Boiler state monitoring and predicting system based on artificial intelligence

Abstract

The invention discloses a boiler state monitoring and predicting system based on artificial intelligence, and relates to the technical field of state monitoring. The system comprises a data acquisition part, a combustion stability analysis part, a hydrodynamic instability risk assessment part, a steam drum water level fluctuation prediction part and a fault probability and service life assessment part. The data acquisition part is used for acquiring boiler operation data and water circulation data; the combustion stability evaluation part is used for calculating a thermodynamic efficiency index; the combustion stability analysis part is used for calculating a combustion stability coefficient; the hydrodynamic instability risk assessment part is used for calculating a hydrodynamic instability risk index; the steam drum water level fluctuation prediction part is used for calculating and predicting steam drum water level fluctuation; and the fault probability and service life evaluation part is used for evaluating the fault probability and the residual service life of the boiler. According to the invention, the comprehensiveness, accuracy and predictability of boiler operation state monitoring are obviously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of state monitoring, and in particular to a boiler state monitoring and prediction system based on artificial intelligence. Background Art

[0002] As the core thermal equipment in industrial production and energy conversion systems, boilers are widely used in many fields such as electricity, petrochemicals, metallurgy, papermaking, and urban heating. Its main function is to convert the chemical energy of fuel into thermal energy, and output high-temperature and high-pressure steam through the water-steam conversion system to drive the steam turbine or provide process heat source. The operating status of the boiler system is directly related to the safety, economy and energy efficiency of the entire industrial process. Therefore, accurate, continuous and dynamic monitoring and prediction of the boiler status has always been an important direction of boiler operation and maintenance technology research. In the traditional technical system, boiler status monitoring mainly relies on the threshold alarm mechanism based on experience and the regular manual inspection system. Typical monitoring parameters include steam pressure, water level, combustion chamber temperature, flue gas composition, etc. By configuring basic pressure transmitters, thermocouples, water level gauges, zirconia probes and other sensors, the collection of some boiler operation data is realized. These technologies are characterized by simple equipment structure, fast response speed and relatively low cost, which are suitable for the basic operation safety guarantee of early boiler systems. However, with the increasing complexity of boiler structure, the increase of operating load fluctuations and the increase in the demand for refined management of thermal efficiency, the traditional monitoring mechanism has gradually exposed the following shortcomings: First, most existing technologies use the "static threshold + logical judgment" method for alarm control, lacking the ability to identify the dynamic trend of the boiler's operating status. For example, the boiler water level is often set to "high / medium / low" three-level alarm. When the water level exceeds the set upper or lower limit, the alarm signal is triggered. However, in many cases, although the water level has not crossed the limit, it has fluctuated greatly in a short period, which may indicate the instability of the water circulation system or the drastic change in evaporation intensity. However, the traditional system has no perception of this. For example, when the boiler load suddenly changes, the combustion system responds inertia, which may cause short-term abnormal oxygen content. The traditional system often cannot determine whether it is a "normal disturbance" or a "potential fault", thus misreporting or missing. Secondly, the traditional system cannot establish the coupling relationship between multiple variables inside the boiler, and cannot form a "global operation portrait" of the boiler system. For example, the change in boiler thermal efficiency is closely related to factors such as fuel composition, water quality, heat load changes, and hydrodynamic state. However, in traditional technologies, each parameter is collected in isolation, lacking a unified modeling mechanism, and it is difficult to establish a quantitative relationship between the change in combustion efficiency and the thermal structural behavior of the boiler. This makes it difficult for operation and maintenance personnel to determine the root cause when thermal efficiency decreases or combustion becomes unstable. They often need to rely on "empirical inference" and trial and error to troubleshoot, resulting in a lot of downtime and waste of maintenance resources. Thirdly, traditional technologies have serious defects in life prediction and fault assessment. At present, boiler life assessment is mostly based on static estimation based on operating years or cumulative operating hours, completely ignoring historical operating information such as thermal load, number of starts, combustion disturbances and structural fatigue that the boiler bears at different stages. For example, if a boiler that has been in operation for 8 years has been frequently started and stopped in the first three years and has experienced several serious load disturbances, its actual structural fatigue may be much higher than that of a boiler that has been running smoothly for 10 years. The traditional method of "replacing parts at a fixed period" or "estimating life by operating time" is obviously unable to meet the high requirements of modern industry for "on-demand maintenance" and "remaining life prediction". Summary of the invention

[0003] The purpose of the present invention is to provide a boiler status monitoring and prediction system based on artificial intelligence, which realizes comprehensive perception and trend prediction of boiler operation status. The system not only improves the accuracy and real-time performance of boiler operation monitoring, but also realizes dynamic prediction of faults and lifespan, has high adaptability, self-learning ability and engineering deployability, and significantly enhances the operational safety, intelligence level and economic benefits of the boiler system.

[0004] The technical solution of the present invention is achieved in this way: The boiler state monitoring and prediction system based on artificial intelligence includes: a data acquisition part, a combustion stability analysis part, a hydrodynamic instability risk assessment part, a drum water level fluctuation prediction part, and a failure probability and life assessment part; the data acquisition part is used to acquire boiler operation data and water circulation data; the combustion stability assessment part is used to calculate the thermodynamic efficiency index reflecting the boiler operation efficiency and thermal energy conversion capacity based on the acquired boiler operation data, taking into account the physical properties of water and the thermal instability factors caused by steam pressure fluctuations; the combustion stability analysis part is used to calculate the thermodynamic efficiency index reflecting the boiler operation efficiency and thermal energy conversion capacity based on the thermodynamic efficiency index and boiler operation data. The combustion stability coefficient is calculated based on the data of the row; the hydrodynamic instability risk assessment part is used to calculate and evaluate the hydrodynamic instability risk index of two-phase flow fluctuations, circulation interruptions or local drying up in the water wall tube according to the combustion stability coefficient and water circulation data; the drum water level fluctuation prediction part is used to calculate the optimal drum geometry coefficient through a preset deep learning model, and calculate and predict the drum water level fluctuation in combination with the hydrodynamic risk index; the failure probability and life assessment part is used to evaluate the boiler failure probability and remaining life according to the thermodynamic efficiency index, the combustion stability coefficient, the hydrodynamic instability risk index and the predicted drum water level fluctuation.

[0005] Furthermore, boiler operation data include: boiler combustion efficiency ; Fuel flow , unit is kg / h; steam temperature , in °C; ambient temperature , in °C; specific heat capacity of water , unit is kJ / (kg·°C); boiler water density , unit is kg / m³; boiler water volume , unit is m³; maximum fluctuation value of steam pressure , unit is MPa; convection heat transfer coefficient , unit is W / (m²·°C); heat exchange area , unit is m²; thermal conductivity of furnace wall insulation layer , unit is W / (m·°C); combustion chamber resonance frequency , unit is Hz; oxygen percentage in flue gas ; Carbon monoxide concentration , in ppm; air flow , unit is m³ / h; stable combustion fuel flow , unit is kg / h; flame temperature , unit is K.

[0006] Furthermore, water circulation data include: system operating pressure , unit is MPa; critical pressure , unit is MPa; water wall tube heat flux density , unit is kW / m²; water temperature rise , unit is °C; inner diameter of water pipe , unit is mm; water pipe length , unit is m; hydrodynamic viscosity , in m² / s; cavitation number ; Reynolds number of water flow ; Steam mass flow rate , unit is kg / s; drum cross-sectional area , in m²; steam density , unit is kg / m³; saturation temperature , unit is °C; load change sensitivity coefficient , the unit is h / kg, and the value range is 0.005 to 0.03.

[0007] Furthermore, the drum geometry coefficient , calculated by a preset deep learning model, specifically including: selecting multiple historical geometric structure parameters of the drum as input features of the deep learning model, the historical geometric structure parameters including: drum diameter, length, cross-sectional area and wall thickness; based on the historical drum water level fluctuations corresponding to each historical geometric structure parameter obtained by the acquisition as label data for the model output; using a composite network structure of a convolutional neural network combined with a long short-term memory network in deep learning to achieve a nonlinear mapping between historical geometric structure parameters and the dynamic process of historical drum water level fluctuations. The specific process includes: firstly using a convolutional neural network to extract spatial features of the input historical geometric structure parameters to capture different historical The implicit characteristic patterns between the combinations of geometric structure parameters; the dynamic feature extraction of historical drum water level fluctuations is then carried out through the long short-term memory network; in the network training stage, the training set with measured geometric structure parameters and measured drum water level fluctuations is explicitly used for supervised learning, and the loss function accurately adopts the mean square error function, and the sum of squares of the errors between the predicted drum water level fluctuations output by the network and the measured drum water level fluctuations is used as the standard for back propagation and gradient optimization until the prediction accuracy of the network model reaches a preset threshold; after the network training is completed, in the model use stage, the geometric structure parameters of the boiler drum to be analyzed are input into the deep learning model, and the deep learning model outputs the drum geometric structure coefficients .

[0008] Furthermore, the thermodynamic efficiency index reflecting the boiler operation efficiency and heat energy conversion capacity is calculated by the following formula: : ; Among them, when the furnace wall is made of high-density refractory bricks, The value range is 0.8 to 1.3; when the furnace wall is ordinary ceramic fiber blanket, The value range is 0.04 to 0.09; when the furnace wall is microporous calcium silicate board, The value range is 0.12 to 0.18; when the furnace wall is a composite insulation coating, The value range is 0.02 to 0.05.

[0009] Furthermore, the combustion stability coefficient is calculated by the following formula: : .

[0010] Furthermore, the hydrodynamic instability risk index is calculated by the following formula: : .

[0011] Furthermore, the drum water level fluctuation is calculated and predicted by the following formula: ; in, is the fuel flow rate change.

[0012] Furthermore, the boiler failure probability is evaluated by the following formula: : ; in, is the boiler factory thermodynamic efficiency index; Design service life for boilers; The actual usage time of the boiler; is the number of completed thermal cycles; Design the maximum number of thermal cycles; evaluate the remaining life by the following formula : .

[0013] The boiler state monitoring and prediction system based on artificial intelligence of the present invention has the following beneficial effects: First, the present invention constructs a high-dimensional, multi-scale boiler operation and water circulation state parameter collection system through the data acquisition part, and monitors the boiler combustion efficiency, steam pressure fluctuations, water temperature changes, flue gas composition, water flow state and other key indicators in real time, forming a set of high-precision data input foundation covering the overall thermodynamic behavior of the boiler, solving the information loss problem caused by traditional single parameter and low-frequency data monitoring. This multi-source data perception mechanism enables the system to accurately capture the changes in the operating characteristics of the boiler, and can reflect the small disturbances of the boiler under actual working conditions in real time, effectively improving the continuity, integrity and reliability of the data.

[0014] Secondly, the present invention proposes a systematic combustion stability and hydrodynamic instability risk assessment method, which can deeply reflect the multi-physical coupling characteristics existing in the boiler combustion and water circulation system. By establishing the combustion stability coefficient and the hydrodynamic instability risk index, the thermal disturbance, structural response and fluid state change inside the boiler are closely combined, so that the system can timely identify potential unstable states. Compared with the traditional alarm threshold method, the present invention can sensitively perceive the small disturbances and nonlinear changes of the boiler operating state by introducing nonlinear mapping mechanisms such as exponential functions, logarithmic functions and sine functions, greatly improving the accuracy and robustness of boiler state monitoring, and significantly reducing the false alarm rate and missed alarm rate. The present invention proposes a boiler failure probability assessment and life prediction model. This model fully integrates multiple factors such as the historical attenuation trend of boiler thermodynamic efficiency, the dynamic response of drum water level fluctuations, the coupling relationship between combustion stability and hydrodynamic risk, and the accumulation effect of thermal cycle fatigue, and establishes a continuous, dynamic and accurate failure probability assessment mechanism and life prediction mechanism.

[0015] Compared with the traditional scheduled maintenance strategy, the evaluation model proposed in this invention can reflect the actual degree of wear and tear and remaining service capacity of the boiler's internal structure and operating status in real time, transforming the operation and maintenance of the boiler from "passive response" to "active prevention". As a result, not only the safety, stability and reliability of boiler equipment are greatly improved, but also the service life of the equipment can be effectively extended, unplanned downtime and maintenance costs can be reduced, and economic benefits can be significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of the system structure of an artificial intelligence-based boiler state monitoring and prediction system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present disclosure clearer and more understandable, the present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present disclosure and are not used to limit the present disclosure.

[0018] Example 1, reference Figure 1:A boiler status monitoring and prediction system based on artificial intelligence, the system includes: a data acquisition part, a combustion stability analysis part, a hydrodynamic instability risk assessment part, a drum water level fluctuation prediction part, and a failure probability and life assessment part; the data acquisition part is used to acquire boiler operation data and water circulation data; the combustion stability assessment part is used to calculate the thermodynamic efficiency index reflecting the boiler operation efficiency and thermal energy conversion capacity based on the acquired boiler operation data, taking into account the physical properties of water and the thermal instability factors caused by steam pressure fluctuations; the combustion stability analysis part is used to calculate the thermodynamic efficiency index reflecting the boiler operation efficiency and thermal energy conversion capacity based on the thermodynamic efficiency index and the boiler The combustion stability coefficient is calculated based on the operating data; the hydrodynamic instability risk assessment part is used to calculate and assess the hydrodynamic instability risk index of two-phase flow fluctuations, circulation interruptions or local drying up in the water wall tube according to the combustion stability coefficient and water circulation data; the drum water level fluctuation prediction part is used to calculate the optimal drum geometry coefficient through a preset deep learning model, and calculate and predict the drum water level fluctuation in combination with the hydrodynamic risk index; the failure probability and life assessment part is used to assess the boiler failure probability and remaining life according to the thermodynamic efficiency index, the combustion stability coefficient, the hydrodynamic instability risk index and the predicted drum water level fluctuation.

[0019] Specifically, during the operation of the boiler, there are many complex mechanisms involved, such as heat energy conversion, gas-liquid interaction, flow and heat transfer coupling, etc. The measurement of a single type of sensor or a single variable is not enough to support intelligent state recognition. Therefore, the data acquisition part of the present invention comprehensively deploys sensor systems of multiple dimensions such as thermodynamics, fluid mechanics, electrical and chemical gas composition, and integrates the discrete real-time data with high-frequency sampling and time synchronization through a unified data fusion platform to form a holographic state mapping of the boiler operation. Boiler operation data includes indicators such as combustion efficiency, fuel flow, flame temperature, and air flow, which together determine the heat energy generation capacity during the combustion process; at the same time, parameters such as ambient temperature, convective heat transfer coefficient, boiler water density, and boiler water volume directly affect the transfer efficiency of heat energy in the boiler structure. In particular, the dynamic fluctuation value of steam pressure not only reflects the output stability of heat energy in the evaporator, but also reveals whether there is thermal lag or load disturbance between combustion and heat transfer. Therefore, the high-speed response and stable output of the steam pressure sensor are crucial to the calculation of the thermodynamic efficiency index. In addition, the joint measurement of oxygen content and carbon monoxide concentration in flue gas provides an important reference for combustion completeness and local hypoxia. These chemical composition parameters reflect whether the combustion mixture is reasonable and are the basic elements for evaluating combustion stability. For the collection of water circulation data, more emphasis is placed on dynamic response and boundary condition recognition capabilities. Parameters such as system operating pressure, critical pressure, water temperature rise and heat flux density need real-time feedback to determine whether phase change instability or critical dryness has occurred in the water wall tube; the water flow Reynolds number and cavitation number jointly determine the flow state of the water flow and whether it is close to the critical zone of vaporization, and the steam mass flow rate and steam drum cross-sectional area are used to determine the response speed of the gas-liquid interface in two-phase flow. It is worth emphasizing that in order to meet the high requirements of deep learning models for data quality and sample consistency, all acquired data must undergo noise filtering, missing value completion, time series alignment and physical constraint verification to ensure that the data input into the neural network has engineering physical meaning and maintains statistical consistency.

[0020] As a key hub, the combustion stability assessment part is mainly responsible for extracting essential thermodynamic information from the complex and changeable boiler operating environment, assessing the stability of the current combustion system and the thermal energy conversion efficiency, and providing a solid energy assessment basis for subsequent hydrodynamic risk assessment and fault prediction. The core principle of this part is to construct a thermodynamic efficiency index that can reflect the dynamic equilibrium relationship between heat input and heat output during boiler combustion, and combine it with the disturbance characteristics of the combustion process to characterize the combustion stability with quantitative indicators. The stability of the combustion process is affected by many factors, including fuel supply fluctuations, uneven air ratios, changes in turbulent structure in the combustion chamber, flame oscillations, and rapid disturbances in steam pressure. These factors can easily disrupt the internal thermal balance of the boiler, resulting in fluctuations in combustion intensity, and even flameout, flashback and other dangerous conditions. Therefore, it is difficult to fully reflect the dynamic evolution characteristics of the combustion state by relying solely on static parameters. The present invention introduces the thermodynamic efficiency index as the basic calculation indicator to combine combustion efficiency and fuel flow rate. , the difference between steam temperature and ambient temperature and other parameters act together in the energy gain term. At the same time, the specific heat capacity, density, volume and steam pressure fluctuation of boiler water are introduced into the energy balance factor, so as to construct a comprehensive evaluation index that combines heat input and thermal system responsiveness.

[0021] In the actual implementation process, the thermodynamic efficiency index is not only regarded as a measure of energy utilization, but also endowed with dynamic response characteristics. The system realizes real-time perception of the combustion state through high-frequency data acquisition and index recalculation. In order to further enhance the dynamic adaptability of the evaluation, the present invention also introduces the heat transfer performance factor of the furnace wall insulation layer, considering the influence of different insulation materials on heat loss, so that the efficiency index can respond to the differences brought by the structural heat transfer characteristics while reflecting the heat output capacity. In addition, in the evaluation of combustion stability, it is also necessary to fully consider the nonlinear disturbances caused by the resonant frequency, pressure fluctuations and turbulence intensity in the system. This part of information is often reflected in the changes in oxygen ratio and abnormal fluctuations in carbon monoxide concentration. Too low oxygen content or increased carbon monoxide concentration may mean incomplete combustion or local oxygen deficiency, resulting in unstable flames. Based on this, the present invention incorporates the temporal changes of chemical gas components into the calculation factors of the stability coefficient during the evaluation process, and combines the matching relationship between air flow, fuel flow and flame temperature, and further extracts the non-stationary characteristics of the flame state through the dynamic stability model constructed by the artificial intelligence algorithm. It is worth noting that combustion stability assessment is not only a state judgment, but also a prerequisite for the accuracy of system prediction. Its assessment results directly affect the accuracy of the subsequent construction of the hydrodynamic instability risk index. Because when the combustion system is unstable, the unevenness of steam generation in the boiler will cause pulsation fluctuations in the water circulation and phase boundary imbalance of the two-phase flow, thereby bringing a higher risk of instability.

[0022] The hydrodynamic instability risk assessment part undertakes the core task of identifying and quantifying potential instability phenomena in the boiler water circulation system. Its design principle is rooted in the two-phase flow stability theory and the dynamic response mechanism under complex coupling field conditions. Combined with the high-dimensional state perception capability provided by the artificial intelligence system, it forms an intelligent model that can evaluate the boiler operation risk in real time. In actual operation, the boiler water circulation system faces the joint action of multiple unstable factors, such as uneven steam generation rate, vaporization mutation caused by local overheating, pressure drop fluctuation in the water wall tube, and drastic changes in heat load. Once these factors are triggered, they may cause drastic changes in the state of the steam-water mixed flow, leading to circulation interruption, local drying up, and even serious faults such as thermal explosion. Therefore, it is difficult to effectively warn by relying solely on conventional pressure difference threshold judgment. The present invention constructs a hydrodynamic instability risk index based on combustion stability, heat load distribution, hydraulic flow characteristics and geometric boundary factors to comprehensively reflect the possibility and intensity of system instability.

[0023] In this module, the combustion stability coefficient is first introduced into the risk modeling logic as a priori condition. This is because the instability of the combustion process often occurs before the hydrodynamic instability, and will affect the flow state on the water side through steam pressure fluctuations and heat flux density changes. The system dynamically identifies the proximity of the current heat flux intensity to the critical vaporization point based on the collected operating pressure, critical pressure, water wall tube heat flux density and water temperature rise parameters. If the heat flux continues to increase and the flow rate is insufficient, it is very easy to form a steam adhesion layer on the inner wall of the tube, causing the so-called "local drying" phenomenon. In order to more realistically reflect this process, the present invention introduces a nonlinear coupling mechanism of energy input-flow response based on exponential function modeling. By parameterizing the flow geometric characteristics composed of the water pipe diameter, length and fluid viscosity, the vaporization sensitivity induced by heat-flow mismatch is simulated. Furthermore, the system incorporates the cavitation number and Reynolds number into the evaluation model to characterize whether the fluid in the pipeline tends to the critical state of bubble formation. If the cavitation number is low or the Reynolds number fluctuates violently, it indicates that signs of near-critical instability have appeared in the flow. In addition, by performing logarithmic mapping on the hydrodynamic parameters, the model can maintain the continuity and sensitivity of calculation when facing high-amplitude fluctuations, avoiding abnormal assessment errors caused by extreme data. At the same time, the risk index is not an isolated value, but is highly coupled with other modules in the system. Its value will be used as a weight factor in the subsequent prediction of drum water level fluctuations, further amplifying or suppressing the response of the neural network to the unstable trend on the water side. Therefore, the risk assessment of hydrodynamic instability is not only physically credible, but also has a dynamic influence on the entire system.

[0024] In order to achieve accurate modeling of the dynamic changes of the drum water level, the present invention innovatively introduces a composite neural structure that combines the convolutional neural network and the long short-term memory network in deep learning. On the basis of retaining the traditional state causal logic based on thermal calculation, the system enhances the learning ability of the mapping relationship between high-dimensional geometric structure parameters and nonlinear water level response. In the specific implementation, the system first selects a large amount of historical drum operation data, including geometric parameters such as drum diameter, length, cross-sectional area and wall thickness as model input features. These parameters not only determine the volume characteristics of the drum, but also affect its dynamic response ability to steam flow and water supply flow. The model extracts spatial features of these parameters through convolutional neural networks, explores the common influence patterns that may exist under different structural combinations, and then connects to the long short-term memory network to model the historical water level fluctuation curve in time series, capturing the periodicity, mutation and inertia characteristics of water level fluctuations. This structure not only retains the decisive role of traditional structural parameters on system inertia and response time, but also integrates the powerful learning ability of artificial intelligence for the changing trend of complex time series data. It is worth noting that in order to ensure prediction accuracy and engineering applicability, the present invention strictly adopts a supervised learning strategy with measured labels during the model training stage, uses known geometric structures and historical fluctuation data to construct a high-quality training set, and uses the mean square error as the loss function. The network weights are back-propagated and optimized with reference to the actual water level changes until the error between the predicted water level and the actual water level is controlled within the set accuracy range. During the model deployment stage, the system will read the structural parameters and current hydrodynamic instability index of the target boiler in real time, and input them into the trained deep learning model, and output the drum geometric structure coefficient as the mediating variable of the nonlinear structural response. Subsequently, combined with the system's current steam mass flow rate, drum cross-sectional area, steam density, and temperature ratio parameters, the drum water level prediction value is finally output to complete the early identification of future short-term water level fluctuation trends.

[0025] The evaluation model first uses the predicted value of drum water level fluctuation as an important trigger factor, because abnormal fluctuations in water level are often early signs of structural disturbances or uncoordinated operation in the boiler system. The system uses it as an amplification factor for the intensity of thermal disturbances, and combines the thermal efficiency reference value of the boiler when it is designed to leave the factory to analyze the deviation of the current thermal efficiency, thereby establishing a performance degradation measure between the current operating state and the optimal design state. At the same time, the system also introduces thermal cycle fatigue theory, comparing the number of hot start-stop cycles completed by the boiler since its commissioning with the maximum number of designed cycles to quantify the degree of structural wear of the boiler. On this basis, the system further introduces the ratio of the cumulative operating time of the boiler to the design life, revealing the evolution stage of the boiler in the life cycle in the time dimension. Finally, by weighted combination of the above four key variables, combined with nonlinear exponential mapping and sinusoidal function modulation mechanism, a comprehensive fault probability model of the boiler at the current moment is constructed, realizing a reasonable characterization of the failure mechanism of multi-dependent variable coupling in complex systems. On the basis of the failure probability assessment, the failure probability and life assessment part further calculates the possible remaining safe use time of the boiler under the current operating state. The core idea is to reduce the actual available life on the basis of the design life, combining the currently assessed failure probability with the stability characteristics of the combustion and hydrodynamic systems. The system uses the failure probability as a risk weight, multiplies and compresses the design life, and introduces the combined effect of combustion stability and hydrodynamic risk coefficient as the angle modulation factor of the life fluctuation function to consider the nonlinear reduction effect of the stability state on the life. In this way, the present invention not only avoids the problem of poor adaptability of the traditional static life estimation model to the actual operating conditions, but also can respond in real time to the life attenuation trend caused by heat load fluctuations, changes in operating strategies or local structural aging during boiler operation, thereby realizing a dynamic, intelligent and adaptive boiler life prediction function.

[0026] Example 2: Boiler operation data includes: boiler combustion efficiency ; Fuel flow , unit is kg / h; steam temperature , in °C; ambient temperature , in °C; specific heat capacity of water , unit is kJ / (kg·°C); boiler water density , unit is kg / m³; boiler water volume , unit is m³; maximum fluctuation value of steam pressure , unit is MPa; convection heat transfer coefficient , unit is W / (m²·°C); heat exchange area , unit is m²; thermal conductivity of furnace wall insulation layer , unit is W / (m·°C); combustion chamber resonance frequency , unit is Hz; oxygen percentage in flue gas ; Carbon monoxide concentration , in ppm; air flow , unit is m³ / h; stable combustion fuel flow , unit is kg / h; flame temperature , unit is K.

[0027] Specifically, boiler combustion efficiency It is a direct reflection of the heat conversion capacity, representing the proportion of the heat released by a unit of fuel that is effectively used for water heating or evaporation. It is a core indicator for measuring the thermal economy of the boiler. This parameter is usually calculated based on the results of flue gas analysis and the calorific value of the fuel. In the intelligent system, it can be used as a direct gain factor for the calculation of the thermodynamic efficiency index. As the basic variable of heat source input, its size directly determines the heat energy supply capacity per unit time, and its unit is , which is in line with the practice of engineering heat balance calculation. Steam temperature With ambient temperature The difference between the two reflects the temperature difference driving force between the system's heat output and the external conditions, and is a reference for the temperature gradient in the entire boiler heat exchange process. The larger the difference, the greater the temperature rise capacity of the boiler per unit fuel, which also represents a higher thermal potential difference. In terms of physical energy storage characteristics, the specific heat capacity of water is , boiler water density , and water capacity The three together determine the heat capacity of the heat storage medium in the boiler, that is, the total amount of heat that can be stored per unit temperature difference. These parameters are reflected as thermal inertia and response capacity indicators in the thermodynamic efficiency calculation. At the same time, the maximum fluctuation value of steam pressure Characterizes the dynamic stability of the thermal state of the steam side of the boiler system. It is not only affected by combustion instability, but also closely related to water cycle imbalance. It is an important indicator for evaluating boiler transient disturbances. , heat exchange area , and thermal conductivity of furnace wall insulation layer The combination of reflects the efficiency level of heat transfer from the combustion chamber to the boiler water, among which The value of needs to be determined in combination with the actual insulation material, which is a correction factor affecting heat loss and thermal efficiency.

[0028] In the combustion state analysis, the combustion chamber resonance frequency It is a key parameter to describe whether there is thermoacoustic instability in the combustion process. It is closely related to the combustion chamber structure and the matching degree of the flame fluctuation frequency. Too high or synchronization with the flame frequency may cause flame oscillation or flame failure. and carbon monoxide concentration From the perspective of chemical reaction, it is used to judge whether the combustion is complete. Low oxygen content and increased carbon monoxide often mean incomplete combustion, which is an indispensable data source for combustion stability assessment. With stable combustion fuel flow The ratio of the flame temperature to the air temperature constitutes the combustion air ratio, which is the core factor of flame stability, temperature distribution and emission control. In AI models, nonlinear discriminant factors are usually constructed based on their combined relationship. It is a direct reflection of the combustion intensity and a comprehensive reflection of the reaction heat release intensity, flame size and thermal enthalpy of the combustion products. It is of great significance for predicting whether the boiler is approaching the fuel overheating zone or the critical value of the high temperature stress of the structure.

[0029] Example 3: Water circulation data includes: system operating pressure , unit is MPa; critical pressure , unit is MPa; water wall tube heat flux density , unit is kW / m²; water temperature rise , unit is °C; inner diameter of water pipe , unit is mm; water pipe length , unit is m; hydrodynamic viscosity , in m² / s; cavitation number ; Reynolds number of water flow ; Steam mass flow rate , unit is kg / s; drum cross-sectional area , in m²; steam density , unit is kg / m³; saturation temperature , unit is °C; load change sensitivity coefficient , the unit is h / kg, and the value range is 0.005 to 0.03.

[0030] Specifically, the system operating pressure It is the primary parameter for evaluating the working conditions of the boiler water side, which directly affects the saturation temperature, evaporation potential and critical state of phase change of water. When the system pressure approaches the critical pressure of water When the interface between liquid and steam gradually blurs, the so-called pseudo-critical phenomenon may occur inside the boiler, causing the thermal disturbance to be significantly amplified, thus inducing the instability or drying of the two-phase flow. The system can evaluate in real time whether the current operation is close to the critical zone, thus giving an early warning of flow stability. and water temperature rise The former reflects the intensity of heat transfer per unit area, while the latter reveals the amplitude of temperature response caused by heat entering the water body. The two together determine whether there is a risk of boiling enhancement or film vaporization in the water body, and are important indicators for evaluating the risk of local drying and high-temperature steam layer formation. The geometric parameters of the water pipe include the inner diameter, and length , which determines the flow resistance and pressure drop distribution of the water flow, and is related to the hydrodynamic viscosity Together, they constitute the necessary conditions for describing the flow state. Hydrodynamic viscosity essentially reflects the flow ability of water under high temperature and high pressure environment, and its change directly affects the value of Reynolds number. It is the core dimensionless parameter for judging the flow state (laminar or turbulent) and is usually regarded as an important criterion for the stability of the flow inside the boiler. A low Reynolds number means a slow flow rate and easy formation of bubble aggregation, while a high Reynolds number indicates a high turbulence intensity, which can more effectively break up bubbles and reduce the risk of local overheating. At the same time, the cavitation number is introduced , used to evaluate whether cavitation may occur in the water flow. This phenomenon often occurs in the area of ​​local pressure drop, forming microbubbles and generating high-frequency impact on the tube wall, which is a typical cause of damage to the boiler water-cooled wall tube structure. In terms of drum dynamic modeling, the steam mass flow rate and drum cross-sectional area The key descriptive variables of the mass flux at the steam-water separation interface are related to the steam density Together, they determine the volume exchange rate of the steam-water mixture, which in turn affects the real-time fluctuation trend of the drum water level. , to determine whether the water is saturated at the current pressure, which is related to the steam temperature Combined use can determine the boundary state of overheating, overcooling or saturated operation. In addition, the load change sensitivity coefficient It is a characterization of the adaptability of the boiler in response to external load disturbances. Its unit is , which indicates the system water level or pressure response rate caused by the change of unit mass fuel load. The value range is 0.005 to 0.03. The larger the value, the more sensitive the system is to the fuel change, and the smaller the value, the stronger the inertia and self-stabilization ability of the system.

[0031] Example 4: Steam Drum Geometry Coefficients , calculated by a preset deep learning model, specifically including: selecting multiple historical geometric structure parameters of the drum as input features of the deep learning model, the historical geometric structure parameters including: drum diameter, length, cross-sectional area and wall thickness; based on the historical drum water level fluctuations corresponding to each historical geometric structure parameter obtained by the acquisition as label data for the model output; using a composite network structure of a convolutional neural network combined with a long short-term memory network in deep learning to achieve a nonlinear mapping between historical geometric structure parameters and the dynamic process of historical drum water level fluctuations. The specific process includes: firstly using a convolutional neural network to extract spatial features of the input historical geometric structure parameters to capture different historical The implicit characteristic patterns between the combinations of geometric structure parameters; the dynamic feature extraction of historical drum water level fluctuations is then carried out through the long short-term memory network; in the network training stage, the training set with measured geometric structure parameters and measured drum water level fluctuations is explicitly used for supervised learning, and the loss function accurately adopts the mean square error function, and the sum of squares of the errors between the predicted drum water level fluctuations output by the network and the measured drum water level fluctuations is used as the standard for back propagation and gradient optimization until the prediction accuracy of the network model reaches a preset threshold; after the network training is completed, in the model use stage, the geometric structure parameters of the boiler drum to be analyzed are input into the deep learning model, and the deep learning model outputs the drum geometric structure coefficients .

[0032] Specifically, the coefficient is automatically generated by a pre-trained deep learning model, whose input is the structural features of multiple drum instances, including the drum's diameter, length, cross-sectional area, and wall thickness. These parameters determine the drum's geometric inertia, heat exchange area, and buffering capacity of water-steam interface motion, and are key variables affecting the dynamic characteristics of the water level. In model construction, the system first inputs these structural parameters into a convolutional neural network (CNN) to extract its spatial features, that is, to identify the implicit structural patterns between different combinations of geometric features, such as "large diameter + small wall thickness" may bring higher thermal response sensitivity, while "small diameter + thick wall" may cause water level response hysteresis. This process can establish a "graphic" expression of the drum's geometric structure, enabling the subsequent network to have the ability to handle complex structural combinations. Next, the model inputs historical water level fluctuation data into a long short-term memory network (LSTM) to capture its dynamic trend over time. As a type of sequence data with strong nonlinearity, periodicity and suddenness, water level fluctuation is easily affected by factors such as heat load disturbance, water pump start and stop, and unstable combustion. Traditional linear prediction methods are difficult to effectively model its complex time series changes. However, the LSTM network can selectively retain historical information through a gating mechanism, effectively model the correlation between short-term disturbances and long-term trends, and thus restore the true fluctuation trajectory of the water level in the time dimension.

[0033] During the network training phase, the system adopts a supervised learning method with labels, using historical samples with measured steam drum geometry and corresponding water level fluctuation data as the training set, and selecting the mean square error (MSE) as the loss function. The square of the difference between the network output and the actual fluctuation value is used as the error metric, and gradient descent and weight adjustment are continuously performed until the model prediction accuracy reaches the preset standard. This precision closed-loop mechanism ensures that the model not only has structural recognition capabilities, but also can output response coefficients with practical significance. When the training is completed, the system enters the use stage. At this time, the user only needs to input the structural parameters of the target boiler drum, and the deep learning model can automatically output its corresponding geometric structure coefficients. This coefficient is essentially a normalized response factor, which characterizes the overall sensitivity of the drum to fluid disturbance, water-steam mixing and thermal inertia changes under the current structural conditions. It will be used as a core parameter in the subsequent water level fluctuation prediction formula, and will be used in conjunction with factors such as the hydrodynamic instability index, steam mass flow rate, drum cross-sectional area and steam density to complete a high-precision prediction of the future state of the drum water level.

[0034] Example 5: The thermodynamic efficiency index reflecting the boiler operation efficiency and heat energy conversion capacity is calculated by the following formula: : ; Among them, when the furnace wall is made of high-density refractory bricks, The value range is 0.8 to 1.3; when the furnace wall is ordinary ceramic fiber blanket, The value range is 0.04 to 0.09; when the furnace wall is microporous calcium silicate board, The value range is 0.12 to 0.18; when the furnace wall is a composite insulation coating, The value range is 0.02 to 0.05.

[0035] Specifically, in the index construction, the combustion efficiency term As the starting point of energy input, it is the efficiency factor of converting fuel chemical energy into thermal energy, and its level directly determines the amount of thermal power that can be provided by unit mass of fuel. , describes the fuel supply capacity of the combustion system per unit time in the form of mass flow rate. The higher the value, the more sufficient the heat source. When these two parameters are combined with the difference between steam temperature and ambient temperature When combined together, they form an input sub-item that reflects the boiler's ability to respond to external temperature differences. The larger the temperature difference, the stronger the boiler's ability to overcome the environmental thermal resistance, which also means that the system still has stable heat transfer capabilities under high load. In the denominator relative to the above input items, the specific heat capacity , boiler water density With pot water volume The product of and together constitutes the representation of the boiler's heat capacity, indicating the total heat energy required to raise the water temperature inside the boiler system from the initial temperature to the target temperature. This item reflects the thermal inertia of the system, that is, the hysteresis of the boiler's absorption of heat input, which determines the response time and energy loss from the input end to the complete absorption of heat energy. If the boiler is large in size, high in density or high in specific heat, it means that the system has strong inertia and responds slowly to temperature changes. Output hysteresis may occur under short-cycle load fluctuations, and the thermodynamic efficiency shows a downward trend.

[0036] In addition, the maximum fluctuation of steam pressure The introduction of the denominator in the form of a square root is an important disturbance term for evaluating thermal-fluid instability in boiler operation. Steam pressure is a direct reflection of system load stability. The more violent its fluctuations are, the more unbalanced the heat exchange and phase change processes within the system are, which will lead to a significant increase in energy loss in the heat transfer process. This term exists in the form of a square root in the denominator, which means that its growth will weaken the final value of the thermal efficiency by a power-weighted manner, thereby producing actual feedback on the operating status. In addition to the above-mentioned ratio structure based on energy input and heat storage response, the index also introduces an exponential correction term in the form of This is an adjustment mechanism specifically for the heat dissipation loss of the furnace wall, which is used to feed back the deviation between the actual heat transfer efficiency of the boiler and the designed thermal structure into the final efficiency index. It indicates the intensity of heat exchange between the hot flue gas in the furnace and the furnace wall. The larger the value, the faster the heat transfer rate per unit area of ​​the boiler. It is an important structural parameter that characterizes the scale of the heat transfer path. The larger the area, the stronger the heat output capacity of the boiler. However, this heat transfer is not completely equal to the heat utilization efficiency. The thermal conductivity of the furnace wall material It reflects the energy loss caused by insufficient insulation performance during the heat transfer to the boiler water system. The range varies significantly, for example, the The value is only 0.02 to 0.05, which is much lower than the 0.8 to 1.3 of traditional high-density refractory bricks. This means that under the same heat flux, the thermal efficiency of composite boilers will be much better than that of traditional structures. The correction of heat loss in this index is expressed in exponential form, reflecting the "exponential" erosion characteristics of heat loss on system efficiency, that is, with the increase of heat exchange intensity and area, if the insulation performance is poor, the overall system thermal efficiency will show nonlinear attenuation. This processing method avoids the problem of underestimation of the effect of heat dissipation under extreme structural conditions by traditional linear regression models, and has higher prediction accuracy and engineering applicability.

[0037] Example 6: Calculate the combustion stability coefficient using the following formula: : .

[0038] Specifically, in the boiler state monitoring and prediction system based on artificial intelligence of the present invention, the combustion stability coefficient It is an important criterion for quantifying the dynamic steady-state level of the combustion system. It not only describes the ability of the boiler combustion process to resist disturbances under current thermal conditions, but also provides a key reference for subsequent hydrodynamic instability risk assessment and fault probability modeling. The stability of the combustion system is the fundamental guarantee for safe, efficient and durable operation of the boiler. Its essence is a dynamic behavior under the mutual coupling of multiple physical fields. It not only depends on whether the heat supply is sufficient and whether the combustion air is sufficient, but also is affected by multiple nonlinear factors such as flame oscillation, acoustic cavity resonance, abnormal gas composition and load changes. The present invention constructs a combustion stability coefficient formula with measurable physical quantities, reasonable structure and nonlinear response characteristics, abstracting these complexities into a comprehensive feature quantity that can be received and utilized by an artificial intelligence model, thereby realizing intelligent identification of the boiler combustion state.

[0039] The coefficient is firstly based on the built-in thermodynamic efficiency index As the basic weight of the energy input part. This index has been defined in the previous embodiment, and it represents the efficiency level of the boiler in converting the chemical energy of the fuel into the thermal energy of steam. In the combustion stability evaluation, if the system can operate at a higher efficiency, it means that the combustion process is relatively complete, the heat exchange is good, and the coupling balance between the heat source and the heat load is high, so The larger it is, the more stable the boiler can be considered to be. The factor it is multiplied by is the resonant frequency of the combustion chamber. , which is the core variable to characterize the acoustic characteristics of boiler combustion. Flame pulsation or thermoacoustic instability often occurs in boilers under high load conditions. Especially when the disturbance frequency of the combustion flame is close to the natural frequency of the structural cavity, heat energy will be released in the form of acoustic vibration, which will induce periodic flame instability or local extinction. It can be used to reveal whether the boiler structure is prone to exciting acoustic resonance modes, thereby more realistically reflecting the stability of the combustion state under the background of thermal-acoustic coupling.

[0040] In contrast to the energy input above, the formula introduces two gas composition indicators in the denominator: oxygen percentage and carbon monoxide concentration . Oxygen is an essential combustion medium, and its content determines whether the mixture can maintain a complete combustion process; while carbon monoxide is a typical product of incomplete combustion, and an increase in its concentration often means flame disorder, uneven fuel distribution or local oxygen deficiency. The product of the two is used as the attenuation term to introduce the coefficient denominator. Its physical meaning is that even if the boiler is operating under high efficiency conditions, if there is too much carbon monoxide or insufficient oxygen supply, it means that the combustion integrity is reduced and the system may enter an unstable state. Therefore, it needs to be deducted from the stability index. The treatment of adding 1 to the carbon monoxide concentration is to prevent numerical instability caused by the denominator approaching zero at extremely low concentrations, while ensuring the high sensitivity of this item to system combustion deviations.

[0041] More importantly, the last part of the formula introduces a sinusoidal function to form a nonlinear regulation mechanism for air-fuel ratio and flame temperature, namely This item reflects the dynamic response capability of the boiler system to the combustion state under the combined influence of the mixed gas ratio and heat intensity. In actual operation, the air flow With stable combustion fuel flow The ratio of ∠ to ∠ constitutes the excess air coefficient during the combustion process, which directly determines the flame structure, temperature distribution and smoke composition. If the air supply is insufficient, it is easy to cause carbon monoxide accumulation. If it is excessive, the flame temperature will be reduced, resulting in incomplete combustion. As a direct indicator of the heat intensity in the system, the higher the value, the more sensitive the system is to the change in the ratio. Its square root appears in the formula as a normalization term, which is used to correct the ratio weight under high temperature conditions. The introduction of the sine function reflects nonlinear modulation: when the gas distribution ratio is in the optimal range, its function value is close to 1, indicating that the combustion is most stable; once the gas distribution is unbalanced, the function drops rapidly, and the stability coefficient drops accordingly, reflecting the boiler's high sensitivity to non-ideal combustion conditions.

[0042] Example 7: Calculate the hydrodynamic instability risk index using the following formula: : .

[0043] Specifically, from the formula structure, the index is based on the system operating pressure. Combustion stability factor The square root of is taken as the primary driving term. This part reflects the dominant energy excitation intensity of the boiler water circulation system, where It indicates the thermal driving pressure on the water side at present, which is the basic power to promote the circulation of boiler water along the water wall tube. The introduction of represents the indirect regulation of the combustion system on thermal stability. When the combustion process is more stable, the steam generation rate is more uniform, the heat distribution of the boiler water is more stable, the water side pressure fluctuation is smaller, and the system water circulation is more capable of maintaining normal operation. Therefore, combining these two and comparing them with the critical pressure The purpose of making a ratio is to determine the relative distance of the system from the critical point of the two-phase flow under the current operating conditions. The closer the critical pressure is, the easier it is for the system to fall into an unstable area due to nonlinear feedback of thermal-fluid coupling.

[0044] The exponential term in the middle It is used to characterize the influence factor of local thermal disturbance on flow stability of water wall tube. This structure is constructed in the form of "1 minus exponential decay". The physical meaning is: when the heat flux density of the tube wall With water temperature rise When it is small, the thermal disturbance is not strong, the exponential term value is close to 1, and the overall contribution is low; and When they rise at the same time, it means that the system water wall pipe has been subjected to strong thermal shock per unit length, especially when the water pipe diameter is Small, length Short, viscosity When the temperature is low, the water flow is more easily disturbed by local steam generation, resulting in the formation of a steam blanket or even local drying or flow backflow. The mathematical form of this term simulates the process of the gradual impact of thermal disturbances on the circulation system: the impact is not obvious under small thermal disturbances, but as the heat load increases, its impact will increase rapidly, forming a nonlinear mutation risk.

[0045] The last term of the formula It is a risk correction factor used to characterize the relationship between the current flow state and the vaporization trend. It is a dimensionless parameter that describes whether the fluid enters the vaporization zone when the local pressure drops. The smaller the value, the closer it is to the critical state of vaporization. Since the cavitation number may be too small or even close to zero, a constant of 10 is added to the formula to ensure that the overall value is stable and has identifiability on an engineering scale. The Reynolds number in the denominator is It is an important feature to describe the flow state of water. The larger its value is, the more turbulent the system is, the stronger the flow stability is, and it is not easy to form a vapor layer. On the contrary, if If it is too low, it means that the water flow is slow, the fluid inertia is small, and it is more susceptible to thermal disturbances and phase changes, resulting in bubble aggregation and channel blockage. Therefore, this item is added in logarithmic form to quickly amplify the overall index value under high risk conditions, while the contribution to the risk index is controllable when the flow state is good, effectively balancing the dynamic responsiveness and anti-abnormal stability of the model.

[0046] Embodiment 8: The drum water level fluctuation is calculated and predicted by the following formula: ; in, is the fuel flow rate change.

[0047] Specifically, in actual operation, excessive fluctuations in the drum water level can easily cause false operation of the low water level protection or steam entrainment, which in turn causes unstable boiler operation or even shutdown. Therefore, it is of great significance to achieve predictive control of its trend. This embodiment introduces the structural response coefficient, flow stability factor and fuel disturbance function output by the deep learning model to construct a prediction model that integrates structural physics and dynamic disturbance mechanisms, which can achieve quantitative estimation of water level fluctuations. The core variable of this prediction model is the drum water level fluctuation quantity. , which is first affected by the geometric structure response coefficient This coefficient is obtained by combining convolutional neural network and long short-term memory network in the above embodiment, and reflects the overall response ability of the drum structure (such as diameter, wall thickness, volume, etc.) to water level fluctuation. Under the same disturbance conditions, drums with different geometric characteristics have significant differences in fluctuation amplitude and response speed, so Individualized correction factors at the structural level are provided to make the prediction adaptable to the boiler structure.

[0048] Multiplied by this is the hydrodynamic instability risk index , this variable comes from Example 7, which comprehensively considers multi-dimensional factors such as system pressure, thermal flow disturbance, pipeline structure and flow state, and quantifies the unstable trend in the boiler water circulation system. The larger the index, the closer the water side flow is to the critical instability zone, which may induce severe disturbances at the steam-water separation interface, thereby significantly amplifying the fluctuation amplitude of the drum water level. Therefore, It is used as an amplifier in the model, making the system highly sensitive to water level fluctuations in the hydrodynamic imbalance state. It represents the volume flow rate of steam flow per unit cross-sectional area, that is, the volume exchange rate of steam flowing through the cross-sectional area of ​​the steam drum. is the steam mass flow rate, is the cross-sectional area of ​​the steam drum, is the steam density. The combination of the three reflects the ability of steam flow to drive the movement of the gas-liquid interface in the drum. The larger this item is, the higher the steam exchange rate in the drum per unit time is, and it is more likely to cause rapid changes in the water level when hydrodynamic disturbances exist. The term reflects the amplification effect of the system under the thermal disturbance dimension. is the current steam temperature, and is the saturation temperature at the corresponding pressure. Their ratio measures whether the current system thermal state is in the overheating or underheating area. If the steam temperature is much higher than the saturation temperature, it means that the system heat load is large, the evaporation intensity is high, and the water level is prone to nonlinear fluctuations due to the rapid change of the evaporation rate. Therefore, this term is introduced by square root to smooth the thermal response while maintaining the physical expression ability of risk amplification under overheating conditions. The last term of the formula is the response function to the fuel disturbance. Here Indicates the change in fuel flow rate, is the load change sensitivity coefficient, which indicates the system thermal load response amplitude caused by unit fuel disturbance. Since the change of fuel input directly causes the change of steam generation rate, which acts on the drum water level, this item is used to simulate the impact of short-term load changes on water level disturbance. The range of is set between 0.005 and 0.03, which is obtained through the identification of the system's historical operating characteristics and represents the dynamic response of different boilers to load fluctuations. The larger the value, the more sensitive the boiler is to fuel disturbances, and the amplitude of water level fluctuations increases accordingly.

[0049] Example 9: Evaluate the boiler failure probability using the following formula: : ; in, is the boiler factory thermodynamic efficiency index; Design service life for boilers; The actual usage time of the boiler; is the number of completed thermal cycles; Design the maximum number of thermal cycles; evaluate the remaining life by the following formula : .

[0050] Specifically, the calculation formula for the failure probability fully integrates four key indicators during the system operation process: structural disturbance response (expressed by water level fluctuation), performance degradation (expressed by thermal efficiency decline), fatigue accumulation (expressed by the number of thermal cycles) and life consumption (expressed by time progress). First, the first term of the index The drum water level fluctuation The water level fluctuation is a comprehensive reflection of the system thermal flow disturbance, hydrodynamic imbalance and load change, which has a direct impact on the boiler structure safety and operational reliability. The exponential function here amplifies the risk brought by the violent fluctuation in a way close to 1. Approaching zero, Close to zero, indicating that the boiler is running stably; As it increases, the risk accumulates rapidly, reflecting the system’s high sensitivity to water level disturbances and its nonlinear response characteristics. It is used to measure the performance degradation of the current boiler thermal efficiency compared to the factory state. is the initial thermodynamic efficiency of the boiler, and The actual efficiency obtained by current monitoring. When the boiler efficiency decreases due to scaling, increased thermal resistance or decreased combustion efficiency, the ratio increases, which means that the system bears a higher thermal burden while maintaining the same output load, and its internal structure (such as water-cooled wall, convection section, etc.) is more susceptible to fatigue and corrosion accumulation, and the risk of failure increases accordingly. The exponential power of 0.35 is a sensitivity adjustment factor set through model fitting and experimental data regression to ensure that the change curve of this item is smooth and has engineering stability throughout the life cycle. The third item It is used to reflect the degree of fatigue accumulation of boiler materials. Frequent start-stop operations during boiler operation will trigger thermal stress cycles, which are the main source of fatigue damage to key parts of the boiler (such as welds, expansion joints, and drum support points). The closer the ratio is to 1, the closer the boiler is to its thermal fatigue design limit and the greater the risk of failure. The exponential function is used to control its growth rate, so that it grows slowly in the early stage, but rapidly amplifies when the thermal cycle approaches the limit, accurately reflecting the nonlinear cumulative characteristics of fatigue damage. The last item It is an expression of life consumption in the time dimension. This term uses a sine function instead of a linear growth method, with the intention of introducing a concept of periodic fatigue: when the boiler operation time is close to the design life, the sine function is close to 1, indicating that the life risk in the time dimension is the greatest; while in the early stage of use, the output value of this function is small, and the system is in a relatively safe stage. This processing method can more accurately describe the boiler's tolerance to failure risks at different stages of its life, and form a complementary relationship with the above three physical risk terms.

[0051] In the artificial intelligence-based boiler state monitoring and prediction system of the present invention, the life prediction formula is the ultimate calculation model for realizing the quantification of the boiler health state and the sustainable operation evaluation. It is not only an extension of the failure probability assessment results, but also a key guiding quantity output after integrating the real-time operating status, structural stability and long-term service capacity of the boiler. Compared with the traditional empirical method of estimating the life by years or operating hours, the present invention adopts a cosine function control formula that integrates the dynamic disturbance response and the degree of state degradation, which realizes the continuity, predictability and structural adaptability of the remaining life of the boiler. Its physical basis, mathematical structure and engineering significance are highly coupled.

[0052] Boiler design life is the theoretical upper limit, which is the maximum service time set under material fatigue, manufacturing quality and standard operating conditions, usually in years or hours. It represents the maximum expected life of the boiler under "ideal working conditions". However, in actual working conditions, due to load disturbances, operating fluctuations, frequent start-ups and shutdowns, etc., the equipment life is far from being completely consumed in a linear manner. Therefore, the formula reduces it by two key factors to reflect the nonlinear characteristics of life consumption under actual operation. The first reduction factor is , that is, the failure probability As a direct quantification of the boiler health depreciation rate, this part expresses the "closeness" of the boiler to the fault state in the current state. When it is close to 1, that is, the system is in a serious instability or degradation state, the life prediction value tends to zero, indicating that the boiler is no longer in a safe state. Close to 0, it means that the system is running stably, the thermal structure is sound, and the life prediction value is close to the design life. This part is a linear reduction item, and its change is directly linked to the long-term health status of the system, reflecting the fundamental impact of historical accumulation and structural degradation on life.

[0053] More critical is the second reduction factor: This part fully reflects the core innovation of the present invention, which is to incorporate real-time operation disturbance dynamics into the life prediction model. It uses the cosine function as a regulator to establish a nonlinear state mapping relationship between boiler thermal instability and life prediction, so that the system is not only sensitive to "accumulated damage", but also can reflect the impact of changes in "operating pressure" on life in real time. The product term in the formula It is a combination of the combustion stability coefficient and the hydrodynamic instability risk index, which reflect the internal operation disturbance intensity of the boiler from the hot end and the water side respectively. If there is acoustic and vibration disturbance, uneven mixing, low oxygen or increased CO concentration in the combustion process of the boiler, then If the value decreases, it means that the system has entered the unstable zone at the hot end; if high heat flux density, low Reynolds number or high cavitation risk occurs in the boiler water cycle, then The higher the product of these two items, the closer the boiler is to the "thermal-fluid resonance zone". Even if there is no fault at present, the potential fatigue damage and micro crack development speed will be greatly accelerated.

[0054] Use the cosine function to map this product term to the interval , a dynamic adjustment can be achieved: when hour, When the product approaches 1, it means that the system is heat-flux stable and the lifespan remains at the theoretical upper limit; when the product approaches 2, The output is -1. Since the formula cannot have a negative life span in the physical sense, the actual model sets an upper limit constraint. In the state, The value will drop sharply to close to 0, and even trigger an alarm mechanism, indicating that the life span has been severely damaged and immediate intervention is required. The significance of this nonlinear control is to provide early feedback on the life consumption caused by operational disturbances. In traditional linear models, only relying on time and failure rate cannot capture the rapid erosion of life by short-term high-intensity disturbances (such as frequent boiler starts and stops or large load mutations). The present invention establishes a dynamic attenuation mechanism through a cosine function, so that in a short period of time, even if the system has not yet entered a high failure probability state, as long as the operating stability is greatly reduced, the life span is quickly compressed, prompting the system to issue an early warning. On the contrary, under long-term stable operation, even if the equipment has been in service for many years, as long as the current combustion and hydraulic system states are stable, the system can still maintain a relatively long remaining life estimate, reflecting the full exploitation of the equipment's potential.

[0055] The preferred embodiments of the present disclosure are described above with reference to the accompanying drawings, but the scope of the present disclosure is not limited thereby. Any modification, equivalent substitution and improvement made by those skilled in the art without departing from the scope and essence of the present disclosure shall be within the scope of the present disclosure.

Claims

1. The boiler status monitoring and prediction system based on artificial intelligence is characterized by: The system includes: a data acquisition part, a combustion stability analysis part, a hydrodynamic instability risk assessment part, a drum water level fluctuation prediction part, and a failure probability and life assessment part; the data acquisition part is used to acquire boiler operation data and water circulation data; the combustion stability assessment part is used to calculate the thermodynamic efficiency index reflecting the boiler operation efficiency and heat energy conversion capacity based on the acquired boiler operation data, taking into account the physical properties of water and the thermal instability factors caused by steam pressure fluctuations; the combustion stability analysis part is used to calculate the combustion stability coefficient based on the thermodynamic efficiency index and the boiler operation data; the hydrodynamic instability risk assessment part is used to calculate and evaluate the hydrodynamic instability risk index of two-phase flow fluctuations, circulation interruption or local drying up in the water wall tube based on the combustion stability coefficient and water circulation data; the drum water level fluctuation prediction part is used to calculate the optimal drum geometric structure coefficient through a preset deep learning model, and calculate and predict the drum water level fluctuation in combination with the hydrodynamic risk index; the failure probability and life assessment part is used to evaluate the boiler failure probability and remaining life based on the thermodynamic efficiency index, the combustion stability coefficient, the hydrodynamic instability risk index and the predicted drum water level fluctuation.

2. The artificial intelligence-based boiler state monitoring and prediction system according to claim 1, characterized in that: Boiler operation data include: boiler combustion efficiency ; Fuel flow , unit is kg / h; steam temperature , in °C; ambient temperature , in °C; specific heat capacity of water , unit is kJ / (kg·°C); boiler water density , unit is kg / m³; boiler water volume , unit is m³; maximum fluctuation value of steam pressure , unit is MPa; convection heat transfer coefficient , unit is W / (m²·°C); heat exchange area , unit is m²; thermal conductivity of furnace wall insulation layer , unit is W / (m·°C); combustion chamber resonance frequency , unit is Hz; oxygen percentage in flue gas ; Carbon monoxide concentration , in ppm; air flow , unit is m³ / h; stable combustion fuel flow , unit is kg / h; flame temperature , unit is K.

3. The artificial intelligence-based boiler state monitoring and prediction system according to claim 2, characterized in that: Water circulation data include: system operating pressure , unit is MPa; critical pressure , unit is MPa; water wall tube heat flux density , unit is kW / m²; water temperature rise , unit is °C; inner diameter of water pipe , unit is mm; water pipe length , unit is m; hydrodynamic viscosity , in m² / s; cavitation number ; Reynolds number of water flow ; Steam mass flow rate , unit is kg / s; drum cross-sectional area , in m²; steam density , unit is kg / m³; saturation temperature , unit is °C; load change sensitivity coefficient , the unit is h / kg, and the value range is 0.005 to 0.

03.

4. The artificial intelligence-based boiler state monitoring and prediction system according to claim 3, characterized in that: Drum geometry factor , calculated by a preset deep learning model, specifically including: selecting multiple historical geometric structure parameters of the drum as input features of the deep learning model, the historical geometric structure parameters including: drum diameter, length, cross-sectional area and wall thickness; based on the historical drum water level fluctuations corresponding to each historical geometric structure parameter obtained by the acquisition as label data for the model output; using a composite network structure of a convolutional neural network combined with a long short-term memory network in deep learning to achieve a nonlinear mapping between historical geometric structure parameters and the dynamic process of historical drum water level fluctuations. The specific process includes: firstly using a convolutional neural network to extract spatial features of the input historical geometric structure parameters to capture different historical The implicit characteristic patterns between the combinations of geometric structure parameters; the dynamic feature extraction of historical drum water level fluctuations is then carried out through the long short-term memory network; in the network training stage, the training set with measured geometric structure parameters and measured drum water level fluctuations is explicitly used for supervised learning, and the loss function accurately adopts the mean square error function, and the sum of squares of the errors between the predicted drum water level fluctuations output by the network and the measured drum water level fluctuations is used as the standard for back propagation and gradient optimization until the prediction accuracy of the network model reaches a preset threshold; after the network training is completed, in the model use stage, the geometric structure parameters of the boiler drum to be analyzed are input into the deep learning model, and the deep learning model outputs the drum geometric structure coefficients .

5. The artificial intelligence-based boiler state monitoring and prediction system according to claim 4, characterized in that: The thermodynamic efficiency index reflecting the boiler operation efficiency and heat energy conversion capacity is calculated by the following formula: : ； Among them, when the furnace wall is made of high-density refractory bricks, The value range is 0.8 to 1.3; when the furnace wall is ordinary ceramic fiber blanket, The value range is 0.04 to 0.09; when the furnace wall is microporous calcium silicate board, The value range is 0.12 to 0.18; when the furnace wall is a composite insulation coating, The value range is 0.02 to 0.

05.

6. The artificial intelligence-based boiler state monitoring and prediction system according to claim 5, characterized in that: The combustion stability coefficient is calculated by the following formula: : 。 7. The artificial intelligence-based boiler state monitoring and prediction system according to claim 6, characterized in that: The hydrodynamic instability risk index is calculated by the following formula: : 。 8. The artificial intelligence-based boiler state monitoring and prediction system according to claim 7, characterized in that: The following formula is used to calculate and predict the drum water level fluctuation: ； in, is the fuel flow rate change.

9. The artificial intelligence-based boiler state monitoring and prediction system according to claim 8, characterized in that: The boiler failure probability is estimated by the following formula : ； in, is the boiler factory thermodynamic efficiency index; Design service life for boilers; The actual usage time of the boiler; is the number of completed thermal cycles; Design the maximum number of thermal cycles; evaluate the remaining life by the following formula : 。

Images