CN-121659091-B - Intelligent monitoring system of coal conveying system of thermal power plant based on multi-source perception and edge calculation

CN121659091BCN 121659091 BCN121659091 BCN 121659091BCN-121659091-B

Abstract

The invention belongs to the technical field of industrial control, and discloses an intelligent monitoring system of a coal conveying system of a thermal power plant based on multi-source perception and edge calculation; the method comprises the steps of collecting multi-mode sensing data through a heterogeneous sensor array, performing deep collaborative extraction on cross-mode characteristics such as vision, acoustics, vibration, temperature, current and the like by using a physical constraint correlation model, dynamically identifying abnormal modes of equipment according to the deep collaborative extraction, and introducing a hierarchical fusion analysis framework to provide multi-level confidence guarantee for fault diagnosis. The invention also realizes an incremental correction mechanism of state deviation real-time feedback, avoids control deviation caused by model mismatch, ensures that each abnormal symptom of a coal conveying link can be accurately captured and adaptively treated, and remarkably improves the safety of the system and the intelligent level of operation and maintenance.

Inventors

ZHU ZHENLIN
JING TAO
SUI XIUKUN
LI RUI
YAN ZHONGXIAN

Assignees

国能吉林江南热电有限公司

Dates

Publication Date: 20260508
Application Date: 20260130

Claims (9)

1. Heat-engine plant coal conveying system intelligent monitoring system based on multisource perception and edge calculation, which is characterized by comprising: The multi-source data acquisition module is used for acquiring multi-mode data of the coal conveying link according to the deployed heterogeneous sensor array to obtain multi-source sensing data, wherein the multi-source sensing data comprise a visual image sequence, equipment acoustic signals, temperature distribution data, vibration time sequence data and current waveform data; the data preprocessing module is used for carrying out environment interference compensation on the multi-source perceived data to obtain multi-source purified data, and carrying out deep collaborative extraction of cross-modal characteristics on the multi-source purified data to obtain a collaborative feature set; The acquisition process of the collaborative feature set comprises the following steps: performing space-time segmentation and multi-scale decomposition on the moving target in the purified visual image, and identifying and reconstructing the complete spatial morphological characteristics of the moving target; Performing combined time-frequency analysis on the purified equipment acoustic signals based on wavelet packet decomposition and short-time Fourier transform, and identifying and extracting fundamental frequency components and fault modulation components of equipment rotating parts; Carrying out space evolution analysis on the purified temperature distribution data in a thermal anomaly region, inverting the position and the intensity of a heat source, and generating thermal anomaly propagation characteristics; Performing empirical mode decomposition on the purified vibration time sequence data, and determining the distribution characteristics of vibration energy in a time-frequency plane in the vibration time sequence data based on the empirical mode decomposition; Carrying out load fluctuation mode identification on the purified current waveform data, and extracting load fluctuation characteristics by combining the load power spectrum density of the current waveform data; Constructing a mutual information matrix based on the extracted cross-modal characteristics, constructing a physical constraint association model, and carrying out characteristic projection transformation on the mutual information matrix based on the mutual information matrix to obtain a collaborative characteristic set, wherein the cross-modal characteristics comprise complete spatial morphological characteristics, fundamental frequency components, fault modulation components, thermal anomaly propagation characteristics, load fluctuation characteristics and distribution characteristics of vibration energy in a time-frequency plane; The edge decision module is used for constructing a layered fusion analysis framework, and carrying out abnormal event identification and deep analysis by combining the collaborative feature set to obtain the global running state of the coal conveying link; The intelligent control and feedback module is used for carrying out progressive diagnosis and propagation path deduction on the running state of the current coal conveying link according to the global running state, generating an adaptive control strategy for the coal conveying link through simulation deduction, converting the adaptive control strategy into a control instruction and transmitting the control instruction to a control mechanism of the coal conveying link.
2. The intelligent monitoring system of the coal conveying system of the thermal power plant based on multi-source sensing and edge calculation according to claim 1, wherein the process of acquiring multi-source purification data comprises the following steps: Performing joint analysis according to the prior distribution of dark channels and the attenuation degree of spatial frequency of each visual image in the visual image sequence to obtain the equivalent transmittance of dust concentration in the current environment; performing multi-scale feature decomposition on the visual image based on the equivalent transmittance of dust concentration, dividing the visual image into a dust-free intrinsic layer and a dust scattering layer, and performing image processing on the original visual image based on the dust-free intrinsic layer to obtain a purified visual image; Performing real-time dynamic tracking on an environmental noise baseline of the acquired device acoustic signal, identifying a steady-state noise component, and constructing a time-varying parameter model based on the steady-state noise component; identifying and extracting electromagnetic interference characteristics of the vibration time sequence data and the current waveform data, distinguishing equipment mechanical impact signals and electromagnetic pulse interference signals, and performing template matching and self-adaptive cancellation on the electromagnetic pulse interference signals based on a pre-established interference characteristic library to respectively obtain purified vibration time sequence data and current waveform data; carrying out non-uniformity correction and temperature calibration compensation on the temperature distribution data to obtain purified temperature distribution data; and summarizing all the data after the purification treatment to obtain corresponding multi-source purification data.
3. The intelligent monitoring system of the coal conveying system of the thermal power plant based on multi-source perception and edge calculation according to claim 1, wherein the implementation process of the edge decision module comprises the following steps: deploying a layered fusion analysis framework in an edge computing node of a corresponding coal conveying link, wherein the layered fusion analysis framework comprises an edge quick response layer, an edge depth analysis layer and a cloud collaborative verification layer; The edge quick response layer is used for constructing a reference feature space in a normal running state according to the cross-modal features, acquiring the mahalanobis distance between the reference feature space and the corresponding cooperative features, and quickly judging based on the mahalanobis distance to obtain an abnormal event log; Arranging a refined fault diagnosis model in an edge depth analysis layer, and inputting a collaborative feature set corresponding to occurrence time in an abnormal event log into the refined fault diagnosis model to obtain an output result; carrying out decision evaluation according to the confidence coefficient of the output result, if the confidence coefficient is smaller than a preset second threshold value, marking the corresponding output result as a low-confidence decision, triggering a cloud collaborative verification layer, carrying out association analysis of global historical data and expert knowledge base verification in the cloud collaborative verification layer, and updating the low-confidence decision based on the verification result; and carrying out data compression and feature extraction on the intermediate result and the final result of the analysis to obtain a global running state.
4. The intelligent monitoring system of the thermal power plant coal conveying system based on multi-source sensing and edge calculation according to claim 3, wherein the implementation process of the intelligent control and feedback module comprises the following steps: according to various discrimination nodes in the multi-level fault discrimination decision tree, carrying out fault combination discrimination on the coal conveying link by combining the cooperative characteristic set and the global running state to obtain an output result of the multi-level fault discrimination decision tree; uncertainty quantization is carried out on the output result of the multi-level fault discrimination decision tree, probability distribution of decision confidence is generated, and variance of the decision confidence is calculated; when the variance of the decision confidence coefficient is not greater than the preset confidence threshold, generating an intelligent monitoring report based on the discrimination result of the corresponding multi-level fault discrimination decision tree, wherein the intelligent monitoring report comprises fault type, fault position, fault severity, fault propagation path, corresponding risk probability and decision confidence coefficient; And generating an adaptive control strategy according to the fault severity and the fault type, packaging the generated adaptive control strategy into a standardized control instruction, and issuing the standardized control instruction to a control mechanism of the coal conveying link.
5. The intelligent monitoring system of the coal conveying system of the thermal power plant based on multi-source sensing and edge calculation according to claim 4, wherein the root node of the multi-level fault distinguishing decision tree is a normal and abnormal classification, the subsequent level nodes are three decision nodes of a fault class distinguishing node, a fault part positioning node and a fault severity evaluation node in sequence, and each decision node is distinguished based on different combinations of a cooperative feature set.
6. The intelligent monitoring system of the thermal power plant coal conveying system based on multi-source sensing and edge calculation according to claim 4, wherein the fault combination discrimination process comprises the following steps: performing modal decomposition on the collaborative feature set based on fault category discrimination nodes in the multi-level fault discrimination decision tree, identifying and marking modal sources of dominant fault features, establishing a modal combination strategy library aiming at fault types, and determining specific fault types based on the modal combination strategy library; Determining the space position of a fault source by a fault part positioning node in the multi-level fault discrimination decision tree according to the space topology of the abnormal sensor array and the propagation delay analysis of fault characteristics; Performing fault degradation curve fitting by a fault severity assessment node in a multi-level fault discrimination decision tree through tracking fault evolution trend of fault features on a time axis, and predicting the residual time of the fault features reaching a dangerous threshold; performing fault propagation path deduction analysis based on the predicted remaining time and fault position, and calculating each fault propagation path And generates the output result of the multi-level fault discrimination decision tree.
7. The intelligent monitoring system of the coal conveying system of the thermal power plant based on multi-source sensing and edge calculation as claimed in claim 4, wherein the generation process of the self-adaptive control strategy comprises the following steps: Performing strategy matching from a preset control strategy library according to the fault type and the fault severity to obtain a candidate control strategy set; Forward simulation deduction is carried out on each candidate control strategy in the candidate control strategy set by utilizing a digital twin model of the coal conveying link, and a system state evolution track after the candidate control strategy is executed is predicted; Performing strategy scoring on each candidate control strategy based on the system state evolution track to obtain comprehensive scores corresponding to each candidate control strategy, and taking the candidate control strategy with the highest comprehensive score as an initial control adjustment strategy; and carrying out parameter fine adjustment on the initial control adjustment strategy, and taking the adjusted initial control adjustment strategy as a self-adaptive control strategy.
8. The intelligent monitoring system of the coal conveying system of the thermal power plant based on multi-source sensing and edge calculation according to claim 2, wherein the purified visual image acquisition process comprises the following steps: performing pixel point traversal on the dust-free intrinsic layer by adopting local windows to obtain local contrast in each local window, and mapping the local contrast into an intuitive contrast distribution thermodynamic diagram; identifying low contrast areas, medium contrast areas, and high contrast areas within the original visual image based on the contrast distribution thermodynamic diagram; and carrying out differential enhancement treatment on each contrast area, and carrying out smooth fusion on each contrast area after the enhancement treatment is completed to obtain a purified visual image.
9. The intelligent monitoring system of the coal conveying system of the thermal power plant based on multi-source sensing and edge calculation according to claim 2, wherein the design process of the adaptive frequency domain notch filter bank comprises the following steps: The method comprises the steps of carrying out short-time Fourier transform on a current analysis frame of an acoustic signal of equipment to obtain a time-frequency plane representation, identifying a spectrum ridge line of environmental noise on the time-frequency plane, synchronously obtaining a corresponding time sequence evolution sequence by locating a local maximum point as a noise center frequency, predicting the noise center frequency at the future time based on the time sequence of the noise center frequency, constructing a notch filter at each predicted noise center frequency, and cascading and combining all the notch filters to form a self-adaptive frequency domain notch filter bank.

Description

Intelligent monitoring system of coal conveying system of thermal power plant based on multi-source perception and edge calculation Technical Field The invention relates to the technical field of industrial control, in particular to an intelligent monitoring system of a coal conveying system of a thermal power plant based on multi-source perception and edge calculation. Background With the deep development of intelligent manufacturing and industrial internet, thermal power plants are undergoing a transition upgrade from traditional artificial operation and maintenance to intelligent monitoring as a key infrastructure. The coal conveying system is used as an energy artery of a thermal power plant, integrates complex industrial scenes of mechanical transmission, material conveying and multi-equipment cooperation, and is safe and stable to operate and directly related to power generation efficiency and personnel safety. The traditional coal conveying system depends on manual inspection and experience judgment, and faces a plurality of technical bottlenecks in complex fault mode identification and preventive maintenance. In recent years, the multisource sensing technology and edge calculation have made remarkable breakthrough in the field of industrial monitoring, and technical support is provided for intelligent transformation of a coal conveying system. However, the existing multi-source sensing monitoring technology faces some problems when being applied to a coal conveying system of a thermal power plant, and particularly presents the problems of feature conflict and decision contradiction when heterogeneous data fusion is used for processing complex coupling relations of multi-mode sensing signals. As the monitoring coverage is expanded, data collected by sensors of different modes such as vision, acoustics, vibration, temperature, current and the like show isomerism and time sequence misalignment characteristics, fault characteristic space expands sharply, and semantic gaps exist among modes. The existing solutions such as weighted voting fusion or back-end cloud centralized processing cannot solve the problem of feature semantic alignment, introduce extra network delay and bandwidth occupation, and are difficult to meet the real-time requirements of quick response of faults. In view of the above, the present invention proposes an intelligent monitoring system for a coal conveying system of a thermal power plant based on multi-source sensing and edge calculation to solve the above problems. Disclosure of Invention In order to overcome the defects in the prior art, the invention provides the following technical scheme for achieving the purposes: Intelligent monitoring system of coal conveying system of thermal power plant based on multisource perception and edge calculation includes: The multi-source data acquisition module is used for acquiring multi-mode data of the coal conveying link according to the deployed heterogeneous sensor array to obtain multi-source sensing data, wherein the multi-source sensing data comprise a visual image sequence, equipment acoustic signals, temperature distribution data, vibration time sequence data and current waveform data; the data preprocessing module is used for carrying out environment interference compensation on the multi-source perceived data to obtain multi-source purified data, and carrying out deep collaborative extraction of cross-modal characteristics on the multi-source purified data to obtain a collaborative feature set; The edge decision module is used for constructing a layered fusion analysis framework, and carrying out abnormal event identification and deep analysis by combining the collaborative feature set to obtain the global running state of the coal conveying link; The intelligent control and feedback module is used for carrying out progressive diagnosis and propagation path deduction on the running state of the current coal conveying link according to the global running state, generating an adaptive control strategy for the coal conveying link through simulation deduction, converting the adaptive control strategy into a control instruction and transmitting the control instruction to a control mechanism of the coal conveying link. Further, the process of acquiring the multi-source purge data includes: Performing joint analysis according to the prior distribution of dark channels and the attenuation degree of spatial frequency of each visual image in the visual image sequence to obtain the equivalent transmittance of dust concentration in the current environment; performing multi-scale feature decomposition on the visual image based on the equivalent transmittance of dust concentration, dividing the visual image into a dust-free intrinsic layer and a dust scattering layer, and performing image processing on the original visual image based on the dust-free intrinsic layer to obtain a purified visual image; Performing real-time dynamic tracking on an