN-121980140-A

Abstract

The invention discloses a touch time sequence compensation system and a touch time sequence compensation method, and relates to the technical field of power systems. The compensation system comprises a touch signal processing module, a physical field modeling module, a wavelet phase alignment module, a space-time diagram network module and a reinforcement learning optimization module. The haptic signal processing module acquires and converts haptic signals into digital signals for the physical field modeling module to construct a wave equation and a delay matrix, the delay matrix output by the physical field modeling module is used as priori knowledge, the power-assisted wavelet phase alignment module aligns phases of the multi-band signals after wavelet decomposition to generate time sequence alignment signals, the time sequence alignment signals are input into the space-time diagram network module to calculate edge weights, the reinforcement learning optimization module dynamically compensates the haptic signals according to the edge weights, haptic time sequence compensation data is output, and the feedback correction physical field modeling module and the wavelet phase alignment module form closed loop optimization, so that the accuracy of mechanical fault recognition is high, and the accuracy and the robustness of haptic time sequence compensation are improved.

Inventors

TANG QI
FU ZHENGXIN
GUO JUNJIE
LI LANYIN
ZENG QINGHUI

Assignees

广东电网有限责任公司佛山供电局

Dates

Publication Date: 20260505
Application Date: 20260127

Claims (10)

1. The haptic timing compensation system is characterized by comprising a haptic signal processing module, a physical field modeling module, a wavelet phase alignment module, a space-time diagram network module and a reinforcement learning optimization module; The tactile signal processing module is used for acquiring a tactile signal of the substation inspection robot, and performing data conversion on the tactile signal to generate a digital signal; The physical field modeling module is used for carrying out physical field modeling by adopting the digital signals to construct a wave equation and a delay matrix; The wavelet phase alignment module is used for carrying out phase alignment on the multiband signal after wave equation wavelet decomposition by using the delay matrix as priori knowledge to generate a time sequence alignment signal; The time-space diagram network module is used for calculating the edge weights of the time sequence alignment signals by adopting a time-space diagram network to generate a plurality of edge weights; the reinforcement learning optimizing module is used for dynamically compensating the touch signal according to the edge weight to generate touch time sequence compensation data.
2. The haptic timing compensation system of claim 1 wherein the haptic signal processing module comprises a haptic sensor, a charge amplifier, a bandpass filter, and an analog-to-digital converter; The touch sensor is used for acquiring a touch signal of the substation inspection robot; The charge amplifier is used for converting the touch signal into a voltage signal and generating an initial voltage signal; the band-pass filter is used for filtering noise signals in the initial voltage signals and generating target voltage signals; the analog-to-digital converter is used for performing digital conversion on the target voltage signal to generate a digital signal.
3. A haptic timing compensation system as recited in claim 1 wherein said physical field modeling module performs the steps of: Calibrating a medium damping coefficient and an electromagnetic mechanical coupling coefficient corresponding to the touch signal through finite element simulation; constructing a wave equation by adopting the material related wave speed, the external excitation function, the medium damping coefficient and the electromagnetic mechanical coupling coefficient corresponding to the digital signal; and constructing a matrix by adopting the three-dimensional coordinates of the sensor nodes corresponding to the digital signals, the preset material correction coefficient and the material-related wave speed, and generating a delay matrix.
4. A haptic timing compensation system as recited in claim 1 wherein said wavelet phase alignment module performs the steps of: performing complex Morlet wavelet transformation on the touch signal to generate a time-frequency energy spectrum; Calculating an instantaneous phase spectrum by adopting the time-frequency energy spectrum to obtain an instantaneous phase spectrum; constructing a cross-frequency band phase alignment loss function by taking phase consistency as a target and adopting a wavelet coefficient phase angle corresponding to the instantaneous phase spectrum; And performing wavelet domain multi-band phase difference conversion on the delay error corresponding to the delay matrix by adopting the cross-band phase alignment loss function to generate a time sequence alignment signal.
5. A haptic timing compensation system as recited in claim 1 wherein said space-time diagram network module performs the steps of: Respectively calculating the characteristic vector of the sensor node by using the contact pressure, the temperature and the angular velocity corresponding to the digital signal and the vibration frequency spectrum vector corresponding to the time sequence alignment signal to generate a plurality of characteristic vectors; and respectively adopting the feature vector and the delay matrix to calculate the edge weight through a time-space diagram attention mechanism, and generating the edge weight corresponding to the sensor node.
6. A haptic timing compensation system as recited in any one of claims 1-5 wherein said reinforcement learning tuning module performs the steps of: carrying out state space construction by adopting an error change rate corresponding to the time sequence alignment signal, the edge weight, a wave velocity gradient corresponding to the wave equation and a phase angle vector, and generating a state space vector; And dynamically compensating strategy network parameters corresponding to the haptic signals by adopting the state space vector through a dual-delay depth deterministic strategy gradient algorithm to generate haptic time sequence compensation data.
7. A haptic timing compensation method, comprising: the method comprises the steps of obtaining a touch signal of a substation inspection robot, and performing data conversion on the touch signal to generate a digital signal; Modeling a physical field by adopting the digital signal, and constructing a wave equation and a delay matrix; using the delay matrix as priori knowledge, performing phase alignment on the multiband signal after wave equation wavelet decomposition to generate a time sequence alignment signal; performing edge weight calculation on the time sequence alignment signals by adopting a space-time diagram network to generate a plurality of edge weights; And dynamically compensating the touch signal according to the edge weight to generate touch time sequence compensation data.
8. An electronic device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the haptic timing compensation method of claim 7.
9. A computer readable storage medium having stored thereon a computer program, wherein the computer program when executed implements the haptic timing compensation method of claim 7.
10. A computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions, wherein the program instructions, when executed by a computer, cause the computer to perform the haptic timing compensation method of claim 7.

Description

Haptic timing compensation system and method Technical Field The invention relates to the technical field of power systems, in particular to a touch time sequence compensation system and a touch time sequence compensation method. Background In the field of intelligent detection and high-precision signal processing of power equipment of a power system, a substation inspection robot is very important for quick and accurate detection of mechanical faults. If mechanical faults such as jamming of the circuit breaker and abrasion of the isolating switch cannot be found and processed in time, serious electric accidents can be caused, and safe and stable operation of a power grid is affected. The haptic timing detection is used as a key technology for identifying the mechanical faults, and the detection precision and reliability directly determine the fault diagnosis accuracy. At present, the traditional technology has a plurality of defects in the sub-millisecond time sequence detection of the mechanical faults of the transformer substation. For example, in the traditional Kalman filtering method, due to the fact that linear assumption is relied on, when nonlinear delay generated in the elastic wave propagation process is processed, actual requirements are difficult to meet, errors exceed 2ms, time sequence detection results have large deviation, and accurate data support cannot be provided for fault diagnosis. Although the LSTM (Long Short-Term Memory) compensation network has certain advantages in time sequence data processing, due to the lack of physical field modeling capability, under the complex strong electromagnetic interference environment of a transformer substation, the phase drift phenomenon is remarkable, so that the false alarm rate exceeds 18%, and the accuracy of fault identification is greatly influenced. The dynamic time warping (DYNAMIC TIME WARPING, DTW) method, although capable of handling alignment problems of time series data to a certain extent, has a computational complexity as high as O #) The processing delay exceeds 50ms, and the requirement of the substation inspection robot on the real-time detection of the mechanical faults cannot be met. In addition, the complex electromagnetic environment of the transformer substation can also cause distortion phenomenon of the touch data, and the accuracy of mechanical fault identification is further reduced. Disclosure of Invention The invention provides a touch time sequence compensation system and a method, which solve the technical problems that the existing touch time sequence detection method is easy to be interfered by electromagnetic environment, and the processing delay time is long, so that the accuracy of mechanical fault identification is low. The invention provides a touch time sequence compensation system, which comprises a touch signal processing module, a physical field modeling module, a wavelet phase alignment module, a space-time diagram network module and a reinforcement learning optimization module; The tactile signal processing module is used for acquiring a tactile signal of the substation inspection robot, and performing data conversion on the tactile signal to generate a digital signal; The physical field modeling module is used for carrying out physical field modeling by adopting the digital signals to construct a wave equation and a delay matrix; The wavelet phase alignment module is used for carrying out phase alignment on the multiband signal after wave equation wavelet decomposition by using the delay matrix as priori knowledge to generate a time sequence alignment signal; The time-space diagram network module is used for calculating the edge weights of the time sequence alignment signals by adopting a time-space diagram network to generate a plurality of edge weights; the reinforcement learning optimizing module is used for dynamically compensating the touch signal according to the edge weight to generate touch time sequence compensation data. Optionally, the haptic signal processing module includes a haptic sensor, a charge amplifier, a band pass filter, and an analog to digital converter; The touch sensor is used for acquiring a touch signal of the substation inspection robot; The charge amplifier is used for converting the touch signal into a voltage signal and generating an initial voltage signal; the band-pass filter is used for filtering noise signals in the initial voltage signals and generating target voltage signals; the analog-to-digital converter is used for performing digital conversion on the target voltage signal to generate a digital signal. Optionally, the physical field modeling module performs the steps of: Calibrating a medium damping coefficient and an electromagnetic mechanical coupling coefficient corresponding to the touch signal through finite element simulation; constructing a wave equation by adopting the material related wave speed, the external excitation function, the medium damping coeff