Combustion balance AI intelligent detection and diagnosis method, system, device and product

Abstract

The present application relates to the technical field of heating furnace, in particular to a combustion balance AI intelligent detection and diagnosis method, system, equipment and product, comprising: when judging combustion imbalance, reading the rated pressure difference, actual pressure difference and working parameters of the burner, the working parameters including: flow orifice plate diameter, diameter ratio, expandable coefficient, outflow coefficient and fluid density; using the pressure difference method to calculate the standard gas and air flow of the burner respectively, and the actual gas and air flow; and collecting the concentration data set obtained by calculation to input it into the fault determination model, when the concentration data set meets the fault determination model, the fault type of the burner is determined as electromagnetic valve blockage (a strong correlation fault). The present application proposes a method of fast prediction based on limited dynamic factor combination for strong correlation faults, thereby reducing the fault detection cost by fast predicting the electromagnetic valve fault, and timely investigating the fault at the early stage of hidden danger.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of horizontal, vertical heating furnace production line combustion systems with radiant tubes, double cross proportional regulation and pulse control, and particularly relates to a combustion balance AI intelligent detection and diagnosis method, system, equipment and product. BACKGROUND

[0002] The metallurgical furnace radiant tube combustion heating is a heating technology widely used in the metallurgical industry. In the cold rolling continuous annealing furnace, the radiant tube is the key equipment for heating the strip steel in the furnace, and the working state of the radiant tube directly affects the annealing quality of the strip steel. Its working principle is that the fuel gas (mostly coal gas) and the combustion-supporting gas (air) are fully mixed and burned in the combustion chamber, the heat generated causes the radiant tube wall to be in a high temperature state, and the outer wall of the radiant tube transfers heat to the furnace atmosphere and the strip steel in the form of heat radiation to achieve the purpose of heating the strip steel without oxidation. The exhaust gas produced by combustion in the radiant tube is circulated into a heat exchanger or a smoke exhaust pipe for exhaust.

[0003] At present, most metallurgical enterprises have been in production for a long time, and the radiant tube combustion system has been running in harsh conditions of high temperature and high pressure, high load and strong impact for a long time. Long-term operation leads to an increase in the failure rate of devices such as electromagnetic valves, combustion equipment, radiant tubes and heat exchangers, which in turn causes combustion imbalance, and the combustion imbalance leads to a vicious cycle of equipment failure. In order to solve the above problems, the following automatic fault identification methods are proposed in the prior art:

[0004] I. Detecting the result - flame state

[0005] For example, patent application CN118168349A discloses a heating furnace flame monitoring method and system, which includes acquiring flame information in the heating furnace; performing data filtering and preprocessing based on the flame information; extracting features to distinguish flame states; establishing a flame state model to match the feature parameters with known flame states; by acquiring the flame information in the heating furnace, the flame state can be continuously monitored and abnormal conditions can be found in time, data filtering and preprocessing can remove noise and outliers, improving the accuracy and reliability of subsequent data analysis, which helps to accurately extract flame feature parameters.

[0006] However, the monitoring of flame-related states is easily affected by environmental conditions, such as light and smoke, which can interfere with the monitoring results, leading to inaccuracy and instability of the monitoring, and it puts high requirements on the performance of the environment and the monitoring equipment; at the same time, this method cannot accurately determine the state and feature parameters of the flame, and cannot conduct in-depth analysis of the combustion of the flame, resulting in low monitoring precision.

[0007] II. Predicting faults using AI mode

[0008] As disclosed in patent application CN117704416A, an AI-based automatic boiler combustion adjustment method includes the following steps: collecting boiler operation data in real time, including temperature, pressure, fuel consumption, and emission data; using the algorithm of an AI unit to analyze the collected data and identify the optimal state of the boiler operation; based on the analysis results, the AI unit automatically generates a combustion adjustment decision; automatically adjusting fuel supply and air flow parameters to optimize combustion; automatically adjusting the combustion strategy according to external environmental changes to solve the problems of combustion efficiency and emissions caused by environmental changes; the AI unit predicts potential failures and maintenance needs based on the performance data and historical maintenance records of the boiler components, and plans maintenance activities in advance.

[0009] For another example, patent application CN117490091A discloses a fuel air distribution control method and system, which includes: collecting historical data and preprocessing the collected data; constructing a deep learning model and training the model; adjusting the parameters of the model or optimizing the performance of the model; deploying the trained model to the actual fuel air distribution, and dynamically adjusting the fuel and air ratio according to the prediction results of the model.

[0010] For another example, patent application CN117144118A also discloses a self-adaptive control method for a continuous annealing furnace and related equipment. The method includes: constructing a first training sample library and a second training sample library for a target area, wherein the first training sample library includes the average value and the measured value of the single burner gas flow of the target area, and the second training sample library includes the average value and the measured value of the single burner air flow of the target area; constructing a regional balanced combustion model based on the first training sample library and the second training sample library; determining the abnormal situation of gas and air based on the regional balanced combustion model.

[0011] However, the existing AI model has relatively limited detection accuracy in application and implementation, and the application cost is also relatively high. SUMMARY

[0012] The purpose of the present application is to provide a combustion balance AI intelligent detection and diagnosis method, system, device and product, which partially solves or alleviates the above-mentioned deficiencies in the prior art, and can improve the reliability of fault prediction while reducing real-time monitoring pressure.

[0013] In order to solve the above-mentioned technical problems, the present application specifically adopts the following technical solutions:

[0014] The first aspect of the present application is to provide a combustion balance AI intelligent detection and diagnosis method, which is applied to a combustion system, the combustion system includes a heating furnace and a burner that provides a heat source for the heating furnace, and correspondingly, the method includes the following steps:

[0015] The burner is subjected to centralized fault judgment, which comprises:

[0016] S101: reading the number of the burner, and centrally reading the rated pressure difference, actual pressure difference and working parameters of the burner according to the number; the rated pressure difference comprises: gas standard pressure difference and air standard pressure difference; the actual pressure difference comprises: gas actual pressure difference and air actual pressure difference; the pressure difference is the pressure difference between the input side pressure and the output side pressure of the flow orifice plate in the pipeline of the burner; the working parameters comprise: the diameter of the flow orifice plate, the diameter ratio of the flow orifice plate to the pipeline, the expandable coefficient, the outflow coefficient and the fluid density;

[0017] S102: using the pressure difference method to calculate the standard gas flow Qgs and the standard air flow Qas of the burner through the gas standard pressure difference and the air standard pressure difference respectively;

[0018] S103: using the pressure difference method to calculate the actual gas flow Qg and the actual air flow Qa of the burner through the gas actual pressure difference and the air actual pressure difference;

[0019] S104: judging the equipment type of the heating furnace according to the actual input side pressure and the actual output side pressure, wherein when the fluctuation values of the input side pressure and the output side pressure both belong to the set fluctuation threshold, it is considered that the equipment type is double-cross proportional regulation;

[0020] S105: when the equipment type is double-cross proportional regulation, it is considered that the burner meets the centralized prediction condition, and the current centralized data set is collected to input the centralized data set into a fault judgment model, the centralized data set comprises: the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco and the tail gas oxygen content Co2, and the fault judgment model comprises:

[0021] |Qa-Qas|÷Qas<A1;

[0022] |Qg-Qgs|÷Qgs>A2;

[0023] Cco>A3;

[0024] Co2>A4;

[0025] Wherein, |Qa-Qas| ÷ Qas is the first index, |Qg-Qgs| ÷ Qgs is the second index, A1, A2, A3, A4 are respectively the first training threshold, the second training threshold, the third training threshold, the fourth training threshold; Correspondingly, the plurality of corresponding training thresholds are obtained by machine learning training.

[0026] S106: When the centralized data set meets the fault judgment model, it is determined that the fault type of the combustor is electromagnetic valve blockage.

[0027] In some embodiments, the step of obtaining the plurality of corresponding training thresholds by machine learning training comprises:

[0028] Obtaining a training sample set, the training sample set comprising: input end training data, the input end training data comprising: the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco, the tail gas oxygen content Co2; and output end training data corresponding to the input end training data, the output end training data comprising: electromagnetic valve blockage data;

[0029] Inputting the training sample set into a set model for at least one iteration training; wherein the set model is used to learn the internal relationship between the first index, the second index, the tail gas carbon monoxide content Cco, the tail gas oxygen content Co2 and the electromagnetic valve blockage data;

[0030] After the set model converges, outputting the prediction value interval of the first index, the second index, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2 corresponding to the electromagnetic valve blockage state through the set model, and setting a plurality of corresponding training thresholds according to the prediction value interval.

[0031] In some embodiments, before the fault judgment of the combustor, it further comprises the steps of:

[0032] (1) obtaining the historical training threshold set according to the set model last time;

[0033] (2) using the update data set formed by the combustor in the first period of time, the update data set comprising: input end update data, which comprises: the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2; and output end update data corresponding to the input end update data, the output end update data comprising: electromagnetic valve blockage data;

[0034] (3) collecting a first update data set from the update data set and inputting the corresponding input end update data into the fault judgment model to obtain a fault prediction result output by the fault judgment model; wherein the fault judgment model uses the historical training threshold to make a judgment;

[0035] (4) comparing the fault prediction result with the output end update data to calculate the current prediction accuracy of the fault judgment model.

[0036] In some embodiments, the method further comprises the step of: when the prediction accuracy is greater than a first set value, collecting a second update data set from the update data set;

[0037] inputting the second update data set into the setting model to update and train the setting model;

[0038] setting a new training threshold based on the updated setting model.

[0039] In some embodiments, before the fault judgment of the combustor, the method further comprises the step of:

[0040] judging the combustion balance state, which comprises:

[0041] calculating the air excess coefficient by collecting the excess air amount and the actual air amount in the tail gas; wherein the air excess coefficient is calculated by the following formula: alpha = 1 / [1-(Delta L / La)], alpha is the air excess coefficient, Delta L is the excess air amount, and La is the actual air amount;

[0042] judging whether the air excess coefficient belongs to a set threshold range, if yes, performing the step of: judging the fault of the combustor; if not, continuing to detect the excess air amount and the actual air amount of the tail gas.

[0043] The present application also provides a combustion balance AI intelligent detection and diagnosis system, wherein the intelligent detection method is applied to a combustion system, the combustion system comprises a heating furnace and a combustor providing a heat source for the heating furnace, and correspondingly, the system comprises a centralized fault judgment subsystem, which comprises:

[0044] a reading module configured to read a number of the burner, and read a rated pressure difference, an actual pressure difference and an operating parameter of the burner collectively according to the number; the rated pressure difference includes a gas standard pressure difference and an air standard pressure difference; the actual pressure difference includes a gas actual pressure difference and an air actual pressure difference; the pressure difference is a pressure difference between an input side pressure and an output side pressure of a flow orifice plate in a pipeline of the burner; and the operating parameter includes a diameter of the flow orifice plate, a diameter ratio of the flow orifice plate to the pipeline, an expandable coefficient, a flow-out coefficient and a fluid density;

[0045] a first calculation module configured to calculate a standard gas flow Qgs and a standard air flow Qas of the burner respectively by the gas standard pressure difference and the air standard pressure difference through a pressure difference method;

[0046] a second calculation module configured to calculate an actual gas flow Qg and an actual air flow Qa of the burner through the pressure difference method by the gas actual pressure difference and the air actual pressure difference;

[0047] a type judgment module configured to judge a device type of the heating furnace according to an actual input side pressure and an actual output side pressure, wherein when fluctuation values of the input side pressure and the output side pressure both belong to a set fluctuation threshold value, it is considered that the device type is a double-cross proportional regulation;

[0048] a data input module configured to consider that the burner meets a collective prediction condition when the device type is the double-cross proportional regulation, and collect a current collective data set to input the collective data set to a fault judgment model, the collective data set including the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, a tail gas carbon monoxide content Cco and a tail gas oxygen content Co2, and the fault judgment model including:

[0049] |Qa-Qas|÷Qas<A1;

[0050] |Qg-Qgs|÷Qgs>A2;

[0051] Cco>A3;

[0052] Co2>A4;

[0053] wherein |Qa-Qas|÷Qas is a first index, |Qg-Qgs|÷Qgs is a second index, A1, A2, A3 and A4 are respectively a first training threshold value, a second training threshold value, a third training threshold value and a fourth training threshold value, and a plurality of corresponding training threshold values are obtained through machine learning training;

[0054] a fault determination module, configured to determine that the fault type of the combustor is electromagnetic valve blockage when the centralized data set meets the fault determination model.

[0055] In some embodiments, the data input module further comprises:

[0056] a sample acquisition unit configured to acquire a training sample set, the training sample set comprising: input training data comprising the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2; and output training data corresponding to the input training data, the output training data comprising electromagnetic valve blockage data;

[0057] an iteration unit configured to input the training sample set into a set model to perform at least one iteration training; wherein the set model is configured to learn the internal correlation between the first indicator, the second indicator, the tail gas carbon monoxide content Cco, the tail gas oxygen content Co2, and the electromagnetic valve blockage data;

[0058] a threshold setting unit configured to output, by the set model, a predicted value interval of the first indicator, the second indicator, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2 corresponding to the electromagnetic valve blockage state after the set model converges, and set a plurality of training thresholds corresponding to the predicted value interval.

[0059] In some embodiments, further comprising: an accuracy prediction subsystem configured to perform the following steps:

[0060] (1) acquiring historical training thresholds set according to the set model last time;

[0061] (2) using an updated data set formed by the combustor in a first period of time, the updated data set comprising: input updated data comprising the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2; and output updated data corresponding to the input updated data, the output updated data comprising electromagnetic valve blockage data;

[0062] (3) collecting a first updated data set from the updated data set and inputting the corresponding input updated data into the fault determination model to obtain a fault prediction result output by the fault determination model; wherein the fault determination model uses the historical training thresholds for determination;

[0063] (4) According to the comparison between the fault prediction result and the output end update data, the prediction accuracy of the fault determination model at the current time is calculated.

[0064] The application further provides an electronic device, comprising a memory and a processor; the memory is used for storing a program; the processor is used for executing the program to realize each step of the intelligent detection method according to any one of the embodiments.

[0065] The application further provides a computer program product, comprising a computer program, when the computer program is executed by one or more processors, causing the one or more processors to execute the intelligent detection method according to any one of the embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference signs. In the drawings, each element or part is not necessarily drawn according to the actual scale. Obviously, the drawings described below are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0067] Figure 1 The method flowchart in an exemplary embodiment of the present application;

[0068] Figure 2 The system module structure diagram in an exemplary embodiment of the present application;

[0069] Figure 3 The interface diagram of the gas solenoid valve blockage fault diagnosis in an exemplary embodiment of the present application;

[0070] Figure 4 The interface diagram of the gas solenoid valve leakage fault diagnosis in an exemplary embodiment of the present application;

[0071] Figure 5 The collection result diagram of the burner related data in an exemplary embodiment of the present application;

[0072] Figure 6 The module structure diagram of the system in an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0073] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will be combined with the accompanying drawings to make a clear and complete description of the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0074] Herein, the suffix such as "module", "part" or "unit" used for representing an element is only for facilitating the description of the present application, and has no specific meaning by itself. Therefore, "module", "part" or "unit" can be mixedly used.

[0075] Herein, the terms "upper", "lower", "inner", "outer", "front", "back", "one end", "the other end" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for facilitating the description of the present application and simplifying the description, and do not indicate or imply that the indicated device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first", "second" are only for the purpose of description, and cannot be understood as indicating or implying relative importance.

[0076] Herein, unless otherwise explicitly specified and limited, the terms "mount", "provided with", "connected" and the like should be understood in a broad sense, for example, "connected" can be fixedly connected, or detachably connected, or integrally connected; can be mechanically connected, can be directly connected, or indirectly connected through an intermediate medium, can be the communication inside two elements. For a person of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0077] Herein, "and / or" includes any and all combinations of one or more listed associated items.

[0078] Herein, "a plurality of" means two or more, that is, it includes two, three, four, five, etc.

[0079] In the present specification, the term "about" typically means + / - 5% of the stated value, more typically + / - 4% of the stated value, more typically + / - 3% of the stated value, more typically + / - 2% of the stated value, even more typically + / - 1% of the stated value, even more typically + / - 0.5% of the stated value.

[0080] In this specification, certain embodiments can be disclosed in a format that is a range. It is to be understood that such a "range" format is used only for convenience and brevity and should be interpreted in the context of the description as a whole. Therefore, the description of a range should be considered to have specifically disclosed all possible subranges as well as individual numerical values within that range. For example, a description of a range 1-6 should be considered to have specifically disclosed the subranges 1-3, 1-4, 1-5, 2-4, 2-6, 3-6, etc., as well as the individual numbers 1, 2, 3, 4, 5, and 6 within this range. This same logic should be applied to ranges recited in the disclosure irrespective of the breadth of the range.

[0081] Since most metallurgical enterprises have been in operation for a long time, the radiant tube combustion system has been running in harsh conditions of high temperature and high pressure, high load, and strong impact for a long time. Long-term operation leads to an increase in the failure rate of devices such as electromagnetic valves, combustion equipment, radiant tubes, and heat exchangers, which in turn causes combustion imbalance, thereby seriously affecting the stable operation of the combustion system.

[0082] To this end, the prior art attempts to detect faults by using AI intelligent detection, but the applicant has noticed that due to the large system and complexity of the combustion system, the existing AI intelligent detection technology is actually very limited in terms of predictive maintenance and fault prediction. In particular, the training workload of the model is very large, while the universality of the model is very limited.

[0083] For example, the adaptive control method for a continuous annealing furnace disclosed in patent application CN117144118A adopts the following technical route: a large number of parameter libraries (furnace temperature set value, strip speed set value, fuel gas flow set value, air-fuel ratio set value, air flow set value, power load, combustion air fan power, combustion air pressure before the burner, fuel gas pressure, flow, negative pressure, flue gas temperature, etc.) are used to comprehensively predict faults, such as learning based on a large amount of data to predict air leakage, proportional valve failure, pipeline blockage, heat exchanger damage, and radiant tube damage. However, the applicant has found that this large amount of data for fault prediction requires a large amount of resources in the training and maintenance process of the model, and the prediction universality is also relatively effective.

[0084] Correspondingly, in order to reduce the pressure of model training on the basis of improving fault prediction coverage and model application range, the application proposes an AI intelligent detection method for centralized prediction of the burner blockage fault of the specific regulation mechanism of the heating furnace. Specifically, the application proposes a limited dynamic factor combination (such as the first index, the second index, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2) for focused prediction of strong correlation faults (i.e., electromagnetic valve blockage). The applicant finds that this limited dynamic factor combination can not only more efficiently and accurately predict electromagnetic valve blockage faults for a combustion system based on a double-cross proportional regulation control mode, but also indirectly predict possible associated faults, such as flue damage, excessive air volume, and heat exchanger burnout, through the prediction results of electromagnetic valve faults. That is, for the combustion system of this special regulation mechanism, the application focuses on selecting the electromagnetic valve fault that is strongly associated with other faults and is directly predicted by limited dynamic factors, thereby greatly improving the cost-effectiveness of fault prediction.

[0085] Therefore, the application can not only improve the accuracy and reliability of fault prediction by predicting specific strong correlation faults (this specific prediction mechanism for strong correlation faults to some extent avoids the technical pressure exerted by the large system and complexity of the combustion system on model training), but also indirectly judge batch faults through limited workload, thereby still reducing the real-time analysis pressure of faults to a large extent.

[0086] Especially, most of the equipment of the heating furnace is in a sealed state, and the arrangement and maintenance of the detection device arranged inside are very difficult and costly. The application can reduce the technical difficulty and cost of the detection system in the implementation and application process by using a smaller amount of detection device. For example, for some small metallurgical enterprises, the cost of AI supporting equipment arrangement that can be invested is relatively limited. This centralized prediction can complete batch fault prediction through limited cost, and thus manual analysis of the predicted faults by manpower can make the cost of fault detection and maintenance lower.

[0087] In other words, the centralized prediction mechanism proposed by the application is also beneficial for centralized data collection. From the perspective of the input end, the pressure of gas and air in the burner pipeline, the oxygen and carbon monoxide content in the tail gas are collected centrally, and the necessary data detection device required for configuration is relatively small, so that the detection system of the application is easier to implement.

[0088] In summary, the application proposes a technical route for centralized prediction of strong correlation faults based on centralized data.

[0089] Embodiment one:

[0090] Referring to Figure 1 As shown in the drawings, the present application provides a combustion balance AI intelligent detection and diagnosis method, the combustion system comprising: a heating furnace, and a burner providing a heat source for the heating furnace, and correspondingly, the method comprising steps of:

[0091] Conducting centralized fault determination on the burner, comprising:

[0092] S101: reading the number of the burner, and centrally reading the rated pressure difference, actual pressure difference and working parameters of the burner according to the number; the rated pressure difference comprises: gas standard pressure difference, air standard pressure difference; the actual pressure difference comprises: gas actual pressure difference, air actual pressure difference; the pressure difference is the pressure difference between the input side pressure and the output side pressure of the flow orifice plate in the burner; the working parameters comprise: the diameter of the flow orifice plate, the diameter ratio of the diameter of the flow orifice plate to the diameter of the pipe, the expandable coefficient, the outflow coefficient, and the fluid density;

[0093] For example, in some embodiments, a plurality of burners are arranged in the combustion system, and the corresponding rated parameters can be queried by the number (such as ID value) of the burner, such as the rated power Ps of M3AO burner is 100kW, and the corresponding rated pressure difference can be calculated according to the current design standard of the combustion system, such as the gas standard pressure difference and the air standard pressure difference corresponding to the M3AO burner are both about 500Pa.

[0094] Specifically, the series of rated pressure difference, actual pressure difference and working parameters associated with the burner can be automatically read by the number.

[0095] S102: calculating the standard gas flow Qgs and the standard air flow Qas of the burner by the gas standard pressure difference and the air standard pressure difference respectively by using the pressure difference method;

[0096] S103: calculating the actual gas flow Qg and the actual air flow Qa of the burner by the gas actual pressure difference and the air actual pressure difference respectively by using the pressure difference method;

[0097] For example, in some embodiments, the calculation formula of the pressure difference method is as follows:

[0098] Wherein, Q v is the flow of the fluid, β is the diameter ratio, d is the diameter of the flow orifice plate, ε is the expandable coefficient, C is the outflow coefficient, ΔP is the pressure difference of the corresponding fluid, and ρ is the density of the corresponding fluid; the fluid is gas or air;

[0099] For example, when Q vQg is the actual gas flow rate, ΔP is the actual gas pressure difference, and ρ is the density of the gas. When Qa is the actual air flow rate, ΔP is the actual air pressure difference, and ρ is the density of the air. v Qg is the actual gas flow rate, ΔP is the actual gas pressure difference, and ρ is the density of the gas. When Qa is the actual air flow rate, ΔP is the actual air pressure difference, and ρ is the density of the air.

[0100] Qg is the actual gas flow rate, ΔP is the actual gas pressure difference, and ρ is the density of the gas. When Qa is the actual air flow rate, ΔP is the actual air pressure difference, and ρ is the density of the air. v Qg is the actual gas flow rate, ΔP is the actual gas pressure difference, and ρ is the density of the gas. When Qa is the actual air flow rate, ΔP is the actual air pressure difference, and ρ is the density of the air. v Qg is the actual gas flow rate, ΔP is the actual gas pressure difference, and ρ is the density of the gas. When Qa is the actual air flow rate, ΔP is the actual air pressure difference, and ρ is the density of the air.

[0101] Referring to FIG. 1, the standard gas flow rate Qgs is calculated by the pressure difference method, and the standard air flow rate Qas is calculated by the pressure difference method. Figure 2 Figure 2 The actual measurement data table of the M3AO burner is shown in FIG. 2, and the standard gas flow rate Qgs of the burner M3AO is calculated by the pressure difference method to be about 20 m 3 / h, and the standard air flow rate Qas is about 110 m 3 / h.

[0102] In some embodiments, the adaptability of the fault diagnosis model is evaluated before the centralized data set is input into the fault diagnosis model, and when it is judged that the heating furnace belongs to double cross proportional regulation, the current fault diagnosis model is recommended to be used.

[0103] S104: judging the equipment type of the heating furnace according to the actual input side pressure and the actual output side pressure, wherein when the fluctuation values of the input side pressure and the output side pressure both belong to the set fluctuation threshold, the equipment type is considered to be double cross proportional regulation;

[0104] S105: when the equipment type is double cross proportional regulation, collecting a current centralized data set and inputting the centralized data set into a fault diagnosis model, the centralized data set comprising: the standard gas flow rate Qgs, the standard air flow rate Qas, the actual gas flow rate Qg, the actual air flow rate Qa, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2, and the fault diagnosis model comprising:

[0105] |Qa-Qas| ÷ Qas < A1;

[0106] |Qg-Qgs| ÷ Qgs > A2;

[0107] Cco > A3;

[0108] Co2 > A4;

[0109] ​Wherein, |Qa-Qas| ÷ Qas is the first index, |Qg-Qgs| ÷ Qgs is the second index, A1, A2, A3, A4 are respectively the first training threshold, the second training threshold, the third training threshold, and the fourth training threshold; correspondingly, a plurality of corresponding training thresholds are obtained through machine learning training;

[0110] S106: When the centralized data set meets the fault judgment model, it is determined that the fault type of the burner is electromagnetic valve blockage.

[0111] In the embodiment, a method for quickly predicting electromagnetic valve blockage based on centralized gas indicators (or limited dynamic factors) is proposed. This centralized prediction method can not only quickly predict strongly correlated faults through limited data to facilitate the rapid mobilization of maintenance resources for timely repair of electromagnetic valve faults, but also can analyze electromagnetic valve blockage to make preliminary manual observations of batch faults, such as flue damage, excessive air volume, and heat exchanger burnout, and timely allocate resources to investigate and suppress the "signs" of faults to limit further expansion of the faults.

[0112] It should be noted that the heating furnace based on double-cross proportional regulation is widely used and usually has a long production life. Moreover, because such equipment is mostly in a sealed state, it is difficult to discover faults in time through regular inspection. In addition, due to the complexity of the heating furnace system based on double-cross proportional regulation, the applicant finds that traditional AI intelligent detection methods are difficult to apply, especially in the early data collection and model training stages, which require a large amount of resources, and the final fault prediction accuracy is relatively limited. In particular, faults such as burner blockage, flue damage, excessive air volume, and heat exchanger burnout have a significant impact on the loss of the heating furnace based on double-cross proportional regulation. Once the fault prediction is incorrect, it is very easy to cause the fault to expand and increase economic losses.

[0113] In view of this special application environment, the application proposes a method for centralized prediction of burner blockage faults of a heating furnace based on double-cross proportional regulation using limited dynamic factors. The applicant finds that through centralized prediction of burner faults, other faults such as flue damage, excessive air volume, and heat exchanger burnout can be indirectly predicted by taking advantage of the strong correlation between burner faults and other faults in the heating furnace based on double-cross proportional regulation. For example, when it is found that the current burner is blocked (or the blockage is serious), the staff will manually start the fault detection of flue damage, excessive air volume, and heat exchanger burnout.

[0114] Further, in some embodiments, the method further comprises:

[0115] The step of training the corresponding training threshold by machine learning comprises:

[0116] obtaining a training sample set, the training sample set comprising: input training data, the input training data comprising: the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2; and output training data corresponding to the input training data, the output training data comprising: solenoid valve blockage data;

[0117] For example, in some embodiments, the solenoid valve blockage data can be an index describing the degree of blockage. For example, the degree of blockage can be divided into low, medium or high according to the blocked area, and for another example, the degree of blockage can be a proportional value describing the size of the blocked area, such as 10%, 20%, … 90%, 100%, and the like.

[0118] Correspondingly, in some embodiments, the output result can be blockage or no blockage. Alternatively, the output result can be the degree of blockage.

[0119] inputting the training sample set into a set model for at least one iteration training; wherein the set model is used to learn the internal correlation between the first index, the second index, the tail gas carbon monoxide content Cco, the tail gas oxygen content Co2, and the solenoid valve blockage data;

[0120] After the set model converges, outputting the predicted value interval of the first index, the second index, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2 corresponding to the solenoid valve blockage state by the set model, and setting a plurality of corresponding training thresholds according to the predicted value interval.

[0121] In some embodiments, before fault determination of the burner, further comprising the steps of:

[0122] (1) obtaining historical training thresholds set according to the set model last time;

[0123] (2) using an updated data set formed by the current burner in a first period, the updated data set comprising: input updated data, the input updated data comprising: the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2; and output updated data corresponding to the input updated data;

[0124] (3) collecting a first update data set from the update data set, inputting the update input end training data into the fault judgment model, obtaining a fault prediction result output by the fault judgment model, and determining the fault judgment model using the historical training threshold;

[0125] (4) comparing the fault prediction result with the output end update data, and calculating the prediction accuracy (such as prediction accuracy) of the fault judgment model at the present time.

[0126] In some embodiments, the method further comprises the step of: when the prediction accuracy is greater than a first set value, collecting a second update data set from the update data set;

[0127] inputting the second update data set into the setting model to update the setting model;

[0128] setting a new training threshold based on the updated setting model.

[0129] In some embodiments, the method further comprises:

[0130] Before the fault judgment of the burner, the method further comprises the step of:

[0131] judging the combustion balance state, which comprises:

[0132] collecting the excess air amount and the actual air amount in the tail gas to calculate the air excess coefficient, wherein the air excess coefficient is calculated using the following formula: α = 1 / [1-(ΔL / La)], α is the air excess coefficient, ΔL is the excess air amount, and La is the actual air amount;

[0133] judging whether the air excess coefficient belongs to a set threshold range, if yes, performing the step of judging the fault of the burner, and if not, continuing to detect the excess air amount and the actual air amount of the tail gas.

[0134] In the present embodiment, the electromagnetic valve of the burner is preferably detected when the combustion state is detected to be unbalanced through the air content in the tail gas, which can further reduce the fault monitoring pressure of the heating furnace based on the double cross proportional regulation.

[0135] Further, in some embodiments, the intelligent detection method further comprises the step of:

[0136] obtaining the fault prediction results of the burners corresponding to a plurality of different numbers output by the fault judgment model;

[0137] The fault prediction result is compared with the actual fault result (for example, the actual fault result can be found by the staff during the maintenance process); when the fault prediction result is the same as the actual fault result, the corresponding fault prediction result is considered to be accurate; when the fault prediction result is different from the actual fault result, the corresponding fault prediction result is considered to be incorrect.

[0138] When the proportion of the number of burners with incorrect fault prediction results to the total number of burners is greater than a first set ratio, a first prompt signal is output to prompt the user to verify the fault judgment model (e.g., update the model using the latest updated dataset).

[0139] When the number of burners with incorrect fault prediction results accounts for a proportion of the total number of burners that is greater than a second set ratio and less than or equal to a first set ratio, a second prompt signal is output to prompt the user to check the gas detectors associated with the corresponding burner (e.g., detectors used to detect the content of gas, air flow, oxygen, and carbon monoxide in exhaust gas, respectively).

[0140] In this invention, by employing a limited set of gas parameters to centrally predict burner blockage faults, the difficulty of fault prediction can be effectively reduced. Conversely, when the model's prediction results show varying degrees of error, this also helps guide staff to analyze the true causes of the errors in a timely manner.

[0141] For example, in some embodiments, when only a small number of burner fault prediction results show false alarms, staff can promptly inspect a small number of gas detectors to maintain the fault detection method in a timely manner.

[0142] Example 2:

[0143] See Figure 6 As shown, the present invention also provides an intelligent detection and diagnosis system, which is applied to a combustion system. The combustion system includes a heating furnace and a burner that provides a heat source for the heating furnace. Correspondingly, the system includes a centralized fault determination subsystem, which includes:

[0144] The reading module 101 is used to read the burner's serial number and, based on the serial number, centrally read the burner's rated differential pressure, actual differential pressure, and operating parameters. The rated differential pressure includes: standard differential pressure of gas and standard differential pressure of air. The actual differential pressure includes: actual differential pressure of gas and actual differential pressure of air. The differential pressure is the pressure difference between the input and output pressures of the flow orifice plate in the burner's pipeline. The operating parameters include: the diameter of the flow orifice plate, the ratio of the diameter of the flow orifice plate to that of the pipeline, the expansion coefficient, the discharge coefficient, and the fluid density.

[0145] The first calculation module 102 is configured to calculate the standard gas flow Qgs and the standard air flow Qas of the burner by the pressure difference method through the standard pressure difference of the gas and the standard pressure difference of the air respectively;

[0146] The second calculation module 103 is configured to calculate the actual gas flow Qg and the actual air flow Qa of the burner by the pressure difference method through the actual pressure difference of the gas and the actual pressure difference of the air;

[0147] The type judgment module 104 is configured to judge the equipment type of the heating furnace according to the actual input side pressure and the actual output side pressure, wherein when the fluctuation values of the input side pressure and the output side pressure both belong to the set fluctuation threshold, it is considered that the equipment type is double-cross proportional regulation;

[0148] The data input module 105 is configured to consider that the burner meets the concentration prediction condition when the equipment type is double-cross proportional regulation, and collect the current concentration data set to input the concentration data set into a fault judgment model, wherein the concentration data set includes the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2, and the fault judgment model includes:

[0149] |Qa-Qas|÷Qas<A1;

[0150] |Qg-Qgs|÷Qgs>A2;

[0151] Cco>A3;

[0152] Co2>A4;

[0153] Wherein, |Qa-Qas|÷Qas is the first index, |Qg-Qgs|÷Qgs is the second index, A1, A2, A3, and A4 are the first training threshold, the second training threshold, the third training threshold, and the fourth training threshold respectively; correspondingly, a plurality of corresponding training thresholds are obtained by machine learning training;

[0154] The fault judgment module 106 is configured to determine that the fault type of the burner is electromagnetic valve blockage when the concentration data set meets the fault judgment model.

[0155] In some embodiments, the data input module further includes:

[0156] a sample acquisition unit configured to acquire a training sample set, the training sample set comprising: input training data, the input training data comprising: the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2; and output training data corresponding to the input training data, the output training data comprising: solenoid valve blockage data;

[0157] an iteration unit configured to input the training sample set into a set model to perform at least one iteration training, wherein the set model is configured to learn an internal correlation between the first index, the second index, the tail gas carbon monoxide content Cco, the tail gas oxygen content Co2, and the solenoid valve blockage data;

[0158] a threshold setting unit configured to output, by the set model, a predicted value interval of the first index, the second index, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2 corresponding to the solenoid valve blockage state after the set model converges, and set a plurality of training thresholds corresponding to the predicted value interval.

[0159] In some embodiments, the system further comprises an accuracy prediction subsystem configured to perform the following steps:

[0160] (1) acquiring historical training thresholds set according to the set model last time;

[0161] (2) using an updated data set formed by the burner in a first time period, the updated data set comprising: input updated data, the input updated data comprising: the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2; and output updated data corresponding to the input updated data, the output updated data comprising: solenoid valve blockage data;

[0162] (3) collecting a first updated data set from the updated data set and inputting the corresponding input updated data into the fault judgment model to obtain a fault prediction result output by the fault judgment model, wherein the fault judgment model uses the historical training thresholds for judgment;

[0163] (4) comparing the fault prediction result with the output updated data to calculate a current prediction accuracy of the fault judgment model.

[0164] In some embodiments, a linear regression algorithm, a K-means algorithm, and a neural network algorithm can be used to complete corresponding model training.

[0165] The application also provides an electronic device, comprising a memory and a processor; the memory is used for storing a program; the processor is used for executing the program to realize each step of the intelligent detection method according to any one of the application.

[0166] The application also provides a computer program product, comprising a computer program, when the computer program is executed by one or more processors, causing the one or more processors to execute the intelligent detection method according to any one of the application.

[0167] It can be understood that the sequence of steps described in the application is only exemplary, and in fact, the steps in the application can be executed according to the actual situation, for example, the steps of S102, S103, S104 and S105 can be executed synchronously or in any order.

[0168] It should be noted that in this document, the terms "comprise", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "comprises a" does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0169] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by software and necessary general hardware platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes a plurality of instructions for making a computer terminal (which can be a mobile phone, computer, server or network equipment) execute the method described in each embodiment of the application.

[0170] The embodiments of the application are described above in combination with the drawings, but the application is not limited to the above specific embodiments, and the above specific embodiments are only illustrative, not restrictive, and those skilled in the art can make many forms under the inspiration of the application without departing from the purpose of the application and the scope protected by the claims. These are all within the protection of the application.

Claims

1. A combustion balance AI intelligent detection and diagnosis method, characterized in that, The method is applied to a combustion system, the combustion system comprising a heating furnace and a burner providing a heat source for the heating furnace, and the method detects an electromagnetic valve of the burner when detecting combustion state imbalance through air content in tail gas; The method comprises: S101: reading a number of the burner, and centrally reading a rated pressure difference, an actual pressure difference and working parameters of the burner according to the number; the rated pressure difference comprises a gas standard pressure difference and an air standard pressure difference; the actual pressure difference comprises a gas actual pressure difference and an air actual pressure difference; the gas standard pressure difference, the air standard pressure difference, the gas actual pressure difference and the air actual pressure difference are pressure difference values between input side pressure and output side pressure of a flow orifice plate in a pipeline of the burner; and the working parameters comprise a diameter of the flow orifice plate, a diameter ratio of the flow orifice plate to the pipeline, an expandable coefficient, a flow-out coefficient and a fluid density; S102: calculating standard gas flow Qgs and standard air flow Qas of the burner through the gas standard pressure difference and the air standard pressure difference respectively by using a pressure difference method; S103: calculating actual gas flow Qg and actual air flow Qa of the burner through the gas actual pressure difference and the air actual pressure difference by using the pressure difference method; S104: judging a device type of the heating furnace according to actual input side pressure and actual output side pressure, wherein when fluctuation values of the input side pressure and the output side pressure both belong to a set fluctuation threshold value, it is considered that the device type is double-cross proportional regulation; S105: when the device type is double-cross proportional regulation, it is considered that the burner meets a central prediction condition, and current central data set is collected to input the central data set into a fault judgment model, the central data set comprising the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, tail gas carbon monoxide content Cco and tail gas oxygen content Co2, and the fault judgment model comprising: |Qa-Qas|÷Qas<A1; |Qg-Qgs|÷Qgs>A2; Cco>A3; Co2>A4; wherein |Qa-Qas|÷Qas is a first index, |Qg-Qgs|÷Qgs is a second index, A1, A2, A3 and A4 are respectively a first training threshold value, a second training threshold value, a third training threshold value and a fourth training threshold value; and the first training threshold value, the second training threshold value, the third training threshold value and the fourth training threshold value are obtained by machine learning training; S106: when the central data set meets the fault judgment model, it is judged that a fault type of the burner is electromagnetic valve blockage.

2. The method of claim 1, wherein, Further comprising: obtaining fault prediction results of burners corresponding to a plurality of different numbers and output by the fault judgment model; comparing the fault prediction results with actual fault results; and When the fault prediction result is the same as the actual fault result, it is considered that the corresponding fault prediction result is accurate; When the fault prediction result is not the same as the actual fault result, it is considered that the corresponding fault prediction result is incorrect; When the number of the burners with incorrect fault prediction results accounts for more than a first set ratio of the total number of the burners, a first prompt signal is output to prompt the user to verify the fault determination model; When the number of the burners with incorrect fault prediction results accounts for more than a second set ratio and less than or equal to the first set ratio of the total number of the burners, a second prompt signal is output to prompt the user to detect the gas detector associated with the corresponding burner.

3. The method of claim 2, wherein, The step of obtaining a plurality of corresponding training thresholds through machine learning training includes: obtaining a training sample set, which includes input training data including the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco and the tail gas oxygen content Co2, and output training data corresponding to the input training data, the output training data including solenoid valve blockage data; inputting the training sample set into a set model for at least one iteration training; wherein the set model is used to learn the internal correlation between the first index, the second index, the tail gas carbon monoxide content Cco, the tail gas oxygen content Co2 and the solenoid valve blockage data; After the set model converges, the set model outputs the prediction value interval of the first index, the second index, the tail gas carbon monoxide content Cco and the tail gas oxygen content Co2 corresponding to the solenoid valve blockage state, and sets the first training threshold, the second training threshold, the third training threshold and the fourth training threshold according to the prediction value interval.

4. The method of claim 3, wherein, Before determining the fault of the burner, the method further includes the steps of: (1) obtaining historical training thresholds set according to the set model last time; (2) obtaining an updated data set formed by the burner in a first time period, the updated data set including input updated data including the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco and the tail gas oxygen content Co2, and output updated data corresponding to the input updated data, the output updated data including solenoid valve blockage data; (3) collecting a first updated data set from the updated data set and inputting the corresponding input updated data into the fault determination model to obtain the fault prediction result output by the fault determination model; wherein the fault determination model uses the historical training thresholds for determination; (4) comparing the fault prediction result with the output updated data to calculate the current prediction accuracy of the fault determination model.

5. The method of claim 4, wherein, Further comprising the steps of: when the prediction accuracy is greater than a first set value, then collecting a second update dataset from the update dataset; inputting the second update dataset into the set model to update train the set model; based on the updated set model, setting new first training threshold, second training threshold, third training threshold and fourth training threshold again.

6. The method of claim 1, wherein, Before fault determination of the burner, further comprising the steps of: determining the combustion balance state; it includes: The excess air coefficient is calculated by collecting the excess air amount and the actual air amount in the tail gas, wherein the excess air coefficient is calculated by the following formula: a = 1 / [1- (L / La)], wherein a is the excess air coefficient, L is the excess air amount, and La is the actual air amount. L / La)], a is the excess air coefficient, L is the excess air amount, and La is the actual air amount. determine whether the excess air coefficient belongs to the set threshold range, if yes, execute the step of determining the fault of the burner; if not, continue to detect the excess air amount and the actual air amount of the tail gas.

7. A combustion balance AI intelligent detection and diagnosis system, characterized in that, The combustion balance AI intelligent detection and diagnosis system is used to implement the method of any one of claims 1-6, the AI intelligent detection and diagnosis system is applied to a combustion system, the combustion system includes a heating furnace and a burner providing a heat source for the heating furnace, wherein the combustion balance AI intelligent detection and diagnosis system detects the solenoid valve of the burner when the combustion state is detected by the air content in the tail gas. The system includes a centralized fault determination subsystem, which includes: a reading module for reading the number of the burner and centrally reading the rated pressure difference, actual pressure difference and working parameters of the burner according to the number; the rated pressure difference includes gas standard pressure difference and air standard pressure difference; the actual pressure difference includes gas actual pressure difference and air actual pressure difference; the gas standard pressure difference, the air standard pressure difference, the gas actual pressure difference and the air actual pressure difference are the pressure difference between the input side pressure and the output side pressure of the flow orifice plate in the pipeline of the burner; the working parameters include the diameter of the flow orifice plate, the diameter ratio of the flow orifice plate to the pipeline, the expandable coefficient, the outflow coefficient and the fluid density; a first calculation module for calculating the standard gas flow Qgs and the standard air flow Qas of the burner by the gas standard pressure difference and the air standard pressure difference respectively by using the pressure difference method; a second calculation module for calculating the actual gas flow Qg and the actual air flow Qa of the burner by the gas actual pressure difference and the air actual pressure difference by using the pressure difference method; a type judgment module for judging the equipment type of the heating furnace according to the actual input side pressure and the actual output side pressure, wherein when the fluctuation values of the input side pressure and the output side pressure both belong to the set fluctuation threshold, it is considered that the equipment type is double cross proportional regulation. The data input module is configured to, when the device type is a double cross proportional regulation, consider that the burner meets a centralization prediction condition, and collect a current centralization data set to input the centralization data set into a fault determination model, the centralization data set including the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, tail gas carbon monoxide content Cco, and tail gas oxygen content Co2, and the fault determination model including: |Qa-Qas| ÷ Qas < A1; |Qg-Qgs| ÷ Qgs > A2; Cco > A3; Co2 > A4; wherein |Qa-Qas| ÷ Qas is a first index, |Qg-Qgs| ÷ Qgs is a second index, A1, A2, A3, and A4 are respectively a first training threshold value, a second training threshold value, a third training threshold value, and a fourth training threshold value; the first training threshold value, the second training threshold value, the third training threshold value, and the fourth training threshold value are obtained through machine learning training; The fault determination module is configured to, when the centralization data set meets the fault determination model, determine that a fault type of the burner is an electromagnetic valve blockage.

8. The system of claim 7, wherein, The data input module further includes: The sample acquisition unit is configured to acquire a training sample set, the training sample set including input end training data including the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2, and output end training data corresponding to the input end training data, the output end training data including electromagnetic valve blockage data; The iteration unit is configured to input the training sample set into a setting model to perform at least one iteration training; wherein the setting model is configured to learn an internal correlation between the first index, the second index, the tail gas carbon monoxide content Cco, the tail gas oxygen content Co2, and the electromagnetic valve blockage data; The threshold value setting unit is configured to, after the setting model converges, output, through the setting model, a prediction value interval of the first index, the second index, the tail gas carbon monoxide content Cco, and the tail gas oxygen content Co2 corresponding to the electromagnetic valve blockage state, and set the first training threshold value, the second training threshold value, the third training threshold value, and the fourth training threshold value according to the prediction value interval.

9. The system of claim 8, wherein, Further including: The accuracy prediction subsystem is configured to perform the following steps: (1) acquiring a historical training threshold value set last time according to the setting model; (2) using the current burner to form an updated data set in the first period, the updated data set includes: input update data, which includes: the standard gas flow Qgs, the standard air flow Qas, the actual gas flow Qg, the actual air flow Qa, the tail gas carbon monoxide content Cco and the tail gas oxygen content Co2; and the output update data corresponding to the input update data, the output update data includes: solenoid valve blockage data; (3) the first update data set is collected from the updated data set, and the corresponding input update data is input into the fault judgment model, and the fault prediction result output by the fault judgment model is obtained; wherein the fault judgment model uses the historical training threshold for judgment; (4) according to the fault prediction result and the output update data, the prediction accuracy of the fault judgment model is calculated.

10. An electronic device, comprising: Comprising: a memory and a processor; the memory is used for storing programs; the processor is used for executing the programs, realizing the steps of the combustion balance AI intelligent detection and diagnosis method in any one of claims 1-6.

11. A computer program product, characterised in that, comprising a computer program, when the computer program is executed by one or more processors, causing the one or more processors to execute the combustion balance AI intelligent detection and diagnosis method in any one of claims 1-6.

Images