Self-diagnosis method and system for unattended heating station boiler equipment faults

Abstract

This invention relates to the field of thermal power plant boiler equipment technology, and in particular to a method and system for self-diagnosis of faults in unattended thermal power plant boiler equipment. The method collects data on thermal efficiency, heat flow distribution, and environmental parameters under different operating conditions to construct a thermal characteristic dataset. Based on changes in thermal gradient and combined with system geometry and material thermodynamic parameters, it analyzes the trends of heat accumulation and dissipation, generating a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data. Based on this, it constructs a thermal path network model, identifies thermal risk propagation paths, determines the fixed and dynamic sensing point locations of potential fault areas, and forms an optimized layout scheme for sensing points throughout the entire operating cycle. Monitoring devices are deployed in key areas to collect operational data, extract thermal dynamic characteristics, identify abnormal heat transfer modes, achieve accurate self-diagnosis of equipment faults, and trigger fault warning and self-repair mechanisms, effectively supporting intelligent operation and maintenance and continuous heating assurance of thermal power plant boilers in unattended scenarios.

Description

Technical Field

[0001] This invention relates to the field of thermal power plant boiler equipment technology, and in particular to a method and system for self-diagnosis of faults in unattended thermal power plant boiler equipment. Background Technology

[0002] With the continuous expansion of urban centralized heating systems and the increasing demand for intelligent operation and maintenance, the boiler in the heating station, as the core heat energy conversion equipment in the heating network, directly affects the heating continuity and energy utilization efficiency of the entire region. Traditional boiler equipment relies heavily on manual inspections and periodic maintenance, making it difficult to achieve real-time and comprehensive perception of the internal thermal status of the equipment. Especially in unattended operation, sudden faults (such as local overheating, heat transfer deterioration, coking and ash accumulation) can easily lead to heating interruptions or equipment damage, seriously affecting people's livelihood and operational economy. Existing fault monitoring methods are mostly based on single temperature or pressure threshold alarms, lacking the ability to deeply analyze the overall thermal behavior of the heating system. They cannot accurately identify the early thermal anomaly evolution trend and its impact path on key components, resulting in delayed fault diagnosis, high false alarm rate, and difficulty in supporting accurate early warning and adaptive response.

[0003] In recent years, although some studies have attempted to introduce multi-sensor data fusion and machine learning techniques for thermal condition monitoring, they generally suffer from problems such as blind sensing layout, shallow feature extraction, and fault reasoning mechanisms divorced from physical mechanisms. These methods fail to fully integrate the geometric structure, material properties, and dynamic heat transfer laws of the thermal system, resulting in poor generalization ability and weak interpretability of diagnostic models. Furthermore, existing methods often focus on anomaly detection under steady-state conditions, neglecting the spatiotemporal cumulative effects and dynamic evolution of thermal risks under complex variable load heating conditions. This makes it difficult to achieve a closed-loop intelligent operation and maintenance transformation from passive alarm to proactive prediction, self-diagnosis, and self-repair.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a self-diagnosis method and system for unattended heating station boiler equipment faults. It aims to solve the technical problems of existing heating station boiler fault monitoring methods, which are difficult to achieve early and accurate perception of thermal anomalies and dynamic identification of fault evolution paths under unattended conditions. These problems include blind perception layout, delayed diagnosis, high false alarm rate, and lack of physical mechanism support.

[0006] To achieve the above objectives, the present invention provides a self-diagnosis method for faults in unattended heating station boiler equipment, the method comprising: By collecting thermal efficiency data, heat flow distribution data, and environmental parameter data under different heating conditions through a multi-modal sensor array pre-arranged in the thermal system of the heating station boiler equipment, a thermal characteristic dataset of the thermal system is obtained. The thermal gradient change between adjacent measuring points in the multimodal sensor array is calculated based on the thermal characteristic dataset. The accumulation and dissipation trend of heat in the system is analyzed by combining the geometric characteristics of the thermal system and the thermodynamic parameters of the material, and a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data are obtained. Based on the three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data, a thermal path network model of the thermal system and key equipment is constructed. The location of fixed sensing points and dynamic sensing points in the potential fault zone of the heat station boiler is determined through the thermal path network model, and a full-cycle operating condition sensing point layout scheme is obtained. According to the full-cycle operating condition sensing point layout scheme, operating condition monitoring devices are deployed in the potential fault area of ​​the boiler in the heating station to collect equipment operation data. Thermal dynamic characteristics are extracted from the equipment operation data, and abnormal heat transfer modes in the thermal system are identified based on the thermal dynamic characteristics to obtain equipment fault self-diagnosis results. Fault warning and self-repair triggering are performed based on the equipment fault self-diagnosis results.

[0007] Optionally, the step of calculating the thermal gradient change between adjacent measuring points in the multimodal sensor array based on the thermal characteristic dataset, and analyzing the accumulation and dissipation trend of heat in the system in combination with the geometric characteristics of the thermal system and the thermodynamic parameters of the materials, to obtain a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data, includes: Environmental noise is eliminated from the thermal efficiency data by adaptive Kalman filtering to obtain filtered thermal efficiency field data. Based on the thermal efficiency field data and the heat flow distribution data, the thermal gradient vector and rate of change of adjacent measuring points in the multimodal sensor array are calculated, and the thermal field is dynamically corrected in combination with the environmental parameter data to obtain a preliminary thermal distribution model. Based on the preliminary thermal distribution model and the geometric characteristics of the thermodynamic system, a thermal equilibrium differential equation is established. The finite volume method is used to solve the thermal equilibrium differential equation to obtain the continuous thermal field distribution inside the thermodynamic system. The thermal gradient abrupt change regions and thermal flow anomaly regions are identified from the continuous thermal field distribution and heat flow distribution data. The thermal risk accumulation index of each thermal gradient abrupt change region and thermal flow anomaly region is calculated by combining the material thermodynamic parameters, so as to obtain a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data.

[0008] Optionally, the step of calculating the thermal gradient vector and rate of change of adjacent measuring points in the multimodal sensor array based on the thermal efficiency field data and the heat flux distribution data, and performing dynamic thermal field correction in conjunction with the environmental parameter data to obtain a preliminary thermal distribution model includes: Based on the thermal efficiency field data, the thermal parameter values ​​of each measuring point are spatially mapped in the three-dimensional spatial coordinate system of the multimodal sensor array to obtain the initial thermal spatial distribution map. The thermal gradient vector is calculated for the thermal parameter values ​​of adjacent measuring points in the initial thermal spatial distribution diagram. The thermal change rate of each measuring point along the principal axis of the thermal system is calculated using the higher-order finite difference method to obtain the thermal gradient vector field. Based on the correspondence between the heat flow distribution data and the thermal gradient vector field, the equivalent thermal resistance coefficient of each region of the thermodynamic system is calculated using the robust regression method, and a heat flow-thermal efficiency relationship model is established. The heat flow-thermal efficiency relationship model is dynamically compensated using the environmental parameter data to obtain a preliminary thermal distribution model.

[0009] Optionally, the step of constructing a thermal path network model of the thermal system and key equipment based on the three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data, and determining the locations of fixed and dynamic sensing points in the potential fault zone of the boiler in the heat station through the thermal path network model to obtain a full-cycle operating condition sensing point layout scheme includes: Density clustering analysis is performed on thermal anomaly points within the thermal system based on the three-dimensional thermal anomaly distribution map. Based on the results of the clustering analysis, similar thermal anomaly points are merged into thermal risk zones. The heat transfer direction and attenuation rate of each thermal risk zone are calculated based on the thermal risk accumulation rate data to obtain a thermal risk impact map. Based on the thermal risk impact map and the geometric topological relationship between the thermal system and key equipment, a weighted graph-form thermal path network model is constructed, where nodes represent thermal risk zones, edges represent heat transfer paths, and weights represent dynamic coefficients of heat conduction. The heat path network model is subjected to critical path analysis of heat propagation. The heat transfer time window and heat attenuation threshold from each heat risk zone to the core component of the heating station boiler are calculated. Based on the heat transfer time window and heat attenuation threshold, each heat risk zone is sorted according to the degree of impact on heating continuity, and a heat risk importance ranking table is obtained. Based on the thermal risk importance ranking table, the locations of fixed and dynamic sensing points in the potential fault zone of the heating station boiler are determined, resulting in a full-cycle operating condition sensing point layout scheme.

[0010] Optionally, the step of determining the locations of fixed and dynamic sensing points in the potential fault zone of the boiler in the heating station based on the thermal risk importance ranking table, and obtaining a full-cycle operating condition sensing point layout scheme, includes: The thermal risk zones are classified according to the thermal risk importance ranking table, and the dynamic coverage radius of each thermal risk zone is calculated to obtain the thermal risk classification table. A working condition stability analysis is performed on the high-risk areas in the heat risk classification table. The root mean square error of the position deviation of each high-risk area under different heating loads is calculated, and the high-risk areas are divided into steady-state risk areas and transient risk areas based on the root mean square error of the position deviation. Fixed sensing points are set according to the location of the steady-state risk zone. The minimum coverage set algorithm is used to optimize the number of fixed sensing points so that each steady-state risk zone is covered by at least one fixed sensing point, while maximizing the sensing redundancy, thus obtaining a fixed sensing point layout scheme. Based on the movement trajectory and frequency characteristics of the transient risk zone, dynamic sensing points are set up in the transient risk zone to obtain a dynamic sensing point layout scheme. The fixed sensing point layout scheme and the dynamic sensing point layout scheme are then integrated to form a full-cycle working condition sensing point layout scheme.

[0011] Optionally, the step of deploying operating condition monitoring devices in the potential fault area of ​​the boiler in the heat station according to the full-cycle operating condition sensing point layout scheme to collect equipment operating data, extracting thermal dynamic characteristics from the equipment operating data, identifying abnormal heat transfer modes in the thermal system based on the thermal dynamic characteristics, obtaining equipment fault self-diagnosis results, and triggering fault early warning and self-repair based on the equipment fault self-diagnosis results includes: According to the full-cycle operating condition sensing point layout scheme, operating condition monitoring devices are deployed in the potential fault area of ​​the thermal station boiler to collect equipment operation data. Multi-scale feature extraction is performed on the equipment operation data by empirical mode decomposition to separate the steady-state and transient components of the equipment operation data and obtain the thermal dynamic feature vector. The thermal state matrix of the thermal system is constructed based on the thermal dynamic feature vector, and the model is matched with a pre-set normal thermal mode benchmark library to obtain the thermal anomaly index. The thermal anomaly index is combined with heating condition parameters for multi-dimensional correlation analysis. An anomaly mode is classified using a fault self-diagnosis algorithm based on the characteristics of the thermal system. The continuous heating guarantee risk assessment index of the equipment is calculated to obtain the equipment fault self-diagnosis results. Fault warning and self-repair triggering are performed based on the equipment fault self-diagnosis results.

[0012] Optionally, the step of performing multi-dimensional correlation analysis on the thermal anomaly index in conjunction with heating condition parameters, classifying anomaly modes using a fault self-diagnosis algorithm based on thermal system characteristics, calculating the equipment continuous heating guarantee risk assessment index, obtaining equipment fault self-diagnosis results, and triggering fault early warning and self-repair based on the equipment fault self-diagnosis results includes: The three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data are combined with historical heating condition parameters to construct a time-series feature space. Based on the structural characteristics of the thermal system, the data points in the feature space are subjected to condition-adaptive clustering to form a condition-driven benchmark fault mode library. The thermal anomaly index is combined with the real-time heating condition parameters and the trajectory is located in the constructed time-series feature space. The dynamic time warping distance is calculated with each mode in the benchmark fault mode library, and the anomaly mode with the smallest distance is selected as the fault mode classification result. Based on the fault mode classification results and the thermal path network model, a digital twin simulation of the heat transfer process in the thermal system is performed to predict the evolution trend of thermal parameters of key equipment. The continuous heating guarantee risk assessment index is calculated according to the thermal tolerance threshold of the equipment to obtain the equipment fault self-diagnosis results. Fault warning and self-repair triggering are performed based on the equipment fault self-diagnosis results.

[0013] Furthermore, to achieve the above objectives, the present invention also provides a self-diagnosis system for unattended heating station boiler equipment faults, the system comprising: The data acquisition module is used to collect thermal efficiency data, heat flow distribution data and environmental parameter data under different heating conditions through a multi-modal sensor array pre-arranged in the thermal system of the heating station boiler equipment, so as to obtain a dataset of thermal characteristics of the thermal system. The thermal field analysis module is used to calculate the thermal gradient change between adjacent measuring points in the multimodal sensor array based on the thermal characteristic dataset, and to analyze the accumulation and dissipation trend of heat in the system by combining the geometric characteristics of the thermal system and the thermodynamic parameters of the material, so as to obtain a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data. The perception planning module is used to construct a thermal path network model of the thermal system and key equipment based on the three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data. The thermal path network model is used to determine the location of fixed and dynamic perception points in the potential fault area of ​​the heat station boiler, and to obtain a full-cycle operating condition perception point layout scheme. The diagnostic and early warning module is used to deploy operating condition monitoring devices in the potential fault area of ​​the boiler in the heating station according to the full-cycle operating condition sensing point layout scheme to collect equipment operation data, extract thermal dynamic characteristics from the equipment operation data, identify abnormal heat transfer modes in the thermal system based on the thermal dynamic characteristics, obtain equipment fault self-diagnosis results, and trigger fault early warning and self-repair based on the equipment fault self-diagnosis results.

[0014] In addition, to achieve the above objectives, the present invention also provides a self-diagnostic device for unattended heating station boiler equipment faults. The device includes: a memory, a processor, and a self-diagnostic program for unattended heating station boiler equipment faults stored in the memory and executable on the processor. The self-diagnostic program for unattended heating station boiler equipment faults is configured to implement the steps of the self-diagnostic method for unattended heating station boiler equipment faults as described in any of the above descriptions.

[0015] In addition, to achieve the above objectives, the present invention also provides a medium storing a self-diagnosis program for unattended heating station boiler equipment faults, wherein when the unattended heating station boiler equipment fault self-diagnosis program is executed by a processor, it implements the steps of the self-diagnosis method for unattended heating station boiler equipment faults as described above.

[0016] This invention provides a self-diagnostic method for unattended heating station boiler equipment. The method integrates multimodal sensor data, the physical characteristics of the thermal system, and dynamic heat transfer analysis to construct a closed-loop technical system encompassing thermal anomaly identification, risk evolution modeling, perception layout optimization, and intelligent diagnosis. This enables refined perception of the thermal state of the heating station boiler throughout its entire lifecycle and early fault warning. The generated three-dimensional thermal anomaly distribution map and thermal path network model accurately depict heat accumulation and propagation paths, improving the accuracy and interpretability of potential fault location. The fixed and dynamic perception point layout, optimized based on the importance of thermal risks, significantly enhances the coverage and adaptability of key area monitoring. Combined with thermal dynamic feature extraction and adaptive diagnostic algorithms, it effectively identifies abnormal heat transfer patterns, reduces false alarm rates, and supports fault warning and self-repair linkage responses, greatly improving the safety, reliability, and intelligent operation and maintenance level of the heating station boiler in unattended scenarios. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an embodiment of the self-diagnosis method for unattended heating station boiler equipment according to the present invention; Figure 2 This is a schematic diagram of the specific steps in obtaining a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data in one embodiment of the self-diagnosis method for unattended heating station boiler equipment of the present invention. Figure 3 This is a schematic diagram illustrating the specific steps of obtaining the full-cycle operating condition sensing point layout scheme in one embodiment of the self-diagnosis method for unattended heating station boiler equipment of the present invention. Figure 4 This is a structural block diagram of an embodiment of the self-diagnosis system for unattended heating station boiler equipment of the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] Reference Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the self-diagnosis method for unattended heating station boiler equipment according to the present invention, which presents an embodiment of the self-diagnosis method for unattended heating station boiler equipment according to the present invention.

[0021] In one embodiment, the self-diagnosis method for unattended heating station boiler equipment faults includes: Step S100: Collect thermal efficiency data, heat flow distribution data and environmental parameter data under different heating conditions through a multi-modal sensor array pre-arranged in the thermal system of the heating station boiler equipment, and obtain a thermal characteristic dataset of the thermal system.

[0022] The multimodal sensor array can be an integrated sensing unit composed of various types of sensors, used to simultaneously collect operational data from different physical dimensions of the boiler in the heating station. It can acquire multi-source heterogeneous data covering thermal efficiency, heat flux distribution, and environmental parameters, supporting comprehensive perception of the thermal state. In an exemplary embodiment, the multimodal sensor array may include, but is not limited to, one or more of the following: a temperature-pressure composite sensor, an infrared thermal imaging and thermocouple fusion array, and an acoustic-vibration-heat flux multi-parameter sensing module.

[0023] Step S200: Calculate the thermal gradient change between adjacent measuring points in the multimodal sensor array based on the thermal characteristic dataset, and analyze the accumulation and dissipation trend of heat in the system by combining the geometric characteristics of the thermal system and the thermodynamic parameters of the material, so as to obtain a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data.

[0024] The geometric characteristics of the thermal system can be the spatial topology and dimensional parameters of the internal structure of the boiler, including pipe routing, heat exchange surface layout, and cavity shape. These can provide spatial constraints for heat propagation path modeling and affect the accuracy of thermal gradient calculation and anomaly location. In one specific embodiment, the geometric characteristics of the thermal system can be obtained through equipment CAD models or 3D laser scanning reconstruction. Material thermodynamic parameters can be physical properties such as thermal conductivity, specific heat capacity, and coefficient of thermal expansion of the materials constituting the boiler components. These can determine the heat conduction, storage, and dissipation capabilities in different media and are the basic input for thermal anomaly evolution modeling. For example, material thermodynamic parameters can be obtained from material handbooks or laboratory testing and calibration. The 3D thermal anomaly distribution map can be a 3D visualization representation of the deviation of the internal spatial thermal state of the boiler from normal operating conditions, reconstructed based on multimodal sensor data and physical models. It can be used to intuitively present the spatial location and intensity level of abnormal areas such as local overheating and heat transfer deterioration. Furthermore, the 3D thermal anomaly distribution map can serve as the input basis for the thermal path network model, guiding the deployment of sensing points; its generation depends on the geometric characteristics of the thermal system and the material thermodynamic parameters. In one exemplary embodiment, the three-dimensional thermal anomaly distribution map may include, but is not limited to, a steady-state thermal anomaly cloud map, a transient thermal shock distribution map, and a thermal fatigue hotspot map under periodic load fluctuations.

[0025] Thermal risk accumulation rate data can be a dynamic indicator that quantifies the rate of increase of thermal anomalies in a specific area per unit time. It can be used to reflect the trend of thermal risk evolution over time and to identify potential faults that develop slowly in the early stages. In a specific embodiment, thermal risk accumulation rate data can be obtained by integral or differential analysis of the rate of change of thermal gradients in adjacent time periods. Combining the geometric characteristics of the thermal system and the thermodynamic parameters of materials to analyze the accumulation and dissipation trends of heat in the system can be achieved by coupling thermal gradient data collected by multimodal sensors with the system geometry and material properties to model and solve the heat conduction-convection-radiation composite equation, thus deducing the spatiotemporal evolution of heat. Furthermore, this operation can be carried out by constructing a transient thermal field simulation model using the finite element method and embedding measured boundary conditions for iterative correction, or by using a graph neural network to encode the geometric topology into node relationships and combining material parameters with weighted edge weights to simulate heat flow propagation. This can generate a thermal anomaly distribution with physical consistency, avoiding overfitting and uninterpretability of purely data-driven methods.

[0026] Step S300: Construct a thermal path network model of the thermal system and key equipment based on the three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data. Determine the locations of fixed and dynamic sensing points in the potential fault zone of the boiler in the heat station through the thermal path network model, and obtain the full-cycle operating condition sensing point layout scheme.

[0027] The thermal path network model can abstract the boiler into a graph structure composed of thermal nodes and thermal edges, describing the propagation path and impedance characteristics of heat from source to sink. It can be used to reveal the propagation path and key bottleneck nodes of thermal anomalies within the system, supporting the prediction of the scope of fault impact. For example, the thermal path network model can include, but is not limited to, steady-state heat conduction path diagrams, transient heat wave propagation networks, and multi-source coupled thermal interference topologies. Fixed sensing points can be permanently installed sensor monitoring locations in high-risk areas of the boiler, used for long-term stable acquisition of key thermal parameters. They can provide continuous, highly reliable benchmark monitoring data and are suitable for structurally fixed areas with a high probability of failure. In a specific embodiment, fixed sensing points can include, but are not limited to, measuring points in the high-temperature zone of the furnace outlet, coking-sensitive points at the inlet of the heat exchanger tube bundle, and wall temperature points in sections of the flue prone to ash accumulation. Dynamic sensing points can be movable / switchable sensing locations that are temporarily activated or adjusted based on real-time thermal risk assessment results. They can be used to enhance the monitoring flexibility and response speed of emerging anomaly areas under varying operating conditions. For example, dynamic sensing points may include, but are not limited to, the scanning focal area of ​​an infrared thermal imager, the pointing point of an adjustable thermocouple probe, and the focusing listening area of ​​an acoustic array.

[0028] Constructing a thermal path network model for the thermal system and key equipment can be achieved by identifying high-gradient regions as thermal nodes based on a 3D thermal anomaly distribution map, and establishing directed weighted edges according to the relationship between heat flow direction and impedance to form a network graph structure. Further, this operation can be performed by using a minimum thermal resistance path algorithm to extract the main heat flow channels and construct a sparse network, or by using a community detection algorithm to cluster thermal anomaly regions into subnets and connect inter-regional thermal bridges. This discretizes the continuous thermal field into a computable network model, facilitating path analysis and key node identification. Determining the locations of fixed and dynamic sensing points for potential fault areas in the heating station boiler can be done by classifying nodes in the thermal path network model according to their thermal risk importance, setting high-importance regions as fixed points and less important but time-varying regions as dynamic points. Further, this operation can be performed by using a multi-objective optimization algorithm to balance the number of fixed points and the switching frequency of dynamic points, or by dynamically adjusting the priority of sensing points based on matching historical fault databases with the current thermal risk map. This allows for optimized allocation of sensing resources, ensuring coverage while improving adaptability to sudden anomalies.

[0029] Step S400: According to the full-cycle operating condition sensing point layout plan, deploy operating condition monitoring devices in the potential fault area of ​​the boiler in the heat station to collect equipment operation data, extract thermal dynamic characteristics from the equipment operation data, identify abnormal heat transfer modes in the thermal system based on the thermal dynamic characteristics, obtain equipment fault self-diagnosis results, and trigger fault warning and self-repair based on the equipment fault self-diagnosis results.

[0030] Among them, thermal dynamic characteristics can be multi-dimensional feature vectors reflecting the time-varying characteristics of the heat transfer process under varying load conditions, such as thermal response delay, phase shift, and nonlinear oscillation. These characteristics can be used to characterize the degree to which the system deviates from the normal heat transfer law and serve as a criterion for identifying abnormal heat transfer patterns. In a specific embodiment, thermal dynamic characteristics can be extracted by performing time-frequency analysis, correlation modeling, or state-space reconstruction on equipment operating data. Abnormal heat transfer patterns can be thermal behavior patterns that significantly deviate from the design heat transfer mechanism under specific operating conditions, such as sudden increases in local thermal resistance, reversed heat flow, and heat transfer stagnation. These can be used as a direct basis for fault self-diagnosis, indicating potential equipment deterioration or abnormal operation. For example, abnormal heat transfer patterns can include, but are not limited to, heat transfer deterioration patterns, thermal short-circuit patterns, and thermal hysteresis patterns.

[0031] Identifying abnormal heat transfer patterns in a thermal system based on dynamic thermal features can be achieved by inputting the extracted dynamic thermal features into an adaptive diagnostic algorithm for operating conditions, comparing them with a normal pattern library, and detecting significantly deviating pattern categories. Furthermore, this operation can be further refined by using a Hidden Markov Model to model the thermal response sequence and identify abnormal state transition paths, or by constructing a Siamese neural network to distinguish between normal and abnormal heat transfer embeddings through contrastive learning. This allows for a shift from threshold alarms to mechanism-driven pattern recognition, reducing false alarms caused by fluctuations in operating conditions.

[0032] Taking early warning of coking during variable load heating as an example, the self-diagnosis method for unattended heating station boiler equipment faults in this embodiment can be as follows: during the stage of sudden drop in heating load, the multi-modal sensor array detects an abnormal decrease in heat flux density in a certain heat exchange tube bundle area. Combining the low thermal conductivity of the stainless steel material in this area with the geometry of the U-shaped bend, the system calculates that the local thermal resistance continues to rise, and generates a three-dimensional thermal anomaly distribution map showing that a low-temperature stagnation zone has appeared at this location. The thermal path network model identifies this node as a heat flow bottleneck, lists it as a high-risk area, and automatically activates nearby infrared dynamic sensing points for high-frequency scanning. The extracted thermal dynamic features show that the thermal response phase is severely lagging, and the diagnostic algorithm matches the "heat transfer deterioration mode", triggering a coking warning and linking the soot blowing device to start self-repair.

[0033] In one embodiment, reference Figure 2 Based on the thermal characteristic dataset, the thermal gradient changes between adjacent measuring points in the multimodal sensor array are calculated. Combined with the geometric characteristics of the thermal system and the thermodynamic parameters of the materials, the accumulation and dissipation trends of heat within the system are analyzed, resulting in a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data, including: Step S201: Remove environmental noise from the thermal efficiency data using adaptive Kalman filtering to obtain filtered thermal efficiency field data; Adaptive Kalman filtering, a recursive state estimation algorithm, adjusts filter parameters in real time based on system noise statistics. It can be used for dynamic denoising of thermal efficiency data, suppressing measurement fluctuations caused by environmental disturbances, and improving the signal-to-noise ratio of thermal field modeling input data. In this embodiment, the operating principle of adaptive Kalman filtering can be explained in context: using thermal efficiency data as the observation input, it uses an adaptive mechanism to estimate the covariance between process noise and observation noise online, dynamically adjusts the filter gain, and outputs smoothed thermal efficiency field data. Furthermore, adaptive Kalman filtering can include, but is not limited to, one or more of the following: Kalman filtering based on residual covariance adaptation, multi-model interactive adaptive Kalman filtering, and adaptive Kalman filtering with a forgetting factor. Eliminating environmental noise from thermal efficiency data through adaptive Kalman filtering can be achieved by updating noise statistics parameters using sliding window covariance estimation or by introducing a fuzzy logic controller to adjust the filter forgetting factor to adapt to sudden changes in operating conditions. This effectively suppresses random noise caused by environmental temperature fluctuations, electromagnetic interference, etc., and improves the stability of subsequent thermal gradient calculations.

[0034] Step S202: Calculate the thermal gradient vector and rate of change of adjacent measuring points in the multimodal sensor array based on the thermal efficiency field data and heat flow distribution data, and perform dynamic correction of the thermal field in combination with environmental parameter data to obtain a preliminary thermal distribution model; Step S203: Based on the preliminary thermal distribution model and the geometric characteristics of the thermodynamic system, establish the thermal equilibrium differential equation, and use the finite volume method to solve the thermal equilibrium differential equation to obtain the continuous thermal field distribution inside the thermodynamic system. The thermal equilibrium differential equation can be a partial differential equation describing the energy conservation relationship inside the furnace of a thermal power plant. It includes terms related to heat conduction, convection, internal heat sources, and boundary exchange. This equation can be used to couple discrete measurement data with the system's geometry and material properties to construct a continuous thermal field evolution model that conforms to physical laws. In an exemplary embodiment, the thermal equilibrium differential equation can be described in context as being derived based on the first law of thermodynamics, combined with the geometric characteristics of the thermodynamic system and the thermodynamic parameters of the materials. The finite volume method can be a numerical method that divides the computational domain into control volumes and solves the partial differential equation by integrating the conservation laws over each volume. It can be used to efficiently solve the thermal equilibrium differential equation and obtain a continuous thermal field distribution that satisfies mass-energy conservation. Exemplarily, the finite volume method can include, but is not limited to, one or more of the following: structured mesh finite volume method, unstructured mesh finite volume method, and implicit time-progression finite volume method.

[0035] By combining a preliminary thermal distribution model with the geometric characteristics of the thermodynamic system, a thermal equilibrium differential equation can be established. This can be achieved by using the preliminary thermal distribution model as the initial field, embedding the geometric boundary conditions of the thermodynamic system and the thermodynamic parameters of the materials, and constructing an energy conservation equation that includes spatial heat conduction, convective heat transfer, and internal heat source terms. This allows for the mapping from discrete sensor data to a physically consistent continuous thermal field, overcoming the sparsity limitations of sensors. Solving the thermal equilibrium differential equation using the finite volume method involves dividing the three-dimensional geometry of the furnace into finite control volumes, integrating the thermal equilibrium equation for each volume, constructing a system of algebraic equations, and iteratively solving them to obtain the overall temperature and heat flux distribution. Furthermore, the finite volume method can be further improved by using the SIMPLE algorithm to handle pressure-velocity-temperature coupling (if fluid-structure interaction is involved) or by employing multigrids to accelerate convergence and improve real-time performance. This allows for obtaining a high-fidelity continuous thermal field that satisfies both local and global energy conservation, supporting precise anomaly identification.

[0036] By identifying regions of abrupt changes in thermal gradient and regions of abnormal thermal flow from continuous thermal field and heat flow distribution data, and combining material thermodynamic parameters, the thermal risk accumulation index of each region of abrupt changes in thermal gradient and region of abnormal thermal flow is calculated, resulting in a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data.

[0037] The thermal gradient abrupt change region can be a spatial region in the thermal field where the thermal gradient at adjacent locations undergoes a significant step or discontinuous change. It can be used to indicate high-risk locations where local heat transfer deterioration, coking, or material degradation may occur. In a specific embodiment, the thermal gradient abrupt change region may include, but is not limited to, one or more of the following: radial thermal gradient abrupt change region, axial heat flow discontinuity region, and interfacial thermal resistance jump region. The heat flow anomaly region can be a region where the measured heat flow direction or intensity deviates significantly from the expected value under design conditions. It can be used to reflect the location of abnormal heat transfer phenomena such as thermal short circuits, blockages, or flow stagnation. For example, the heat flow anomaly region may include, but is not limited to, one or more of the following: reverse heat flow region, heat flow attenuation region, and heat flow concentration vortex region. The thermal risk accumulation index can be a quantitative risk indicator calculated by comprehensively considering the degree of thermal gradient abrupt change, the intensity of heat flow anomaly, and material thermodynamic parameters (such as thermal conductivity decay rate). It can be used to characterize the probability and speed at which each anomaly region develops into a failure under current operating conditions, supporting the generation of thermal risk accumulation rate data. In this embodiment, the thermal risk accumulation index can be explained in context as being obtained by calculating it through a weighted fusion of thermal anomaly characteristics and material heat resistance parameters.

[0038] Identifying regions of abrupt changes in thermal gradients and anomalies in thermal flow from continuous thermal field and heat flow distribution data can be achieved by calculating the full-field thermal gradient tensor and heat flow vector, and detecting significant deviations in amplitude, direction, or continuity through threshold settings or cluster analysis. Furthermore, this identification can be implemented by utilizing Canny edge detection to identify thermal gradient abrupt boundary conditions or by extracting anomalous patterns in the heat flow vector field using principal component analysis (PCA), thereby accurately locating potential fault initiation points and avoiding the limitations of relying solely on threshold values ​​for a single measurement point. Calculating the cumulative thermal risk index for each region of abrupt thermal gradient changes and anomalies in thermal flow, combined with material thermodynamic parameters, can be achieved by weighting and fusing the rate of change of thermal gradients and heat flow deviation in anomalies with parameters such as material thermal conductivity and heat capacity to calculate a comprehensive index reflecting the rate of thermal damage accumulation. Further, this calculation can be implemented by using the Analytic Hierarchy Process (AHP) to determine the weights of each parameter or by introducing an Arrhenius-type thermal aging model to correlate the index with remaining life, thereby enabling differentiated assessments of thermal risk in regions with different materials and improving the physical rationality of early warnings.

[0039] For example, in scenarios involving denoising and anomaly identification of thermal efficiency data under high humidity conditions, the self-diagnosis method for unattended heating station boiler equipment in this embodiment can be as follows: In the high humidity environment of the southern heating season, the thermal efficiency sensor generates high-frequency noise due to interference from condensate; adaptive Kalman filtering adjusts the observation noise covariance in real time to output a smooth thermal efficiency field; a thermal balance differential equation containing the latent heat of phase change term is established in combination with the geometric structure of the U-shaped flue, and a continuous thermal field is obtained by solving it using the finite volume method; the system identifies a sudden change in thermal gradient and a deflection of heat flow direction on the outside of a certain bend, calculates a high thermal risk accumulation index in combination with the low thermal conductivity of the carbon steel material at that location, generates a three-dimensional thermal anomaly distribution map marked as a red high-risk area, and updates the thermal risk accumulation rate data to trigger subsequent redeployment of sensing points.

[0040] In one embodiment, the thermal gradient vector and rate of change of adjacent measuring points in the multimodal sensor array are calculated based on thermal efficiency field data and heat flow distribution data, and the thermal field is dynamically corrected by combining environmental parameter data to obtain a preliminary thermal distribution model, including: Based on the thermal efficiency field data, the thermal parameter values ​​of each measuring point are spatially mapped in the three-dimensional spatial coordinate system of the multimodal sensor array to obtain the initial thermal spatial distribution map. The thermal gradient vector is calculated for the thermal parameter values ​​of adjacent measuring points in the initial thermal spatial distribution diagram. The thermal change rate of each measuring point along the principal axis of the thermodynamic system is calculated using the higher-order finite difference method to obtain the thermal gradient vector field. The initial thermal spatial distribution map is a geometric representation of the thermal efficiency field data in three-dimensional space, providing a structured input basis for subsequent gradient calculations. The thermal gradient vector field can be a vector field composed of thermal parameter gradients at various spatial locations within the boiler of the heating station. It characterizes the direction and intensity changes of heat in three-dimensional space and can be used to accurately describe the spatial evolution trend of local thermal anomalies, providing a high-resolution basis for identifying heat transfer deterioration or abrupt changes in thermal resistance. In this embodiment, the thermal gradient vector field can be part of the intermediate thermal state representation generated by coupling the thermal gradient vector field with a dynamically compensated heat flow-thermal efficiency relationship model. For example, the thermal gradient vector field can include, but is not limited to, one or more of the following: radial thermal gradient vector field, axial heat flow gradient field, and multi-scale fused gradient field. The higher-order difference method can be a numerical calculation method that uses second-order or higher difference schemes to approximate partial derivatives, used to improve the accuracy of thermal gradient estimation. It can be used to reduce gradient calculation errors caused by sensor sparsity or noise, and enhance sensitivity to weak thermal anomalies.

[0041] The thermal gradient vector is calculated from the thermal parameter values ​​of adjacent measuring points in the initial thermal spatial distribution map. A higher-order finite difference method is used to calculate the rate of thermal change at each measuring point along the principal axis of the thermodynamic system. This can be achieved by constructing a higher-order finite difference scheme in a three-dimensional coordinate system using the thermal parameter values ​​of neighboring measuring points, calculating partial derivatives in each direction, and synthesizing the gradient vector. A higher-precision scheme is used, particularly along the principal axis, to extract the rate of thermal change. Furthermore, this operation can be mitigated by using an adaptive order finite difference scheme in non-uniformly distributed regions, or by combining spline interpolation preprocessing with higher-order finite difference to alleviate the sparsity effect. This improves the spatial resolution and numerical stability of the thermal gradient estimation, effectively capturing slowly developing local thermal anomalies.

[0042] Based on the correspondence between heat flow distribution data and thermal gradient vector field, the equivalent thermal resistance coefficient of each region of the thermodynamic system is calculated using the robust regression method, and a heat flow-thermal efficiency relationship model is established. The equivalent thermal resistance coefficient can be an equivalent physical parameter that comprehensively maps complex degradation effects such as material aging, ash accumulation, and coking into the ratio of heat flux to thermal gradient. It can be used to quantify the impact of latent faults on heat transfer performance, allowing invisible degradation states to be perceived and assessed by the model. In a specific embodiment, the equivalent thermal resistance coefficient may include, but is not limited to, the interfacial contact thermal resistance coefficient, the equivalent thermal resistance of the deposited layer, and the reciprocal of the equivalent thermal conductivity of the composite material. Robust regression is a statistical regression method that is insensitive to outliers. It is used to stably estimate the relationship between variables when outlier data exists. It can reliably fit the physical relationship between heat flux and thermal gradient even when some sensors are interfered with or there are sudden local disturbances. The heat flux-thermal efficiency relationship model can be a physical mapping model between heat flux density and thermal efficiency (or temperature gradient) established based on the equivalent thermal resistance coefficient. It can be used to reflect the current heat transfer capacity of the system and serve as a benchmark for judging whether it deviates from normal operating conditions. Furthermore, the heat flux-thermal efficiency relationship model can accept dynamic compensation of environmental parameters, and its output can be used to construct a preliminary thermal distribution model.

[0043] Based on the correspondence between heat flux distribution data and the thermal gradient vector field, the equivalent thermal resistance coefficient of each region of the thermodynamic system is calculated using robust regression. This can be achieved by using heat flux density as the dependent variable and thermal gradient amplitude as the independent variable, performing robust regression on each subset of regions to fit a local equivalent thermal resistance coefficient that is resistant to outlier interference. Furthermore, this operation can be achieved by using sliding window local robust regression to reconstruct a spatially continuous thermal resistance field, or by introducing regularization constraints to ensure a smooth transition of thermal resistance coefficients between adjacent regions. This allows factors that cannot be directly measured, such as material degradation and dust accumulation, to be transformed into modelable thermal resistance parameters, enhancing the physical interpretability of the model.

[0044] By dynamically compensating the heat flow-thermal efficiency relationship model using environmental parameter data, a preliminary thermal distribution model is obtained.

[0045] The initial thermal distribution model can be an intermediate thermal state representation formed by integrating the thermal gradient vector field, equivalent thermal resistance coefficient, and the heat flow-thermal efficiency relationship after environmental compensation. This model can serve as a high-fidelity input basis for subsequently constructing continuous thermal fields and three-dimensional thermal anomaly distribution maps. Dynamic compensation of the heat flow-thermal efficiency relationship model using environmental parameter data can be achieved by using parameters such as ambient temperature, humidity, and wind speed as covariates to correct the bias terms or gain coefficients in the model, making it adaptable to external disturbances. Furthermore, this operation can be achieved by constructing a lookup table or neural network mapping relationship between environmental parameters and thermal resistance correction factors, or by using an online recursive least squares method to update the compensation parameters in real time. This can eliminate the interference of environmental fluctuations on the judgment of the thermal state and improve the model's generalization ability under varying operating conditions.

[0046] For example, in the scenario of thermal field modeling and anomaly identification during a winter cold wave, the self-diagnosis method for unattended heating station boiler equipment in this embodiment can be as follows: Under the invasion of strong cold air, the sudden drop in ambient temperature causes the thermal efficiency sensor readings to drift; the system first performs spatial mapping of the thermal efficiency field to generate an initial thermal spatial distribution map, and uses the fourth-order central difference method to calculate a high-precision thermal gradient vector field; combined with infrared heat flow meter data, RANSAC robust regression is used to identify an abnormally high equivalent thermal resistance coefficient in the flue bend area, indicating severe ash accumulation; then, real-time wind speed and outdoor temperature are used to dynamically compensate the heat flow-thermal efficiency relationship model to eliminate false heat loss signals caused by the cold wave, and finally output a robust preliminary thermal distribution model, providing a reliable initial field for subsequent heat balance equation solving.

[0047] In one embodiment, reference Figure 3 Based on the three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data, a thermal path network model of the thermal system and key equipment is constructed. The location of fixed and dynamic sensing points in the potential fault zone of the boiler in the heating station is determined through this thermal path network model, resulting in a full-cycle operating condition sensing point layout scheme, including: Step S301: Perform density cluster analysis on thermal anomaly points in the thermal system based on the three-dimensional thermal anomaly distribution map. Based on the results of the cluster analysis, merge similar thermal anomaly points into thermal risk zones. Calculate the heat transfer direction and attenuation rate of each thermal risk zone based on the thermal risk accumulation rate data to obtain a thermal risk impact map. The thermal risk impact map can be a spatiotemporal evolution characterization map of thermal risk generated by fusing spatial clustering results of thermal anomalies with dynamic characteristics of heat transfer (direction, decay rate). It can be used to integrate heat dissipation anomaly points into risk region units with physical propagation significance, supporting subsequent network modeling and path analysis. In this embodiment, the thermal risk impact map can be based on the thermal risk zones output by density clustering, combined with thermal risk accumulation rate data to calculate the heat flow vector and exponential decay coefficient of each region, and mapped to the system geometric space. For example, the thermal risk impact map can include, but is not limited to, one or more of the following: steady-state heat diffusion impact map, transient thermal shock wavefront map, and thermal fatigue impact domain under periodic load modulation.

[0048] Density clustering analysis of thermal anomaly points within a thermal system based on a 3D thermal anomaly distribution map can be performed. This can be achieved by treating the anomaly points in the 3D thermal anomaly distribution map as a spatial point set, and using a density clustering algorithm (such as DBSCAN) to identify high-density neighboring regions and merge them into connected thermal risk zones. Furthermore, this operation can be achieved by directly applying DBSCAN in 3D Euclidean space with thermal anomaly intensity as an additional density weight, or by projecting thermal anomaly points onto a device surface mesh and clustering them based on geodesic distance to adapt to complex curved surface structures. This can eliminate interference from isolated noise points, form physically continuous thermal risk zone units, and improve the robustness of subsequent modeling.

[0049] Step S302: Construct a weighted graph-based thermal path network model based on the thermal risk impact map and the geometric topological relationship between the thermal system and key equipment, where nodes represent thermal risk zones, edges represent heat transfer paths, and weights represent dynamic coefficients of heat conduction. A weighted graph-based thermal path network model can be constructed based on the thermal risk impact map and the geometric topological relationships between the thermal system and key equipment. This can be achieved by using thermal risk zones as graph nodes, establishing directed edges based on the heat transfer direction in the thermal risk impact map, and using the dynamic coefficient of heat conduction (determined by material properties and geometric spacing) as the edge weights. In an exemplary embodiment, this operation can be achieved by determining the edge weights using the equivalent method of thermal resistance networks, combined with Fourier's law and geometric cross-sectional area calculations, or by automatically learning the thermal coupling strength between nodes and dynamically adjusting the edge weights through a graph attention mechanism. This allows the continuous thermal field to be abstracted into a computable graph structure, preserving the physical laws of heat propagation while supporting efficient path analysis.

[0050] Step S303: Perform critical path analysis on the heat path network model, calculate the heat transfer time window and heat attenuation threshold from each heat risk zone to the core components of the heating station boiler, and sort each heat risk zone according to the degree of impact on heating continuity based on the heat transfer time window and heat attenuation threshold to obtain a heat risk importance ranking table. The thermal risk importance ranking table can be a quantified priority list generated based on the impact of thermal risk zones on heating continuity (determined by the heat transfer time window and heat attenuation threshold). This list can provide a basis for decision-making regarding the deployment of sensing points, enabling differentiated allocation of monitoring resources according to risk levels. In one specific embodiment, the thermal risk importance ranking table can be generated by performing critical path analysis on a thermal path network model, calculating the thermal response delay and energy attenuation degree from each thermal risk zone to core components, and ranking them after comprehensive scoring. For example, the thermal risk importance ranking table may include, but is not limited to, one or more of the following: a core component thermal sensitivity ranking table, a thermal fault propagation timeliness grading table, and a multi-objective thermal risk comprehensive assessment table.

[0051] Critical path analysis of heat propagation in the thermal path network model is performed to calculate the heat transfer time window and heat attenuation threshold from each thermal risk zone to the core components of the boiler in the heat station. This can be achieved by performing a shortest thermal response path search on a weighted graph, and then using the heat conduction differential equation to inversely deduce the time required for heat to propagate from the source region to the core components and the proportion of remaining energy. Furthermore, this operation can be achieved by using a variant of Dijkstra's algorithm to convert the edge weights into a composite index of heat propagation delay and attenuation factor, or by using Monte Carlo simulation to perform random walks on the graph to statistically analyze the time distribution and intensity attenuation of heat signals reaching the core components. This allows for the quantification of the potential threat level of each thermal risk zone to the system's critical functions, providing a physical basis for risk prioritization.

[0052] Step S304: Determine the locations of fixed and dynamic sensing points in the potential fault zone of the boiler in the heating station according to the thermal risk importance ranking table, and obtain the full-cycle operating condition sensing point layout scheme.

[0053] The locations of fixed and dynamic sensing points for potential fault areas in thermal power plant boilers are determined based on a thermal risk importance ranking table. This can be achieved by designating high-ranking and stable thermal risk areas as fixed sensing points, and areas with moderate ranking but time-varying migration characteristics as the coverage area for dynamic sensing points. In an exemplary embodiment, this operation can be implemented by setting an importance threshold; areas exceeding the threshold are deployed as fixed points, while the rest are included in a dynamic scanning scheduling queue. Alternatively, a hybrid "fixed + dynamic" monitoring strategy can be used for high-importance but sporadic risk areas in conjunction with equipment maintenance cycles. This allows for a shift in sensing layout from static full coverage to risk-driven, adaptive focusing, improving monitoring efficiency and response sensitivity.

[0054] Taking the monitoring of progressive heat transfer deterioration caused by flue ash accumulation as an example, the self-diagnosis method for unattended heating station boiler equipment faults in this embodiment can be as follows: the three-dimensional thermal anomaly distribution map shows that multiple low-temperature points are clustered at the flue bend, and density clustering merges them into a single thermal risk zone; the thermal risk accumulation rate data shows that the heat decay rate in this area is accelerating, and the thermal risk impact map marks it as a heat flow blocking area; after the thermal path network model is constructed, the critical path analysis shows that the heat transfer time window from this area to the economizer exceeds the safety threshold, the heat decay reaches the critical value, and it is listed as a high-importance risk area; accordingly, the system deploys fixed temperature-infrared composite sensing points on the outer wall of the bend, and sets dynamic infrared scanning points in the downstream straight pipe section; when the thermal dynamic characteristics show that the thermal response is further delayed, the diagnostic algorithm identifies it as a heat transfer deterioration mode, triggers an early warning and starts soot blowing self-repair.

[0055] In one embodiment, the locations of fixed and dynamic sensing points in the potential fault zone of the boiler in the heating station are determined according to a thermal risk importance ranking table, resulting in a full-cycle operating condition sensing point layout scheme, including: The thermal risk zones are classified according to the thermal risk importance ranking table, and the dynamic coverage radius of each thermal risk zone is calculated to obtain the thermal risk zone classification table. The thermal risk classification table can be a structured data table that classifies thermal risk areas according to their impact level based on a thermal risk importance ranking table, and assigns a dynamic coverage radius to each level. It can be used to provide a classification basis for the differentiated deployment of fixed and dynamic sensing points, enabling precise allocation of monitoring resources according to risk level. In this embodiment, the thermal risk classification table can divide the thermal risk importance ranking table into intervals (e.g., high / medium / low), and combine historical operating data to statistically analyze the maximum spatial disturbance range of each level of area under varying loads, defining it as the dynamic coverage radius. For example, the thermal risk classification table can include, but is not limited to, one or more of the following: a high-risk compact coverage table, a medium-risk flexible coverage table, and a low-risk sparse coverage table.

[0056] The thermal risk zones are classified according to a thermal risk importance ranking table, and the dynamic coverage radius of each level is calculated. This can be achieved by discretizing the continuous scores in the thermal risk importance ranking table into several levels, and statistically analyzing the spatial offset range of each level under varying operating conditions based on historical operational data, taking the upper bound of the statistics as the dynamic coverage radius. Furthermore, this operation can be implemented by using the quantile method to classify risk levels and taking the maximum offset distance within the 95% confidence interval as the dynamic coverage radius, or by using a clustering algorithm to group the joint features of importance scores and offsets, automatically determining the classification boundaries and corresponding radii. This establishes a mapping relationship between risk levels and spatial uncertainty, providing a quantitative basis for subsequent sensing point coverage strategies.

[0057] A working condition stability analysis was performed on the high-risk areas in the thermal risk classification table. The root mean square error of the position deviation of each high-risk area under different heating loads was calculated, and the high-risk areas were divided into steady-state risk areas and transient risk areas based on the root mean square error of the position deviation. The steady-state risk zone can be a high-risk area where the root mean square error of the positional deviation is less than a set threshold under different heating loads. Its spatial distribution exhibits long-term stability and can be used as a target area for deploying fixed sensing points, suitable for long-term, high-reliability monitoring. In an exemplary embodiment, the steady-state risk zone can be derived from the high-risk area through operational stability analysis and serves as a direct input to the fixed sensing point layout scheme. For example, the steady-state risk zone may include, but is not limited to, one or more of the following: the high-temperature stable zone at the furnace outlet, the fixed coking hotspot at the economizer inlet, and the constant ash accumulation zone at the flue bend.

[0058] Transient risk zones can be high-risk areas where the root mean square error (RMSE) of their location deviation exceeds a set threshold under different heating loads. Their spatial location dynamically migrates with changing operating conditions and can be used as target areas for dynamic sensing point scheduling. Adaptive coverage requires reconfigurable or adjustable sensing methods. In one specific embodiment, transient risk zones and steady-state risk zones are mutually exclusive, together constituting a complete classification of high-risk zones and driving the generation of dynamic sensing point layout schemes. For example, transient risk zones may include, but are not limited to, one or more of the following: thermal shock migration zones caused by load switching, thermal stress fluctuation zones during start-up and shutdown, and localized backflow hotspots under variable flow conditions. Performing operating condition stability analysis on high-risk zones, calculating the RMSE of their location deviation, and classifying them into steady-state and transient risk zones can be achieved by tracking the changes in the center coordinates of high-risk zones under multiple typical heating load conditions, calculating the RMSE of their location sequence, and performing binary classification with a preset threshold as the boundary. Furthermore, this operation can achieve online stability judgment by updating RMSE in real time based on a sliding window, or introduce a time series similarity metric (such as DTW) to replace RMSE to improve sensitivity to nonlinear offset patterns, thereby enabling dynamic characteristic identification of high-risk areas and supporting differentiated design of fixed and dynamic sensing strategies.

[0059] Fixed sensing points are set according to the location of the steady-state risk zone. The minimum coverage set algorithm is used to optimize the number of fixed sensing points so that each steady-state risk zone is covered by at least one fixed sensing point, while maximizing the sensing redundancy, thus obtaining a fixed sensing point layout scheme. A minimum coverage set algorithm is employed to optimize the number of fixed sensing points, ensuring that each steady-state risk area is covered by at least one fixed sensing point while maximizing sensing redundancy. This can be achieved by treating the steady-state risk area as the target set to be covered, and the candidate locations of the fixed sensing points as the sensor installation points, constructing a set coverage problem. The goal is to minimize the number of points while satisfying the full coverage constraint, and prioritize retaining configurations with high overlap coverage in the solution set to improve redundancy. Furthermore, this operation can be achieved by using a greedy approximation algorithm to solve for the minimum coverage set (selecting the candidate point that covers the most uncovered areas each time), or by constructing an integer linear programming model (the objective function is to minimize the number of points, with constraints including coverage requirements and a lower limit on redundancy). This approach reduces hardware costs while ensuring monitoring reliability and enhances system fault tolerance through redundant design.

[0060] Based on the movement trajectory and frequency characteristics of the transient risk zone, dynamic sensing points are set up in the transient risk zone to obtain a dynamic sensing point layout scheme. The fixed sensing point layout scheme and the dynamic sensing point layout scheme are then integrated to form a full-cycle working condition sensing point layout scheme.

[0061] Setting up dynamic sensing points based on the movement trajectory and frequency characteristics of transient risk zones can involve analyzing the spatial trajectory sequence and frequency of occurrence of transient risk zones in historical operating conditions. Based on this, the scanning path, dwell time, and activation priority of dynamic sensing devices (such as rotatable infrared probes or sensors mounted on mobile robots) can be planned. Furthermore, this operation can be achieved by generating an optimal sensing scheduling strategy through risk zone transition probabilities modeled using Markov chains, or by using reinforcement learning to train a dynamic sensing controller to autonomously adjust the scanning strategy with the goal of maximizing the anomaly capture rate. This enables proactive tracking and efficient coverage of time-varying high-risk areas, avoiding monitoring blind spots caused by static point deployment.

[0062] Taking the monitoring of thermal stress migration at the economizer inlet under variable load operation as an example, the self-diagnosis method for unattended boiler equipment faults in this embodiment can be as follows: the economizer inlet is listed as a high-risk area in the thermal risk importance ranking table; the operating condition stability analysis shows that it shifts upstream by 15cm under low load and retreats under high load, and the RMSE exceeds the threshold, so it is classified as a transient risk area; based on its periodic migration trajectory in the 20% to 80% load range, the system configures a horizontally movable infrared thermal imager as a dynamic sensing point, predicts the position according to the load command and stays there in advance; at the same time, the high-temperature area at the adjacent furnace outlet is classified as a steady-state risk area because of its stable position, and two fixed thermocouple points are selected from three candidate positions through the minimum coverage set algorithm to achieve double coverage; when the dynamic sensing point detects a sudden change in thermal gradient, it confirms the abnormal heat transfer mode by combining the fixed point data, triggers an early warning and starts the water-side flow regulation self-repair.

[0063] In one embodiment, operating condition monitoring devices are deployed in the potential fault zone of the boiler in the heating station according to the full-cycle operating condition sensing point layout scheme to collect equipment operation data. Thermal dynamic characteristics are extracted from the equipment operation data, and abnormal heat transfer modes in the thermal system are identified based on these characteristics to obtain equipment fault self-diagnosis results. Fault warning and self-repair triggering are then performed based on these self-diagnosis results, including: Based on the full-cycle operating condition sensing point layout plan, operating condition monitoring devices are deployed in the potential fault area of ​​the thermal station boiler to collect equipment operation data. Empirical Mode Decomposition (EMD) can be an adaptive signal processing method used to decompose nonlinear, non-stationary time series into several intrinsic mode functions (IMFs) and residual terms. It can be used to separate steady-state and transient components reflecting thermal response characteristics at different time scales from equipment operation data, providing a foundation for dynamic feature extraction. In this embodiment, EMD is explained in context: the original equipment operation data is input into the method, and it is adaptively decomposed into multiple IMFs. Low-frequency IMFs and residuals are considered as steady-state components, while high-frequency IMFs are considered as transient components. For example, EMD can include, but is not limited to, one or more of ensemble empirical mode decomposition (EEMD), complementary ensemble empirical mode decomposition (CEEMDAN), and noise-assisted empirical mode decomposition.

[0064] Empirical mode decomposition is used to extract multi-scale features from equipment operation data, separating the steady-state and transient components of the equipment operation data to obtain thermal dynamic feature vectors. Among them, the thermal dynamic eigenvector can be a multi-dimensional vector formed by feature engineering of the steady-state and transient components obtained from empirical mode decomposition. It characterizes the dynamic behavior of the heat transfer process and can be used to quantify the degree of deviation of the system's thermal response under varying operating conditions. Furthermore, the thermal dynamic eigenvector can be explained in context as follows: it is generated by empirical mode decomposition and used to construct the thermal state matrix; its structural design needs to match the feature dimensions of the normal thermal mode benchmark library.

[0065] The thermal state matrix of the thermal system is constructed based on the thermal dynamic feature vector, and the model is matched with a pre-set normal thermal model benchmark library to obtain the thermal anomaly index. The thermal state matrix can be a structured data matrix composed of thermal dynamic feature vectors as rows or columns, reflecting the overall thermal behavior of the boiler in a specific time period. It can provide a standardized input format for pattern matching and support similarity measurement with a normal thermal mode benchmark library. In an exemplary embodiment, the thermal state matrix can be described in context as being obtained by organizing and arranging the thermal dynamic feature vectors according to time windows or operating conditions. The normal thermal mode benchmark library can be a set of typical thermal behavior patterns constructed based on historical normal operating data and the physical mechanisms of the thermal system. It includes standard thermal response characteristics under different loads and environmental conditions and can be used as a reference system for pattern matching to identify whether the current thermal state deviates from the normal range. In a specific embodiment, the normal thermal mode benchmark library may include, but is not limited to, one or more of the following: steady-state full-load benchmark mode, variable load transition benchmark mode, and low-temperature start-up benchmark mode.

[0066] Constructing a thermal state matrix of a thermal system based on thermal dynamic feature vectors and performing pattern matching with a pre-defined normal thermal mode benchmark library can be achieved by arranging multiple thermal dynamic feature vectors within the same operating window into a matrix according to time or spatial dimensions, calculating their similarity to various normal modes in the benchmark library, and outputting the anomaly index corresponding to the minimum distance. Furthermore, this operation can be achieved by using Dynamic Time Warping (DTW) to align thermal state sequences of different lengths before matching, or by using autoencoder reconstruction error as an anomaly index to replace explicit pattern library matching. This allows for a transition from point-level features to system-level states, enhancing the ability to identify complex and progressive faults.

[0067] A multi-dimensional correlation analysis was conducted on the thermal anomaly index and heating condition parameters. An anomaly mode was classified using a fault self-diagnosis algorithm based on the characteristics of the thermal system. The continuous heating guarantee risk assessment index of the equipment was calculated to obtain the equipment fault self-diagnosis results. Fault warning and self-repair triggering were carried out based on the equipment fault self-diagnosis results.

[0068] Among them, the fault self-diagnosis algorithm based on the characteristics of the thermal system can be an intelligent diagnostic model that integrates the geometric structure, material properties, and heat transfer law constraints of the heating station boiler. This model is used to verify and classify the physical consistency of abnormal thermal modes, improving the accuracy and interpretability of anomaly classification and avoiding physically unreasonable outputs from purely data-driven models. For example, this algorithm can include, but is not limited to, one or more of the following: a diagnostic model embedding the heat conduction equation into a graph neural network; a Bayesian inference model based on thermal resistance network topology; and an attention mechanism classifier incorporating CFD prior knowledge. The equipment continuous heating guarantee risk assessment index can be a quantitative indicator calculated by comprehensively considering the degree of thermal anomaly, fault type, the impact weight of key components, and current heating demand. It characterizes the risk level of the equipment's ability to maintain continuous heating over a future period and can provide an actionable quantitative basis for operation and maintenance decisions, supporting graded early warning and self-repair strategy triggering. In a specific embodiment, the acquisition method of this index can be explained in context: by weighted fusion of anomaly classification results and operating parameters, a risk value in the 0-1 range is generated through normalization mapping.

[0069] A multi-dimensional correlation analysis is conducted on the thermal anomaly index combined with heating condition parameters. A fault self-diagnosis algorithm based on thermal system characteristics is used to classify anomaly modes and calculate the continuous heating guarantee risk assessment index. This can be achieved by inputting the thermal anomaly index along with operating parameters such as current load rate, outdoor temperature, and water supply pressure into the fault self-diagnosis algorithm, outputting the fault category and corresponding risk weight, and then weighting and synthesizing the continuous heating guarantee risk assessment index. Furthermore, this operation can be achieved by using a multi-task learning framework to simultaneously output fault type and risk index, or by introducing a fuzzy logic system to integrate expert rules and data-driven results to generate risk assessments. This transforms the diagnostic results from qualitative anomaly judgments into quantitative operational risk assessments, supporting differentiated early warning and self-repair strategies.

[0070] Taking the early diagnosis of the slow decline in heat exchange efficiency of a heating station boiler during a cold wave as an example, the self-diagnosis method for unattended heating station boiler equipment in this embodiment can be as follows: under continuous low temperature weather, the wall temperature data collected by fixed sensing points is separated into a slowly decaying steady-state component and a weakly oscillating transient component through empirical mode decomposition; the thermal dynamic feature vector shows that the phase lag of the thermal response in the heat exchange tube bundle area gradually increases; the matching degree between the constructed thermal state matrix and the "variable load transition benchmark mode" continues to decrease, and the thermal anomaly index increases; combined with the current high load rate and low return water temperature, the fault self-diagnosis algorithm identifies that ash accumulation leads to heat transfer deterioration, and calculates the equipment continuous heating guarantee risk assessment index as 0.78, triggering a medium-level early warning and starting the pulse soot blowing self-repair program.

[0071] In one embodiment, a multi-dimensional correlation analysis is performed on the thermal anomaly index combined with heating condition parameters. A fault self-diagnosis algorithm based on the characteristics of the thermal system is used to classify anomaly modes and calculate the equipment continuous heating guarantee risk assessment index to obtain equipment fault self-diagnosis results. Based on these results, fault warnings and self-repair triggers are implemented, including: A time-series feature space is constructed by combining three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data with historical heating condition parameters. Based on the structural characteristics of the thermal system, the data points in the feature space are subjected to condition-adaptive clustering to form a condition-driven benchmark fault mode library. The temporal feature space can be a multi-dimensional dynamic representation space constructed from a three-dimensional thermal anomaly distribution map, thermal risk accumulation rate data, and historical heating condition parameters. It describes the thermal state evolution trajectory of the boiler in the heating station over time. The temporal feature space can provide a unified semantic coordinate framework for condition-adaptive clustering and anomaly trajectory localization, enhancing the adaptability of pattern recognition to varying operating conditions. In an exemplary embodiment, the temporal feature space may include, but is not limited to, one or more of the following: steady-state load-dominated feature subspace, transient start-stop disturbance feature subspace, and multi-source coupled thermal interference feature subspace. The condition-driven benchmark fault mode library can be a set of typical fault evolution trajectories formed by condition-adaptive clustering in the temporal feature space. Each pattern is associated with specific heating conditions and the structural response characteristics of the thermal system. The condition-driven benchmark fault mode library can serve as a reference for dynamic pattern matching, supporting physically consistent classification of real-time anomaly trajectories. Furthermore, the operating condition-driven benchmark fault mode library can be generated from a temporal feature space and used to calculate the dynamic time warping distance with the real-time trajectory; its clustering process depends on the structural characteristics of the thermal system. For example, the operating condition-driven benchmark fault mode library may include, but is not limited to, one or more of the following: low-load ash accumulation evolution mode, high-load local overheating mode, and variable-load thermal stress fatigue mode.

[0072] A time-series feature space is constructed by combining three-dimensional thermal anomaly distribution maps and thermal risk accumulation rate data with historical heating condition parameters. Based on the structural characteristics of the thermal system, data points in the feature space are then subjected to condition-adaptive clustering to form a condition-driven baseline fault mode library. This can be achieved by aligning the three-dimensional thermal anomaly distribution, thermal risk accumulation rate, and corresponding operating condition parameters from historical operating data over time and embedding them into a unified feature space. Based on the structural constraints of the thermal system, clustering boundaries are constrained, and an adaptive clustering algorithm is used to divide fault evolution clusters. Furthermore, this operation can be further improved by introducing a graph regularization term to smooth the clustering results between adjacent nodes in the thermal path network, or by employing an online clustering mechanism to continuously update the mode library with new operating data. This allows for the generation of a fault mode library strongly coupled with the physical structure and operating conditions, improving the physical consistency and generalization ability of subsequent matching.

[0073] After combining the thermal anomaly index with real-time heating condition parameters, the trajectory is located in the constructed time-series feature space, and the dynamic time warping distance is calculated with each mode in the benchmark fault mode library. The anomaly mode with the smallest distance is selected as the fault mode classification result. Dynamic time warping distance (VTW) can be used as a similarity metric to measure the minimum cumulative distance between two time series of unequal durations under nonlinear time alignment. VTW can be used to achieve accurate matching between real-time thermal anomaly trajectories and baseline fault modes under flexible time axis alignment, overcoming the time series offset problem caused by operating condition fluctuations. In a specific embodiment, VTW can be solved using a dynamic programming algorithm to find the cumulative distance under the optimal alignment path.

[0074] After combining thermal anomaly indices with real-time heating condition parameters, trajectory localization is performed in the constructed time-series feature space. Dynamic time warping distances are then calculated between this trajectory and each mode in the benchmark fault mode library. The anomaly mode with the smallest distance is selected as the fault mode classification result. This can be achieved by concatenating the thermal anomaly sequence within the current window with real-time condition parameters to form a trajectory vector, which is then mapped to a sequence of trajectory points in the time-series feature space. Dynamic time warping distances are then calculated between this sequence and each mode trajectory in the benchmark fault mode library, and the mode corresponding to the smallest distance is selected as the classification result. Furthermore, this operation can be achieved by normalizing the trajectory to eliminate dimensional differences, or by using weighted dynamic time warping to assign higher matching weights to key time periods. This enables robust identification of unsteady, variable-rate thermal anomaly evolution, avoiding misclassification caused by operating condition switching.

[0075] Based on the failure mode classification results and the thermal path network model, digital twin simulation of the heat transfer process in the thermal system is performed to predict the evolution trend of thermal parameters of key equipment. The continuous heating guarantee risk assessment index is calculated according to the thermal tolerance threshold of the equipment, and the equipment fault self-diagnosis results are obtained. Fault warning and self-repair triggering are carried out based on the equipment fault self-diagnosis results.

[0076] Digital twin simulation can be a virtual mapping of a thermal system based on a thermal path network model and the current failure mode, used to simulate the dynamic evolution of heat transfer processes in future time periods. Digital twin simulation can be used to predict future trends in the thermal parameters of critical equipment, supporting proactive risk assessment. In an exemplary embodiment, digital twin simulation may include, but is not limited to, one or more of the following: graph neural network-based heat propagation simulation, reduced-order thermal field simulation coupled with CFD simplification, and fast transient simulation based on thermal resistance networks.

[0077] Based on failure mode classification results and thermal path network models, digital twin simulations are performed on the heat transfer process in the thermal system to predict the evolution trend of thermal parameters of key equipment. A continuous heating assurance risk assessment index is calculated based on the equipment's thermal tolerance threshold. This can be achieved by using identified failure modes as initial disturbances input to the thermal path network model, driving digital twin simulations to predict temperature and heat flow changes at key nodes over several future periods. When the predicted values ​​approach or exceed the thermal tolerance threshold of the equipment materials, the risk assessment index is calculated based on the degree and duration of the exceedance. Furthermore, this operation can be achieved by using Monte Carlo simulations to consider parameter uncertainties and outputting risk probability distributions, or by combining reinforcement learning strategies to simulate the suppression effect of different self-healing actions on the risk index. This extends the diagnostic results from current state judgment to future risk prediction, supporting proactive self-healing decision-making.

[0078] Taking the early warning of thermal fatigue of heat exchanger tube bundles under frequent load fluctuations as an example, the self-diagnosis method of unattended heating station boiler equipment in this embodiment can be that the system detects that the cumulative rate of thermal risk in a certain area is rising periodically, and combines historical data to form a variable load thermal stress fatigue mode in the time-series feature space; the current real-time trajectory is classified into this mode after dynamic time warping and matching; the digital twin simulation predicts that the tube bundle wall temperature will experience 12 thermal cycles exceeding the material creep threshold in the next 24 hours based on the thermal path network model; based on this, the continuous heating guarantee risk assessment index is calculated to be 0.85, triggering an advanced early warning and automatically adjusting the load scheduling strategy to reduce the frequency of thermal shock.

[0079] In addition, refer to Figure 4 To achieve the above objectives, the present invention also provides a self-diagnosis system for unattended heating station boiler equipment faults, the system comprising: Data acquisition module 10 is used to collect thermal efficiency data, heat flow distribution data and environmental parameter data under different heating conditions through a multi-modal sensor array pre-arranged in the thermal system of the heating station boiler equipment, so as to obtain a dataset of thermal characteristics of the thermal system. The thermal field analysis module 20 is used to calculate the thermal gradient change between adjacent measuring points in the multimodal sensor array based on the thermal characteristic dataset, and to analyze the accumulation and dissipation trend of heat in the system in combination with the geometric characteristics of the thermal system and the thermodynamic parameters of the material, so as to obtain a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data. The perception planning module 30 is used to construct a thermal path network model of the thermal system and key equipment based on the three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data. The thermal path network model is used to determine the location of fixed and dynamic perception points in the potential fault area of ​​the heat station boiler, and to obtain a full-cycle operating condition perception point layout scheme. The diagnostic early warning module 40 is used to deploy operating condition monitoring devices in the potential fault area of ​​the boiler in the heat station according to the full-cycle operating condition sensing point layout scheme to collect equipment operation data, extract thermal dynamic characteristics from the equipment operation data, identify abnormal heat transfer modes in the thermal system based on the thermal dynamic characteristics, obtain equipment fault self-diagnosis results, and trigger fault early warning and self-repair based on the equipment fault self-diagnosis results.

[0080] Other embodiments or specific implementations of the unattended heating station boiler equipment fault self-diagnosis system described in this invention can be referred to the above-mentioned method embodiments, and will not be repeated here.

[0081] In addition, to achieve the above objectives, the present invention also provides a self-diagnostic device for unattended heating station boiler equipment faults. The device includes: a memory, a processor, and a self-diagnostic program for unattended heating station boiler equipment faults stored in the memory and executable on the processor. The self-diagnostic program for unattended heating station boiler equipment faults is configured to implement the steps of the self-diagnostic method for unattended heating station boiler equipment faults as described in any of the above descriptions.

[0082] In addition, to achieve the above objectives, the present invention also provides a medium storing a self-diagnosis program for unattended heating station boiler equipment faults, wherein when the unattended heating station boiler equipment fault self-diagnosis program is executed by a processor, it implements the steps of the self-diagnosis method for unattended heating station boiler equipment faults as described above.

[0083] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A self-diagnosis method for faults in unattended heating station boiler equipment, characterized in that, The method includes: By collecting thermal efficiency data, heat flow distribution data, and environmental parameter data under different heating conditions through a multi-modal sensor array pre-arranged in the thermal system of the heating station boiler equipment, a thermal characteristic dataset of the thermal system is obtained. The thermal gradient change between adjacent measuring points in the multimodal sensor array is calculated based on the thermal characteristic dataset. The accumulation and dissipation trend of heat in the system is analyzed by combining the geometric characteristics of the thermal system and the thermodynamic parameters of the material, and a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data are obtained. Based on the three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data, a thermal path network model of the thermal system and key equipment is constructed. The location of fixed sensing points and dynamic sensing points in the potential fault zone of the heat station boiler is determined through the thermal path network model, and a full-cycle operating condition sensing point layout scheme is obtained. According to the full-cycle operating condition sensing point layout scheme, operating condition monitoring devices are deployed in the potential fault area of ​​the boiler in the heating station to collect equipment operation data. Thermal dynamic characteristics are extracted from the equipment operation data, and abnormal heat transfer modes in the thermal system are identified based on the thermal dynamic characteristics to obtain equipment fault self-diagnosis results. Fault warning and self-repair triggering are performed based on the equipment fault self-diagnosis results.

2. The self-diagnosis method for unattended heating station boiler equipment as described in claim 1, characterized in that, The process involves calculating the thermal gradient change between adjacent measuring points in the multimodal sensor array based on the thermal characteristic dataset, and analyzing the accumulation and dissipation trends of heat within the system in conjunction with the geometric characteristics of the thermal system and the thermodynamic parameters of the materials, to obtain a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data, including: Environmental noise is eliminated from the thermal efficiency data by adaptive Kalman filtering to obtain filtered thermal efficiency field data. Based on the thermal efficiency field data and the heat flow distribution data, the thermal gradient vector and rate of change of adjacent measuring points in the multimodal sensor array are calculated, and the thermal field is dynamically corrected in combination with the environmental parameter data to obtain a preliminary thermal distribution model. Based on the preliminary thermal distribution model and the geometric characteristics of the thermodynamic system, a thermal equilibrium differential equation is established. The finite volume method is used to solve the thermal equilibrium differential equation to obtain the continuous thermal field distribution inside the thermodynamic system. The thermal gradient abrupt change regions and thermal flow anomaly regions are identified from the continuous thermal field distribution and heat flow distribution data. The thermal risk accumulation index of each thermal gradient abrupt change region and thermal flow anomaly region is calculated by combining the material thermodynamic parameters, so as to obtain a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data.

3. The self-diagnosis method for unattended heating station boiler equipment as described in claim 2, characterized in that, The process involves calculating the thermal gradient vector and rate of change of adjacent measuring points in the multimodal sensor array based on the thermal efficiency field data and the heat flow distribution data, and then dynamically correcting the thermal field using the environmental parameter data to obtain a preliminary thermal distribution model, including: Based on the thermal efficiency field data, the thermal parameter values ​​of each measuring point are spatially mapped in the three-dimensional spatial coordinate system of the multimodal sensor array to obtain the initial thermal spatial distribution map. The thermal gradient vector is calculated for the thermal parameter values ​​of adjacent measuring points in the initial thermal spatial distribution diagram. The thermal change rate of each measuring point along the principal axis of the thermal system is calculated using the higher-order finite difference method to obtain the thermal gradient vector field. Based on the correspondence between the heat flow distribution data and the thermal gradient vector field, the equivalent thermal resistance coefficient of each region of the thermodynamic system is calculated using the robust regression method, and a heat flow-thermal efficiency relationship model is established. The heat flow-thermal efficiency relationship model is dynamically compensated using the environmental parameter data to obtain a preliminary thermal distribution model.

4. The self-diagnosis method for unattended heating station boiler equipment as described in claim 1, characterized in that, The process involves constructing a thermal path network model of the thermal system and key equipment based on the three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data. This model is used to determine the locations of fixed and dynamic sensing points in the potential fault zone of the boiler in the heat station, resulting in a full-cycle operating condition sensing point layout scheme. This includes: Density clustering analysis is performed on thermal anomaly points within the thermal system based on the three-dimensional thermal anomaly distribution map. Based on the results of the clustering analysis, similar thermal anomaly points are merged into thermal risk zones. The heat transfer direction and attenuation rate of each thermal risk zone are calculated based on the thermal risk accumulation rate data to obtain a thermal risk impact map. Based on the thermal risk impact map and the geometric topological relationship between the thermal system and key equipment, a weighted graph-form thermal path network model is constructed, where nodes represent thermal risk zones, edges represent heat transfer paths, and weights represent dynamic coefficients of heat conduction. The heat path network model is subjected to critical path analysis of heat propagation. The heat transfer time window and heat attenuation threshold from each heat risk zone to the core component of the heating station boiler are calculated. Based on the heat transfer time window and heat attenuation threshold, each heat risk zone is sorted according to the degree of impact on heating continuity, and a heat risk importance ranking table is obtained. Based on the thermal risk importance ranking table, the locations of fixed and dynamic sensing points in the potential fault zone of the heating station boiler are determined, resulting in a full-cycle operating condition sensing point layout scheme.

5. The self-diagnosis method for unattended heating station boiler equipment as described in claim 4, characterized in that, The step of determining the locations of fixed and dynamic sensing points in the potential fault zone of the boiler in the heating station based on the thermal risk importance ranking table, and obtaining a full-cycle operating condition sensing point layout scheme, includes: The thermal risk zones are classified according to the thermal risk importance ranking table, and the dynamic coverage radius of each thermal risk zone is calculated to obtain the thermal risk classification table. A working condition stability analysis is performed on the high-risk areas in the heat risk classification table. The root mean square error of the position deviation of each high-risk area under different heating loads is calculated, and the high-risk areas are divided into steady-state risk areas and transient risk areas based on the root mean square error of the position deviation. Fixed sensing points are set according to the location of the steady-state risk zone. The minimum coverage set algorithm is used to optimize the number of fixed sensing points so that each steady-state risk zone is covered by at least one fixed sensing point, while maximizing the sensing redundancy, thus obtaining a fixed sensing point layout scheme. Based on the movement trajectory and frequency characteristics of the transient risk zone, dynamic sensing points are set up in the transient risk zone to obtain a dynamic sensing point layout scheme. The fixed sensing point layout scheme and the dynamic sensing point layout scheme are then integrated to form a full-cycle working condition sensing point layout scheme.

6. The self-diagnosis method for unattended heating station boiler equipment as described in claim 1, characterized in that, The process involves deploying condition monitoring devices in the potential fault zone of the boiler in the heating station according to the full-cycle operating condition sensing point layout scheme to collect equipment operation data, extracting thermal dynamic characteristics from the equipment operation data, identifying abnormal heat transfer modes in the thermal system based on the thermal dynamic characteristics, obtaining equipment fault self-diagnosis results, and triggering fault early warning and self-repair based on the equipment fault self-diagnosis results, including: According to the full-cycle operating condition sensing point layout scheme, operating condition monitoring devices are deployed in the potential fault area of ​​the thermal station boiler to collect equipment operation data. Multi-scale feature extraction is performed on the equipment operation data by empirical mode decomposition to separate the steady-state and transient components of the equipment operation data and obtain the thermal dynamic feature vector. The thermal state matrix of the thermal system is constructed based on the thermal dynamic feature vector, and the model is matched with a pre-set normal thermal mode benchmark library to obtain the thermal anomaly index. The thermal anomaly index is combined with heating condition parameters for multi-dimensional correlation analysis. An anomaly mode is classified using a fault self-diagnosis algorithm based on the characteristics of the thermal system. The continuous heating guarantee risk assessment index of the equipment is calculated to obtain the equipment fault self-diagnosis results. Fault warning and self-repair triggering are performed based on the equipment fault self-diagnosis results.

7. The self-diagnosis method for unattended heating station boiler equipment as described in claim 6, characterized in that, The process involves multi-dimensional correlation analysis of the thermal anomaly index with heating condition parameters, classification of anomaly modes using a fault self-diagnosis algorithm based on thermal system characteristics, calculation of the equipment continuous heating guarantee risk assessment index, obtaining equipment fault self-diagnosis results, and triggering fault early warning and self-repair based on the equipment fault self-diagnosis results, including: The three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data are combined with historical heating condition parameters to construct a time-series feature space. Based on the structural characteristics of the thermal system, the data points in the feature space are subjected to condition-adaptive clustering to form a condition-driven benchmark fault mode library. The thermal anomaly index is combined with the real-time heating condition parameters and the trajectory is located in the constructed time-series feature space. The dynamic time warping distance is calculated with each mode in the benchmark fault mode library, and the anomaly mode with the smallest distance is selected as the fault mode classification result. Based on the fault mode classification results and the thermal path network model, a digital twin simulation of the heat transfer process in the thermal system is performed to predict the evolution trend of thermal parameters of key equipment. The continuous heating guarantee risk assessment index is calculated according to the thermal tolerance threshold of the equipment to obtain the equipment fault self-diagnosis results. Fault warning and self-repair triggering are performed based on the equipment fault self-diagnosis results.

8. A self-diagnostic system for unattended heating station boiler equipment, characterized in that, The system includes: The data acquisition module is used to collect thermal efficiency data, heat flow distribution data and environmental parameter data under different heating conditions through a multi-modal sensor array pre-arranged in the thermal system of the heating station boiler equipment, so as to obtain a dataset of thermal characteristics of the thermal system. The thermal field analysis module is used to calculate the thermal gradient change between adjacent measuring points in the multimodal sensor array based on the thermal characteristic dataset, and to analyze the accumulation and dissipation trend of heat in the system by combining the geometric characteristics of the thermal system and the thermodynamic parameters of the material, so as to obtain a three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data. The perception planning module is used to construct a thermal path network model of the thermal system and key equipment based on the three-dimensional thermal anomaly distribution map and thermal risk accumulation rate data. The thermal path network model is used to determine the location of fixed and dynamic perception points in the potential fault area of ​​the heat station boiler, and to obtain a full-cycle operating condition perception point layout scheme. The diagnostic and early warning module is used to deploy operating condition monitoring devices in the potential fault area of ​​the boiler in the heating station according to the full-cycle operating condition sensing point layout scheme to collect equipment operation data, extract thermal dynamic characteristics from the equipment operation data, identify abnormal heat transfer modes in the thermal system based on the thermal dynamic characteristics, obtain equipment fault self-diagnosis results, and trigger fault early warning and self-repair based on the equipment fault self-diagnosis results.

9. A self-diagnostic device for unattended heating station boiler equipment, characterized in that, The device includes: a memory, a processor, and an unattended heating station furnace equipment fault self-diagnosis program stored in the memory and executable on the processor, the unattended heating station furnace equipment fault self-diagnosis program being configured to implement the steps of the unattended heating station furnace equipment fault self-diagnosis method as described in any one of claims 1 to 7.

10. A medium, characterized in that, The medium stores a self-diagnostic program for unattended heating station boiler equipment faults. When the processor executes the self-diagnostic program for unattended heating station boiler equipment faults, it implements the steps of the self-diagnostic method for unattended heating station boiler equipment faults as described in any one of claims 1 to 7.

Images