Solid waste heat treatment prediction and anomaly pre-judgment method based on deep learning

Abstract

The invention discloses a solid waste heat treatment prediction and anomaly prejudgment method based on deep learning, and belongs to the technical field of solid waste heat treatment. The problems of comprehensive monitoring and intelligent early warning in the solid waste heat treatment process are solved. The method comprises the following steps: constructing a first hidden layer of a solid waste heat treatment prediction model based on deep learning; constructing a second hidden layer of the solid waste heat treatment prediction model based on deep learning; constructing an output layer of the solid waste heat treatment prediction model based on deep learning; designing a loss function of the solid waste heat treatment prediction model based on deep learning; the input layer, the first hidden layer, the second hidden layer and the output layer are sequentially connected to obtain a solid waste heat treatment prediction model based on deep learning; and a solid waste heat treatment abnormity pre-judgment method is constructed, analysis is carried out from the aspects of substance conversion abnormity, substance distribution abnormity and energy conversion abnormity when the solid waste heat treatment abnormity problem is considered, and an abnormity diagnosis and judgment scheme is given.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of solid waste heat treatment, and particularly relates to a solid waste heat treatment prediction and abnormality prediction method based on deep learning. BACKGROUND

[0002] The solid waste heat treatment process involves complex conversion of matter and energy. First, the components of solid waste have high heterogeneity and uncertainty, including combustibles, moisture, inorganics, and other components, the proportions of which change over time and source. Second, drying, pyrolysis, combustion, and other physical and chemical processes occur simultaneously during the heat treatment process, and these processes have significant coupling effects. The drying process affects the subsequent pyrolysis and combustion effects, the composition of pyrolysis products affects the combustion characteristics, and the combustion heat release in turn affects drying and pyrolysis. At the same time, the distribution of temperature field, flow field, and concentration field also interact, forming a complex multi-field coupling. In addition, there are multi-scale processes such as heat and mass transfer, chemical reaction, and phase change in the furnace, and the nonlinear characteristics and dynamic changes of these processes bring great challenges to process control.

[0003] In the solid waste heat treatment process, conversion efficiency directly reflects the degree of material conversion and is related to the completeness of incineration and the generation of pollutants. For example, carbon conversion rate affects the completeness of combustion and carbon dioxide emissions, and sulfur and chlorine conversion rates are closely related to the generation of acid gases. Product distribution determines the pollution control strategy and resource utilization direction, and accurate prediction helps to optimize the subsequent treatment process. Energy conversion reflects the level of energy utilization, which is of great significance to improve system efficiency and reduce operating costs. Accurate prediction of these indicators can provide decision-making basis for process optimization and intelligent control, and is of great significance to improve system operation efficiency, reduce pollutant emissions, and ensure safe and stable operation.

[0004] The current prediction research on key indicators in the solid waste heat treatment process mainly has the following shortcomings: First, the prediction accuracy is generally not high, especially when the working condition fluctuates greatly, the prediction result deviates significantly from the actual value. Second, the existing models consider only a single influencing factor, often focusing on only some main process parameters, while ignoring the influence of important factors such as material properties and environmental conditions, resulting in poor adaptability of the model to complex working conditions. In addition, most models only predict a single or a few indicators, and lack comprehensive prediction capability for conversion efficiency, product distribution, and energy conversion of multiple key indicators. It is urgent to establish a comprehensive model that considers more comprehensive factors and has better prediction effect to meet the actual engineering needs.

[0005] On the other hand, in the operation management of the solid waste heat treatment process, due to the complexity and uncertainty of the process, how to discover and predict the operation abnormity in time has always been a research difficulty. The existing abnormal diagnosis method often only focuses on the fluctuation of a single index, lacks systematic evaluation of multi-dimensional abnormity such as material conversion, product distribution and energy conversion, and is difficult to accurately reflect the overall operation state of the system. Therefore, it is urgent to establish a scientific solid waste heat treatment abnormal prediction method to realize the comprehensive monitoring and intelligent early warning of the solid waste heat treatment process, which can provide reliable basis for process adjustment and fault prevention. SUMMARY

[0006] The problem to be solved by the present application is to realize comprehensive monitoring and intelligent early warning of the solid waste heat treatment process, and a solid waste heat treatment prediction and abnormal prediction method based on deep learning is proposed.

[0007] To achieve the above-mentioned purpose, the present application realizes by the following technical scheme:

[0008] A solid waste heat treatment prediction and abnormal prediction method based on deep learning, comprising the following steps:

[0009] S1. Determine the input parameters of the input layer of the solid waste heat treatment prediction model based on deep learning, including material characteristic parameters, process operation parameters and process parameters;

[0010] S2. Construct the first hidden layer of the solid waste heat treatment prediction model based on deep learning;

[0011] S3. Construct the second hidden layer of the solid waste heat treatment prediction model based on deep learning;

[0012] S4. Construct the output layer of the solid waste heat treatment prediction model based on deep learning, divide the output layer into conversion efficiency group, product distribution group and energy conversion group, and determine the output parameters of the output layer of the solid waste heat treatment prediction model based on deep learning;

[0013] S5. Design the loss function of the solid waste heat treatment prediction model based on deep learning;

[0014] S6. Connect the input layer, the first hidden layer, the second hidden layer and the output layer of the solid waste heat treatment prediction model based on deep learning in sequence to obtain the solid waste heat treatment prediction model based on deep learning;

[0015] S7. Collect data through the solid waste treatment plant and the automation control system, normalize the data, divide the data into a training set, a validation set and a test set according to a ratio of 7:2:1, train, validate and test the deep learning-based solid waste thermal treatment prediction model obtained in step S6, and obtain a trained deep learning-based solid waste thermal treatment prediction model;

[0016] S8. After the prediction result of the trained deep learning-based solid waste thermal treatment prediction model is denormalized, the prediction result of the deep learning-based solid waste thermal treatment prediction model with actual physical quantities is obtained, an abnormality prediction method for solid waste thermal treatment is constructed, and when the abnormality problem of solid waste thermal treatment is considered, the abnormality problem is analyzed from the aspects of material conversion abnormality, material distribution abnormality and energy conversion abnormality, and an abnormality diagnosis and judgment scheme is given.

[0017] Further, the number of nodes of the input layer in step S1 is determined as 35, which are respectively denoted as X1-X 35 ;

[0018] X1~X 14 is a material characteristic parameter, and the specific parameters are as follows: X1 is moisture content, X2 is ash content, X3 is combustible content, X4 is carbon content, X5 is hydrogen content, X6 is oxygen content, X7 is nitrogen content, X8 is sulfur content, X9 is chlorine content, X 10 is calorific value, X 11 is bulk density, X 12 is particle size, X 13 is sand content, and X 14 is metal content;

[0019] X 15 ~X 25 is a process operation parameter, and the specific parameters are as follows: X 15 is grate speed, X 16 is feeding rate, X 17 is primary air volume, X 18 is secondary air volume, X 19 is furnace temperature, X 20 is furnace negative pressure, X 21 is flue temperature, X 22 is steam temperature, X 23 is steam pressure, X 24 is steam flow, and X 25 is ambient temperature;

[0020] X 26 ~X 35 is a process parameter, and the specific parameters are as follows: X 26 is feed moisture content, X 27for moisture content of the outfeed, X 28 for volatile matter of the infeed, X 29 for volatile matter of the outfeed, X 30 for mass flow rate of the infeed, X 31 for mass flow rate of the outfeed, X 32 for heating value of the infeed, X 33 for heating value of the outfeed, X 34 for flue gas oxygen concentration, X 35 for ash content.

[0021] Further, step S2 divides the nodes of the first hidden layer into a material property group, a reaction property group, and an energy property group. The material property group focuses on the changes in the physical and chemical properties of the material during the heat treatment process, contains 30 neuron nodes, and the ith neuron node has a value of 1-30. The reaction property group focuses on describing the degree of progress of various chemical reactions, contains 25 neuron nodes, and the ith neuron node has a value of 31-55. The energy property group focuses on representing the energy conversion and transmission characteristics, contains 25 neuron nodes, and the ith neuron node has a value of 56-80.

[0022] The specific implementation method of step S2 includes the following steps:

[0023] S2.1. Temperature correction and material moisture correction are introduced in the construction of the material property group. The temperature correction considers the driving effect of temperature on the conversion rate of the material and reflects the differences in the conversion law of the material in different temperature intervals. The material moisture correction considers the influence of the evaporation process of the material on the subsequent conversion.

[0024] S2.2. Mass conversion term is introduced in the construction of the reaction property group, which directly reflects the degree of material reaction and embodies the depth of reaction. The air excess term is added to represent the sufficiency of the combustion environment and reflect the influence of the oxidation reaction condition on the conversion effect.

[0025] S2.3. Heat value conversion term is introduced in the construction of the energy property group, which directly represents the energy conversion effect. The temperature gradient term is added to reflect the driving force and transmission strength of heat transfer and describe the spatial distribution law of energy.

[0026] Further, step S3 divides the nodes of the second hidden layer into a drying process group, a pyrolysis process group, and a combustion process group. The drying process group describes the influence of the drying process on the subsequent pyrolysis and combustion, contains 15 neuron nodes, and the ith neuron node has a value of 1-15. The pyrolysis process group describes the composition change characteristics of the pyrolysis products, contains 20 neuron nodes, and the ith neuron node has a value of 16-35. The combustion process group describes the process characteristics of the final oxidation stage, contains 15 neuron nodes, and the ith neuron node has a value of 36-50.

[0027] The specific implementation method of step S3 includes the following steps:

[0028] S3.1. Constructing a drying process group by characterizing the process characteristics of water removal in solid waste, reflecting the driving effect of temperature on water evaporation, describing the influence of the drying process on subsequent pyrolysis and combustion;

[0029] S3.2. Constructing a pyrolysis process group by characterizing the kinetic characteristics of organic matter pyrolysis decomposition, reflecting the influence law of temperature on volatile matter, describing the composition change characteristics of pyrolysis products, and realizing the extraction of key characteristics of the pyrolysis stage;

[0030] S3.3. Constructing a combustion process group to characterize the oxidation process of fixed carbon and volatile matter, reflecting the influence of oxygen concentration on combustion effect, and describing the process characteristics of the final oxidation stage.

[0031] Further, step S4 is based on the evaluation requirements of the solid waste heat treatment process, and the output layer is divided into a conversion efficiency group, a product distribution group and an energy conversion group; the conversion efficiency group index reflects the integrity and sufficiency of material conversion, the product distribution group index reveals the migration law of material and environmental impact, and the energy conversion group index embodies the energy utilization level;

[0032] The conversion efficiency group index includes Y1-Y6, Y1 is the moisture conversion rate, Y2 is the volatile conversion rate, Y3 is the fixed carbon conversion rate, Y4 is the carbon conversion rate, Y5 is the sulfur conversion rate, and Y6 is the chlorine conversion rate;

[0033] The product distribution group index includes the material distribution proportion index Y 71 ~ Y 75 and the residue distribution index Y8~ Y9, Y 71 is the mass distribution proportion of moisture after drying into flue gas, Y 72 is the mass distribution proportion of volatile matter after pyrolysis into flue gas, Y 73 is the distribution proportion of carbon element converted into carbon dioxide into flue gas, Y 74 is the distribution proportion of sulfur element converted into sulfur dioxide into flue gas, Y 75 is the distribution proportion of chlorine element converted into hydrogen chloride into flue gas; Y8 is the mass distribution proportion of fixed carbon remaining in the slag after combustion, and Y9 is the mass distribution proportion of fly ash component into fly ash after fixed carbon combustion;

[0034] The energy conversion group index includes Y 10 ~Y 12 , Y 10 is the thermal efficiency, Y 11 is the combustion efficiency, and Y 12 is the heat loss rate.

[0035] Further, the prediction loss and the heat treatment loss are considered in the loss function of step S5, and the specific implementation method includes the following steps:

[0036] S5.1. Constructing a prediction loss based on a conversion efficiency prediction loss, a product distribution prediction loss, and an energy conversion prediction loss, and the expression is:

[0037]

[0038] wherein, is the conversion efficiency prediction loss; is the product distribution prediction loss; is the energy conversion prediction loss;

[0039]

[0040] wherein, is a mean square error calculation function; is a different index output value in the conversion efficiency group predicted by the model; is a different index output value in the conversion efficiency group actually measured;

[0041]

[0042] wherein, is a different index output value in the product distribution group predicted by the model; is a different index output value in the product distribution group actually measured;

[0043]

[0044] wherein, is a different index output value in the energy conversion group predicted by the model; is a different index output value in the energy conversion group actually measured;

[0045] S5.2. Constructing a heat treatment loss based on a material conversion constraint, a composition conversion constraint, and a heat conversion constraint , and the expression is:

[0046]

[0047] wherein, is the material conversion constraint, is a weight corresponding to the material conversion constraint; is the composition conversion constraint, is a weight corresponding to the composition conversion constraint; is the heat conversion constraint, is a weight corresponding to the heat conversion constraint;

[0048] The material conversion constraint is established as:

[0049]

[0050] wherein, is the volatile matter conversion term; is the fixed carbon allocation term;

[0051] The component conversion constraint is established as:

[0052]

[0053] wherein, is the sulfur element conversion constraint, is the chlorine element conversion constraint;

[0054] The heat conversion constraint is established as:

[0055] ;

[0056] S5.3. Constructing total loss based on predicted loss and heat treatment loss ;

[0057]

[0058] wherein, is the weight corresponding to the predicted loss, is the weight corresponding to the heat treatment loss, which is determined by expert experience.

[0059] Further, the model training in step S7 adopts the back propagation algorithm; in the training process, the model calculates the predicted value through forward propagation, substitutes the predicted value and the true value into the loss function to calculate the loss, and then updates the network parameters through back propagation; to prevent overfitting, the early stopping strategy is adopted, and the training is stopped when the validation set loss does not decrease for 10 consecutive epochs; the output result is restored to the actual physical quantity through inverse normalization.

[0060] Further, the specific implementation method of step S8 includes the following steps:

[0061] S8.1. In the solid waste heat treatment anomaly identification process, the material conversion anomaly degree reflects the independent deviation degree of each process parameter, which is helpful for accurately positioning the abnormal link and taking targeted adjustment, and the material conversion anomaly degree The calculation formula is:

[0062]

[0063] wherein, is the different index output value predicted by the model, including , , ; is the corresponding standard value, determined by expert experience, design documents, experimental research methods, etc. is the corresponding standard value, determined by expert experience, design documents, experimental research methods, etc. is the corresponding allowable fluctuation range in the formula for the degree of abnormality of the conversion of matter, determined by expert experience, design documents, experimental research, etc. is the corresponding standard value, determined by expert experience, design documents, experimental research methods, etc. is the corresponding standard value, determined by expert experience, design documents, experimental research methods, etc. is the corresponding distribution weight coefficient.

[0064] S8.2. The degree of abnormality of matter distribution includes the degree of abnormality of flue gas distribution and the degree of abnormality of solid distribution.

[0065] The degree of abnormality of flue gas distribution is calculated by considering the distribution deviation of each component in the flue gas in the processes of water vaporization, volatile decomposition and analysis, carbon oxidation, sulfur oxidation, and hydrogen chloride generation, and the expression is:

[0066]

[0067] wherein, is the distribution weight coefficient, s2=71, 72, 73, 74, 75. is the corresponding standard value, determined by expert experience, design documents, experimental research methods, etc. is the corresponding standard value, determined by expert experience, design documents, experimental research methods, etc. is the corresponding allowable fluctuation range in the formula for the degree of abnormality of flue gas distribution.

[0068] The degree of abnormality of solid distribution is calculated by considering the distribution balance state of inorganic matter in slag and fly ash, and the influence of process parameters such as combustion temperature, residence time, and gas-solid separation efficiency on the slag-fly ash ratio, and the expression is:

[0069]

[0070] wherein, , are the distribution weight coefficients of slag and fly ash, respectively. , are the corresponding allowable fluctuation ranges of the distribution weight coefficients of slag and fly ash in the formula for the degree of abnormality of solid distribution, respectively.

[0071] S8.3. The degree of abnormality of energy conversion includes the energy abnormality and the energy balance abnormality .

[0072] The energy abnormality is calculated by considering the influence of factors such as the heat value fluctuation of solid waste, the distribution of combustion air, the temperature distribution of the furnace, and the heat exchange efficiency on the system thermal efficiency, the combustion efficiency, and the heat loss, and the expression is:

[0073] ​​​

[0074] wherein, is the energy index weight, s3=10, 11, 12; is the corresponding allowable fluctuation range in the energy anomaly degree formula;

[0075] Considering the balance relationship between the effective heat output and the heat loss of the system, combining the influence of the furnace wall insulation performance, the cooling system efficiency and other factors, the energy balance anomaly degree is calculated , and the expression is:

[0076]

[0077] wherein, is the energy balance allowable error, which is determined by expert experience, design documents, experimental research and other methods;

[0078] S8.4. Constructing a comprehensive anomaly index ZH;

[0079]

[0080] wherein, , , , , respectively, are the corresponding process weight coefficients, which are determined by expert experience, design documents, experimental research and other methods; , , , ,

[0081] S8.5. Abnormal diagnosis and determination are carried out based on the comprehensive anomaly index obtained in step S4;

[0082] The anomaly level is divided into four levels, namely normal operation, slight anomaly, moderate anomaly and serious anomaly, wherein the critical comprehensive anomaly indexes corresponding to the slight anomaly, the moderate anomaly and the serious anomaly are respectively , , ; then is normal operation; is slight anomaly; is moderate anomaly; is serious anomaly.

[0083] The beneficial effects of the present application are:

[0084] ​The solid waste heat treatment prediction and abnormality prediction method based on deep learning provided by the application analyzes the solid waste heat treatment process from the aspects of conversion efficiency, product distribution and energy conversion. The conversion efficiency can reflect the completeness and sufficiency of material conversion, the product distribution can reveal the migration rule of the material and the environmental impact, and the energy conversion reflects the energy utilization level. The three aspects constitute a complete characterization system of the solid waste heat treatment process, which can comprehensively reflect the operation effect of the heat treatment process and provide an important basis for process optimization and control.

[0085] The solid waste heat treatment prediction and abnormality prediction method based on deep learning provided by the application is designed and optimized for the characteristics of the solid waste heat treatment in the input parameters, the first hidden layer, the second hidden layer and the output layer, so that the scientificity and rationality of the prediction of the conversion efficiency, the product distribution and the energy conversion in the solid waste heat treatment process are realized. The established deep learning model can determine the state of the solid waste heat treatment from the aspects of material conversion abnormality, material distribution abnormality and energy conversion abnormality, so as to comprehensively evaluate the performance of the system, discover and handle potential operation problems in time, and ensure the continuity and safety of the solid waste treatment process. BRIEF DESCRIPTION OF DRAWINGS

[0086] Figure 1 The flowchart of the solid waste heat treatment prediction and abnormality prediction method based on deep learning provided by the application is shown in the figure.

[0087] Figure 2 The loss curve diagram of the solid waste heat treatment prediction model training process provided by the application is shown in the figure.

[0088] Figure 3 The comparison diagram of the predicted value and the true value of the solid waste heat treatment prediction model provided by the application is shown in the figure.

[0089] Figure 4 The calculation result diagram of the comprehensive abnormality index provided by the application is shown in the figure. DETAILED DESCRIPTION

[0090] In order to make the purpose, technical scheme and advantages of the application clearer, the application is further described in detail below in combination with the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and are not used to limit the application, that is, the described specific embodiments are only part of the embodiments of the application, but not all the specific embodiments. The components of the specific embodiments of the application described and shown in the drawings herein can be arranged and designed in various different configurations, and the application can also have other embodiments.

[0091] Therefore, the following detailed description of the specific embodiments of the application provided in the drawings is not intended to limit the scope of the claimed application, but merely represents selected specific embodiments of the application. Based on the specific embodiments of the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the protection of the application.

[0092] In order to further understand the invention content, characteristics and effects of the present application, the following specific embodiments are exemplified, and the accompanying drawings are provided Figure 1 -Appendix Figure 4 The detailed description is as follows:

[0093] Example 1:

[0094] A solid waste heat treatment prediction and abnormal prediction method based on deep learning, comprising the following steps:

[0095] S1. Determine the input parameters of the input layer of the solid waste heat treatment prediction model based on deep learning, including material characteristic parameters, process operation parameters and process parameters;

[0096] The material characteristic parameters reflect the basic properties of solid waste, including composition (moisture, ash, combustible material, etc.), elemental composition (C, H, O, N, S, Cl) and physical properties (calorific value, bulk density, etc.), which directly affect the material conversion process; The process operation parameters include the controllable operating conditions such as grate speed, air volume, temperature and pressure, which determine the reaction environment; The process parameters reflect the dynamic change characteristics of matter and energy in the conversion process, such as moisture content, volatile matter, mass flow rate and calorific value of the inlet and outlet materials. This classification method comprehensively considers the key factors affecting the process characteristics, which helps the model to establish the internal relationship between material properties, operating conditions and process characteristics.

[0097] Further, the number of nodes of the input layer in step S1 is determined as 35, respectively denoted as X1-X 35 ;

[0098] X1~X 14 are material characteristic parameters, and the specific parameters are as follows: X1 is moisture content, X2 is ash content, X3 is combustible content, X4 is carbon content, X5 is hydrogen content, X6 is oxygen content, X7 is nitrogen content, X8 is sulfur content, X9 is chlorine content, X 10 is calorific value, X 11 is bulk density, X 12 is particle size, X 13 is sand content, and X 14 is metal content;

[0099] X 15 ~X 25is the process operating parameter, and the corresponding specific parameters are as follows: X 15 is the grate speed, X 16 is the feeding rate, X 17 is the primary air volume, X 18 is the secondary air volume, X 19 is the furnace temperature, X 20 is the furnace negative pressure, X 21 is the flue temperature, X 22 is the steam temperature, X 23 is the steam pressure, X 24 is the steam flow rate, X 25 is the ambient temperature;

[0100] X 26 ~X 35 is the process parameter, and the corresponding specific parameters are as follows: X 26 is the feed moisture content, X 27 is the moisture content of the discharge material, X 28 is the feed volatile matter, X 29 is the volatile matter of the discharge, X 30 is the feed mass flow rate, X 31 is the discharge mass flow rate, X 32 is the feed calorific value, X 33 is the calorific value of the discharge material, X 34 is the smoke oxygen concentration, X 35 is the ash content.

[0101] All of the above parameters can be directly collected through the automated control system. The data comes from historical data from the solid waste treatment plant. In order to ensure the training effect and prediction accuracy of the deep learning model, it is necessary to normalize the input and output parameters. Normalization can eliminate the influence of different dimensions on the model, so that all features are within the same scale range, which helps to speed up the model convergence and improve model performance. At the same time, normalization can also avoid the problem of numerical overflow or underflow during the numerical calculation process. For the element content indicators such as X1~X9 in the material characteristic parameters, since they are all mass percentages, they are directly divided by 100 and mapped to the range of 0-1; X 10 ~ X 12 Then a linear mapping is performed based on the maximum and minimum value range of historical data statistics; X 13 and X 14 Directly divide by 100 as a percentage. In the process operation parameters, X 15 and X 16 Mapping based on device rated range, X 17 and X 18 Based on the blower rated range map, X 19 ~ X 25Linear mapping based on historical operation data statistical range. X 26 ~ X 29 Linear mapping based on historical operation data statistical range. X 30 、X 31 Linear mapping based on design operating range. X 32 、X 33 Linear mapping based on historical data range. X 34 Linear mapping based on theoretical air excess coefficient range. X 35 Linear mapping based on historical operation data statistical range. Through such normalization processing, all parameters are uniformly mapped to the 0-1 interval, laying a foundation for subsequent training and prediction of deep learning model.

[0102] S2. Constructing a first hidden layer of a deep learning-based solid waste heat treatment prediction model;

[0103] Further, step S2 divides the nodes of the first hidden layer into a material property group, a reaction property group, and an energy property group. The material property group focuses on the physical and chemical property changes of the material during the heat treatment process, contains 30 neuron nodes, and the ith neuron node has a value of 1-30. The reaction property group focuses on describing the degree of progress of various chemical reactions, contains 25 neuron nodes, and the ith neuron node has a value of 31-55. The energy property group focuses on representing energy conversion and transmission characteristics, contains 25 neuron nodes, and the ith neuron node has a value of 56-80.

[0104] The specific implementation method of step S2 includes the following steps:

[0105] S2.1. Introducing temperature correction and material moisture correction in the material property group. The temperature correction considers the driving effect of temperature on the conversion rate of the material and reflects the differences in material conversion rules in different temperature intervals. The material moisture correction considers the influence of the evaporation process on subsequent conversion. The calculation method of the 30 neuron nodes in the material property group is as follows:

[0106]

[0107] wherein, represents the output value of the ith neuron node in the first hidden layer, and il takes a value of 1-30; is an activation function used to introduce nonlinear features; is the weight coefficient of the ith neuron node in the first hidden layer corresponding to the jth input parameter; is the input parameter, ; is the bias term of the ith neuron node in the first hidden layer corresponding to the jth input parameter; is a temperature correction coefficient, determined by expert experience, design documents, experimental research, etc. is a moisture correction coefficient, determined by expert experience, design documents, experimental research, etc. is a temperature correction term; is a material moisture correction term; while the traditional deep learning method only extracts features through weight and bias terms, lacking mechanism expression of material conversion process. This simple mathematical mapping is difficult to accurately describe the dynamic change law of material characteristics with temperature, and cannot reflect the material conversion characteristics in the pretreatment stage.

[0108] S2.2. Constructing the reaction characteristic group introduces the mass conversion term, directly reflecting the reaction degree of the substance and embodying the depth of the reaction; adding the air excess term to represent the sufficiency of the combustion environment and reflect the influence of oxidation reaction conditions on the conversion effect; the calculation method of the 25 neuron nodes in the reaction characteristic group is as follows:

[0109]

[0110] wherein, is the output value of the i2th neuron node in the first hidden layer, i2 takes a value of 31-55; is the weight coefficient of the i2th neuron node in the first hidden layer corresponding to the jth input parameter; is the bias term of the i2th neuron node in the first hidden layer corresponding to the jth input parameter; is a mass conversion coefficient, determined by expert experience, design documents, experimental research, etc. is an air excess coefficient, determined by expert experience, design documents, experimental research, etc. is air density, determined by actual measurement; is a mass conversion term; is an air excess term; while the traditional neural network model is too simple to describe the reaction process, only relying on data and lacking consideration of chemical reaction mechanism. The model cannot accurately reflect the influence mechanism of reaction conditions on the conversion process, resulting in lack of rationality of the prediction result. This improvement in the embodiment can enhance the perception ability of the model to reaction kinetics and environmental factors.

[0111] S2.3. Constructing the energy characteristic group introduces the heat value conversion term to directly represent the energy conversion effect; adding the temperature gradient term to reflect the driving force and transmission strength of heat transfer and describe the spatial distribution rule of energy; the calculation method of the 25 neuron nodes in the energy characteristic group is as follows:

[0112]

[0113] wherein, is the output value of the i3th neuron node in the first hidden layer, i3 takes values from 56 to 80; is the weight coefficient of the jth input parameter and the i3th neuron node in the corresponding first hidden layer; is the bias term of the i3th neuron node in the first hidden layer corresponding to the jth input parameter; is the heat value conversion term; is the heat value conversion coefficient, determined by expert experience, design documents, experimental research, etc. is the temperature gradient term; is the temperature gradient coefficient, which can be determined by expert experience, design documents, experimental research, etc. The conventional deep learning model often ignores the basic physical laws of energy transfer and conversion when dealing with energy conversion problems, leading to prediction results that may violate the law of conservation of energy. This improvement enables the model to consider both energy conversion and transfer characteristics, improving the prediction accuracy of energy utilization processes.

[0114] Step S3 divides the nodes of the second hidden layer into a drying process group, a pyrolysis process group, and a combustion process group. The drying process group describes the influence of the drying process on subsequent pyrolysis and combustion, contains 15 neuron nodes, and the i4th neuron node takes values from 1 to 15. The pyrolysis process group describes the composition change characteristics of pyrolysis products, contains 20 neuron nodes, and the i5th neuron node takes values from 16 to 35. The combustion process group describes the process characteristics of the final oxidation stage, contains 15 neuron nodes, and the i6th neuron node takes values from 36 to 50.

[0115] S3. Constructing the second hidden layer of the solid waste thermal treatment prediction model based on deep learning; this design is based on the basic reaction mechanism of solid waste thermal treatment and reflects the sequence and mutual influence of material conversion.

[0116] The specific implementation method of step S3 includes the following steps:

[0117] S3.1. Constructing a drying process group by characterizing the process characteristics of water removal in solid waste, reflecting the driving effect of temperature on water evaporation, and describing the influence of the drying process on subsequent pyrolysis and combustion. The calculation method of the 15 neuron nodes in the drying process group is as follows:

[0118]

[0119] wherein, is the output value of the i4th neuron node in the second hidden layer, i4 takes values from 1 to 15; is the weight coefficient of the i4th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer, =1~80; is the bias term of the i4th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer; is the output value of the kth neuron in the first hidden layer; is the drying process term, is the drying process coefficient, determined by expert experience, design documents, experimental research, etc.

[0120] S3.2. The pyrolysis process group is constructed to reflect the influence of temperature on the volatilization, describe the composition change characteristics of the pyrolysis products, and realize the extraction of the key characteristics of the pyrolysis stage. The calculation method of the 20 neuron nodes in the pyrolysis process group is as follows:

[0121]

[0122] wherein, is the output value of the i5th neuron node in the second hidden layer, i5 takes a value of 16-35; is the weight coefficient of the i5th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer; is the bias term of the i5th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer; is the pyrolysis process term, is the pyrolysis process coefficient, determined by expert experience, design documents, experimental research, etc.

[0123] Further, the extraction of the key characteristics of the pyrolysis stage can be realized.

[0124] S3.3. The combustion process group is constructed to characterize the oxidation process of fixed carbon and volatile matter, reflect the influence of oxygen concentration on the combustion effect, and describe the process characteristics of the final oxidation stage. The calculation method of the 15 neuron nodes in the combustion process group is as follows:

[0125]

[0126] wherein, is the output value of the i6th neuron node in the second hidden layer, i6 takes a value of 36-50; is the weight coefficient of the i6th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer; is the bias term of the i6th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer; is the combustion process term, is the combustion process coefficient, is the reference oxygen concentration, determined by expert experience, design documents, experimental research, etc.

[0127] S4. The output layer of the deep learning-based solid waste thermal treatment prediction model is constructed, and the output layer is divided into a conversion efficiency group, a product distribution group, and an energy conversion group, and the output parameters of the output layer of the deep learning-based solid waste thermal treatment prediction model are determined;

[0128] Further, step S4 divides the output layer into a conversion efficiency group, a product distribution group, and an energy conversion group based on the evaluation requirements of the solid waste thermal treatment process; the conversion efficiency group index reflects the integrity and sufficiency of material conversion, the product distribution group index reveals the migration rule of material and environmental impact, and the energy conversion group index embodies the energy utilization level;

[0129] The conversion efficiency group index includes Y1-Y6, Y1 is the moisture conversion rate, Y2 is the volatile conversion rate, Y3 is the fixed carbon conversion rate, Y4 is the carbon conversion rate, Y5 is the sulfur conversion rate, and Y6 is the chlorine conversion rate;

[0130] The product distribution group index includes the material distribution proportion index Y 71 ~ Y 75 and the residue distribution index Y8-Y9, Y 71 is the mass distribution proportion of moisture after drying into flue gas, Y 72 is the mass distribution proportion of volatile after pyrolysis into flue gas, Y 73 is the distribution proportion of carbon converted into carbon dioxide into flue gas, Y 74 is the distribution proportion of sulfur converted into sulfur dioxide into flue gas, Y 75 is the distribution proportion of chlorine converted into hydrogen chloride into flue gas; Y8 is the mass distribution proportion of fixed carbon remaining in the slag after combustion, and Y9 is the mass distribution proportion of fly ash component into fly ash after combustion of fixed carbon;

[0131] The energy conversion group index includes Y 10 ~Y 12 , Y 10 is the thermal efficiency, Y 11 is the combustion efficiency, and Y 12 is the heat loss rate;

[0132] For the output parameters, a normalization method is first used to map and convert the data. The conversion rate index (Y1-Y6) itself is within the range of 0-1; the material distribution proportion index (Y 71 -Y 75 ) and the residue distribution index (Y8-Y9) are naturally limited within the range of 0-1 as percentage values; and the efficiency index (Y 10 -Y 12)Based on the actual operation data of the system to determine the reasonable range of normalization mapping. Through such normalization processing, all parameters are uniformly mapped to the same numerical interval, laying a foundation for subsequent deep learning model training and prediction.

[0133] The specific implementation method of step S4 includes the following steps:

[0134] S4.1. The calculation method of the output value of different indicators in the conversion efficiency group is as follows:

[0135]

[0136] Wherein, is the output value of different indicators in the conversion efficiency group, s1 is the different indicators in the conversion efficiency group; is the weight coefficient of the output value of different indicators in the conversion efficiency group corresponding to the vth neuron of the second hidden layer; is the bias term of the output value of different indicators in the conversion efficiency group corresponding to the vth neuron of the second hidden layer; is the output value of the vth neuron of the second hidden layer;

[0137] s1=1, v∈{1,2,…,15}; s1=2, v∈{16,17,…,35}; s1=3~6, v∈{36,37,…,50};

[0138] Further, the conversion efficiency group respectively reflects the conversion characteristics of the most key components in solid waste in the heat treatment process: the moisture conversion rate represents the drying effect, the volatile matter and fixed carbon conversion rates reflect the pyrolysis and combustion degree, and the carbon, sulfur and chlorine conversion rates reflect the conversion law of potential pollutants, which can comprehensively reflect the material conversion effect. The moisture conversion rate Y1 is associated with the drying process group because water evaporation mainly occurs in the drying stage; the volatile matter conversion rate Y2 is associated with the pyrolysis process group because volatile matter release occurs in the pyrolysis stage; and the fixed carbon, carbon, sulfur and nitrogen conversion rates (Y3-Y6) are associated with the combustion process group because the conversion of these elements is mainly completed in the combustion stage.

[0139] S4.2. The calculation method of the output value of different indicators in the product distribution group is as follows:

[0140]

[0141] Wherein, is the output value of different indicators in the product distribution group, s2 is the different indicators in the product distribution group; is the weight coefficient of the output value of different indicators in the product distribution group corresponding to the vth neuron of the second hidden layer; is the bias term of the output value of different indicators in the product distribution group corresponding to the vth neuron of the second hidden layer;

[0142] s2=71, v e {1, 2, …, 15}; s2=72, v e {16, 17, …, 35}; s2=73, 74, 75, 8, 9, v e {36, 37, …, 50}.

[0143] The product distribution group selects three component distribution indexes of flue gas, slag and fly ash. The flue gas component distribution includes: the mass distribution proportion Y71 of moisture after drying into flue gas, the mass distribution proportion Y72 of volatile matter after pyrolysis into flue gas, the distribution proportion Y73 of carbon element converted into carbon dioxide into flue gas, the distribution proportion Y74 of sulfur element converted into sulfur dioxide into flue gas, and the distribution proportion Y75 of chlorine element converted into hydrogen chloride into flue gas. The slag component distribution is the mass distribution proportion Y8 of fixed carbon remaining in the slag after combustion. The fly ash component distribution is the mass distribution proportion Y9 of fixed carbon into fly ash after combustion. These indexes not only meet the material balance requirement, but also reflect the influence of process parameters on product distribution, and are closely related to environmental impact control. Through the three indexes, the migration law of matter in different phases can be completely characterized. The moisture proportion Y71 in flue gas comes from the drying stage, so it is associated with the drying process group. The volatile matter proportion Y72 into flue gas is derived from the pyrolysis stage, so it is associated with the pyrolysis process group. The flue gas proportions (Y73-Y75) of carbon, sulfur and nitrogen elements converted and the slag and fly ash proportions (Y8-Y9) are all generated in the combustion stage, so they are associated with the combustion process group.

[0144] S4.3. The calculation method of the output values of different indexes in the energy conversion group is as follows:

[0145]

[0146] wherein, is the output value of different indexes in the energy conversion group, and s3 is different indexes in the energy conversion group. is the weight coefficient of the output value of different indexes in the energy conversion group corresponding to the vth neuron of the second hidden layer; is the bias term of the output value of different indexes in the energy conversion group corresponding to the vth neuron of the second hidden layer;

[0147] s3=10, 12, v e {1, 2, …, 50}; s3=11, v e {36, 37, …, 50}.

[0148] The energy conversion group comprises three indexes of thermal efficiency, combustion efficiency and heat loss rate, which are sequentially recorded as Y10-Y12, i.e., Y10 thermal efficiency, Y11 combustion efficiency and Y12 heat loss rate. They represent the energy conversion characteristics from different angles: the thermal efficiency reflects the overall energy utilization level, the combustion efficiency embodies the chemical energy conversion integrity, and the heat loss rate represents the process energy loss. The three indexes are directly related to the input heat value parameter and temperature parameter, and can comprehensively evaluate the energy utilization effect of the heat treatment process. The thermal efficiency Y10 and the heat loss rate Y12 need to consider the energy conversion of the whole heat treatment process, and therefore are associated with all process groups. The combustion efficiency Y11 only reflects the energy conversion effect of the combustion stage, and therefore is only associated with the combustion process group.

[0149] S5. Designing a loss function of the solid waste heat treatment prediction model based on deep learning;

[0150] Further, the prediction loss and the heat treatment loss are considered in the loss function of step S5, and the specific implementation method comprises the following steps:

[0151] S5.1. Constructing the prediction loss based on the conversion efficiency prediction loss, the product distribution prediction loss and the energy conversion prediction loss, and the expression is:

[0152]

[0153] wherein, is the conversion efficiency prediction loss; is the product distribution prediction loss; is the energy conversion prediction loss;

[0154]

[0155] wherein, is a mean square error calculation function; is the output value of different indexes in the conversion efficiency group predicted by the model; is the output value of different indexes in the conversion efficiency group actually measured;

[0156]

[0157] wherein, is the output value of different indexes in the product distribution group predicted by the model; is the output value of different indexes in the product distribution group actually measured;

[0158]

[0159] wherein, is the output value of different indexes in the energy conversion group predicted by the model; different index output values in the energy conversion group for actual measurement;

[0160] S5.2. Constructing heat treatment loss based on material conversion constraint, component conversion constraint and heat conversion constraint , the expression is:

[0161]

[0162] wherein, is the material conversion constraint, is the weight corresponding to the material conversion constraint; is the component conversion constraint, is the weight corresponding to the component conversion constraint; is the heat conversion constraint, is the weight corresponding to the heat conversion constraint;

[0163] The material conversion constraint is established as:

[0164]

[0165] wherein, is the volatile conversion term; is the fixed carbon distribution term;

[0166] Further, in the material conversion constraint, two key material conversion balance relationships are mainly considered: one is the conversion balance of volatile, that is, the material balance of volatile in the original material which is converted into gaseous products by pyrolysis and then enters the flue gas, which is represented by comparing the difference between the conversion amount of volatile in the feed and the amount of volatile finally entering the flue gas; the other is the distribution balance of fixed carbon, that is, the distribution balance of fixed carbon in the original material in the slag and fly ash after combustion, which is represented by comparing the difference between the content of fixed carbon in the feed and the amount of fixed carbon finally distributed in the slag and fly ash.

[0167] The component conversion constraint is established as:

[0168]

[0169] wherein, is the sulfur element conversion constraint, is the chlorine element conversion constraint;

[0170] The heat conversion constraint is established as:

[0171] ;

[0172] In the component conversion constraint, the conversion rules of two key elements in the high-temperature heat treatment process are mainly considered: the material balance of sulfur element converted into SO2 in the flue gas at high temperature, which is characterized by comparing the difference between the conversion amount of sulfur element in the feed and the amount of sulfur element in the flue gas after SO2 is finally formed; the material balance of chlorine element converted into HCl in the pyrolysis and combustion process, which is characterized by comparing the difference between the conversion amount of chlorine element in the feed and the amount of chlorine element in the flue gas after HCl is finally formed.

[0173] S5.3. Constructing total loss based on predicted loss and heat treatment loss ;

[0174]

[0175] wherein, is the weight corresponding to the predicted loss, is the weight corresponding to the heat treatment loss, which is determined by expert experience.

[0176] S6. Connecting the input layer, the first hidden layer, the second hidden layer and the output layer of the deep learning-based solid waste heat treatment prediction model in sequence to obtain the deep learning-based solid waste heat treatment prediction model;

[0177] S7. Collecting data through the solid waste treatment plant and the automatic control system, normalizing the data, dividing the data into a training set, a validation set and a test set according to a ratio of 7:2:1, training, validating and testing the deep learning-based solid waste heat treatment prediction model obtained in step S6 to obtain the trained deep learning-based solid waste heat treatment prediction model;

[0178] Further, the hyperparameters set by the human based on expert experience include: network structure parameters, the number of neurons in each layer; hyperparameters set by the human during the model training process include: learning rate, batch size, training rounds, and weight coefficients in the loss function; parameters determined by the model during the training process: weight coefficients and bias terms in the model.

[0179] Further, the model training in step S7 adopts the back propagation algorithm; during the training process, the model calculates the predicted value through forward propagation, substitutes the predicted value and the true value into the loss function to calculate the loss, and then updates the network parameters through back propagation; to prevent overfitting, the early stopping strategy is adopted, and the training is stopped when the validation set loss does not decrease for 10 consecutive epochs; the output result is restored to the actual physical quantity through inverse normalization.

[0180] S8. After denormalizing the prediction results of the trained deep learning-based solid waste heat treatment prediction model, the prediction results of the deep learning-based solid waste heat treatment prediction model with actual physical quantities are obtained. A solid waste heat treatment anomaly prediction method is constructed. When considering the solid waste heat treatment anomaly problem, it is analyzed from the perspectives of material transformation anomaly, material distribution anomaly, and energy conversion anomaly, and an anomaly diagnosis and judgment scheme is proposed.

[0181] During the thermal treatment of solid waste, due to the complex composition of solid waste and large fluctuations in calorific value, the thermal treatment process includes multiple stages such as drying, pyrolysis, and combustion, and there is a strong coupling relationship between each stage. Therefore, in order to ensure the normal operation of solid waste thermal treatment, higher requirements are needed for operation management. It is necessary to promptly discover abnormal problems in the solid waste thermal treatment process and guide optimization and adjustment decisions through prediction.

[0182] Furthermore, the specific implementation method of step S8 includes the following steps:

[0183] S8.1. During the identification of abnormalities in solid waste thermal treatment, the material conversion abnormality reflects the degree of independent deviation of each process parameter, which helps to accurately locate the abnormal link and take targeted adjustments. The calculation formula is:

[0184]

[0185] in, Output values ​​for different indicators predicted by the model, include 、 、 ; For The corresponding standard values ​​are determined by expert experience, design documents, and experimental research methods; The allowable fluctuation range corresponding to the material conversion abnormality formula is determined by expert experience, design documents, experimental research, etc. for and The corresponding allocation weight coefficient;

[0186] S8.2. Material distribution abnormality includes flue gas distribution abnormality and solid distribution abnormality;

[0187] Considering the distribution deviation of each component in the flue gas during the process of water vaporization, volatile matter analysis, carbon oxidation, sulfur oxidation and hydrogen chloride generation, the flue gas distribution abnormality is calculated. , the expression is:

[0188]

[0189] wherein, s2=71, 72, 73, 74, 75 are the distribution weight coefficients, and wherein, the corresponding standard value is determined by expert experience, design documents, experimental research, etc. s4=71, 72, 73, 74, 75 are the corresponding allowable fluctuation ranges of the distribution weight coefficients of the slag and fly ash in the formula of the solid distribution abnormality degree;

[0190] Considering the distribution balance state of inorganic matters in the slag and fly ash, and combining the influences of the combustion temperature, residence time and gas-solid separation efficiency and other process parameters on the slag-fly ash ratio, the solid distribution abnormality degree is calculated, and the expression is:

[0191]

[0192] wherein, , s1=71, 72, 73, 74, 75 are the distribution weight coefficients of the slag and fly ash, respectively; , s4=71, 72, 73, 74, 75 are the corresponding allowable fluctuation ranges of the distribution weight coefficients of the slag and fly ash in the formula of the solid distribution abnormality degree;

[0193] S8.3. The energy conversion abnormality degree includes the energy abnormality degree and the energy balance abnormality degree ;

[0194] Considering the influences of the solid waste heat value fluctuation, combustion air distribution, furnace temperature distribution and heat exchange efficiency and other factors on the system thermal efficiency, combustion efficiency and heat loss, the energy abnormality degree is calculated, and the expression is:

[0195]

[0196] wherein, s3=10, 11, 12 are the energy index weights; s5=10, 11, 12 are the corresponding allowable fluctuation ranges in the formula of the energy abnormality degree;

[0197] Considering the balance relationship between the effective heat output and the heat loss of the system, and combining the influences of the furnace wall insulation performance, cooling system efficiency and other factors, the energy balance abnormality degree is calculated, and the expression is:

[0198]

[0199] wherein, s6 is the energy balance allowable error, which is determined by expert experience, design documents, experimental research, etc.

[0200] S8.4. The comprehensive abnormality index ZH is constructed;

[0201]

[0202] wherein, 、 、 、 、 respectively, are the corresponding process weight coefficients, which are determined by expert experience, design documents, experimental research, etc. 、 、 、 、

[0203] S8.5. Abnormal diagnosis is determined based on the comprehensive abnormal index obtained in step S4.

[0204] The abnormal level is divided into four levels, namely normal operation, slight abnormality, moderate abnormality and serious abnormality, wherein the critical comprehensive abnormal indexes corresponding to the slight abnormality, moderate abnormality and serious abnormality are respectively 、 、 ; then is normal operation; is slight abnormality; is moderate abnormality; is serious abnormality.

[0205] The embodiment proposes a solid waste thermal treatment process characteristic analysis deep learning model construction method. The method is designed and optimized for solid waste thermal treatment characteristics in input parameters, first hidden layer, second hidden layer and output layer, realizing the scientificity and rationality of the prediction of conversion efficiency, product distribution and energy conversion in the solid waste thermal treatment process.

[0206] The embodiment proposes a solid waste thermal treatment abnormality prediction method. The established deep learning model is used to determine the state of solid waste thermal treatment from the aspects of material conversion abnormality, material distribution abnormality and energy conversion abnormality, which can comprehensively evaluate the performance of the system, discover and handle potential operation problems in time, and ensure the continuity and safety of the solid waste treatment process.

[0207] ​It has to be noted that the terms "first", "second", and the like in connection with an entity or action refer to this entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without further constraints, exclude the presence of additional elements of the process, method, article, or apparatus.

[0208] While the application has been described with reference to specific implementations thereof, it should be understood that various modifications and substitutions can be made by those skilled in the art without departing from the scope of the present application. In particular, any one of the features of the present application disclosed above can be utilized independently of any other and the scope of the application should not be limited by the specific embodiments disclosed herein, but should be given the widest coverage possible in its true scope.

Claims

1. A method for predicting and predicting abnormalities in solid waste thermal treatment based on deep learning, characterized in that: The steps include: S1. Determine the input parameters of the input layer of the deep learning-based solid waste thermal treatment prediction model, including material characteristic parameters, process operation parameters, and process parameters; S2. Construct the first hidden layer of a deep learning-based solid waste thermal treatment prediction model. S3. Construct the second hidden layer of the deep learning-based solid waste thermal treatment prediction model. S4. Construct an output layer for the deep learning-based solid waste thermal treatment prediction model, divide the output layer into conversion efficiency groups, product distribution groups, and energy conversion groups, and determine output parameters for the output layer of the deep learning-based solid waste thermal treatment prediction model. S5. Design a loss function for a deep learning-based solid waste thermal treatment prediction model. S6. Sequentially connecting the input layer, the first hidden layer, the second hidden layer, and the output layer of the deep learning-based solid waste thermal treatment prediction model to obtain a deep learning-based solid waste thermal treatment prediction model; S7. Data is collected from the solid waste treatment plant and automated control system, normalized, and divided into training, validation, and test sets in a ratio of 7:2:

1. The deep learning-based solid waste thermal treatment prediction model obtained in step S6 is trained, validated, and tested to obtain a trained deep learning-based solid waste thermal treatment prediction model. S8. After denormalizing the prediction results of the trained deep learning-based solid waste heat treatment prediction model, the prediction results of the deep learning-based solid waste heat treatment prediction model with actual physical quantities are obtained. A solid waste heat treatment anomaly prediction method is constructed. When considering the solid waste heat treatment anomaly problem, it is analyzed from the perspectives of material transformation anomaly, material distribution anomaly, and energy conversion anomaly, and an anomaly diagnosis and judgment scheme is proposed.

2. The method for predicting and abnormality prediction of solid waste heat treatment based on deep learning according to claim 1 is characterized in that: In step S1, the number of nodes in the input layer is determined to be 35, which are denoted as X1-X 35 ; X1~X 14 The material characteristic parameters are as follows: X1 is the moisture content, X2 is the ash content, X3 is the combustible content, X4 is the carbon content, X5 is the hydrogen content, X6 is the oxygen content, X7 is the nitrogen content, X8 is the sulfur content, X9 is the chlorine content, X 10 is the calorific value, X 11 is the bulk density, X 12 is the particle size, X 13 is the sand content, X 14 is the metal content; X 15 ~X 25 is the process operating parameter, and the corresponding specific parameters are as follows: X 15 is the grate speed, X 16 is the feeding rate, X 17 is the primary air volume, X 18 is the secondary air volume, X 19 is the furnace temperature, X 20 is the furnace negative pressure, X 21 is the flue temperature, X 22 is the steam temperature, X 23 is the steam pressure, X 24 is the steam flow rate, X 25 is the ambient temperature; X 26 ~X 35 is the process parameter, and the corresponding specific parameters are as follows: X 26 is the feed moisture content, X 27 is the moisture content of the discharge material, X 28 is the feed volatile matter, X 29 is the volatile matter of the discharge, X 30 is the feed mass flow rate, X 31 is the discharge mass flow rate, X 32 is the feed calorific value, X 33 is the calorific value of the discharge material, X 34 is the smoke oxygen concentration, X 35 is the ash content.

3. The method for predicting and abnormality prediction of solid waste heat treatment based on deep learning according to claim 2 is characterized in that: Step S2 divides the nodes of the first hidden layer into a material property group, a reaction property group, and an energy property group. The material property group focuses on the changes in the physical and chemical properties of the material during the heat treatment process, and includes 30 neuron nodes, with the value of the i1th neuron node ranging from 1 to 30. The reaction property group focuses on describing the degree of various chemical reactions, and includes 25 neuron nodes, with the value of the i2th neuron node ranging from 31 to 55. The energy property group focuses on characterizing energy conversion and transfer characteristics, and includes 25 neuron nodes, with the value of the i3th neuron node ranging from 56 to 80. The specific implementation method of step S2 includes the following steps: S2.

1. Introduce temperature correction and material moisture correction into the material property group. The temperature correction considers the driving effect of temperature on the material conversion rate and reflects the differences in material conversion patterns within different temperature ranges. The material moisture correction considers the impact of water evaporation on subsequent conversion. The calculation method for the 30 neuron nodes in the material property group is as follows: in, Represents the output value of the i1th neuron node in the first hidden layer, where i1 ranges from 1 to 30; is an activation function used to introduce nonlinear features; is the weight coefficient of the i1th neuron node in the first hidden layer corresponding to the jth input parameter; is the input parameter, ; is the bias term of the i1th neuron node in the first hidden layer corresponding to the jth input parameter; is the temperature correction factor, which is determined by expert experience, design documents, experimental research, etc. is the moisture correction factor, which is determined by expert experience, design documents, experimental research, etc. is the temperature correction term; is the material moisture correction item; S2.

2. The reaction characteristic group is constructed by introducing a mass conversion term to directly reflect the degree of material reaction and the depth of the reaction. An excess air term is added to characterize the adequacy of the combustion environment and reflect the impact of oxidation reaction conditions on the conversion effect. The 25 neuron nodes in the reaction characteristic group are calculated as follows: in, is the output value of the i2th neuron node in the first hidden layer, and i2 ranges from 31 to 55; is the weight coefficient of the i2th neuron node in the first hidden layer corresponding to the jth input parameter; is the bias term of the i2th neuron node in the first hidden layer corresponding to the jth input parameter; is the mass conversion coefficient, which is determined by expert experience, design documents, experimental research, etc. is the excess air coefficient, which is determined by expert experience, design documents, experimental research, etc. is the air density, determined by actual measurement; is the quality conversion item; is the excess air term; S2.

3. Construct the energy characteristic group by introducing a calorific value conversion term to directly characterize the energy conversion effect. A temperature gradient term is added to reflect the driving force and transfer intensity of heat transfer and describe the spatial distribution of energy. The 25 neuron nodes in the energy characteristic group are calculated as follows: in, is the output value of the i3th neuron node in the first hidden layer, and the value of i3 ranges from 56 to 80; is the weight coefficient of the jth input parameter and the corresponding i3th neuron node in the first hidden layer; is the bias term of the i3th neuron node in the first hidden layer corresponding to the jth input parameter; is the calorific value conversion item; is the calorific value conversion coefficient, which is determined by expert experience, design documents, experimental research, etc. is the temperature gradient term; is the temperature gradient coefficient, which can be determined by expert experience, design documents, experimental research, etc.

4. The method for predicting and abnormality prediction of solid waste heat treatment based on deep learning according to claim 3 is characterized in that: Step S3 divides the nodes of the second hidden layer into a drying process group, a pyrolysis process group, and a combustion process group. The drying process group describes the effect of the drying process on subsequent pyrolysis and combustion, and includes 15 neuron nodes, with the i4th neuron node taking a value of 1 to 15. The pyrolysis process group describes the composition change characteristics of the pyrolysis products, and includes 20 neuron nodes, with the i5th neuron node taking a value of 16 to 35. The combustion process group describes the process characteristics of the final oxidation stage, and includes 15 neuron nodes, with the i6th neuron node taking a value of 36 to 50. The specific implementation method of step S3 includes the following steps: S3.

1. Construct a drying process group to characterize the process of water removal from solid waste, reflect the driving effect of temperature on water evaporation, and describe the impact of the drying process on subsequent pyrolysis and combustion. The 15 neuron nodes in the drying process group are calculated as follows: in, is the output value of the i4th neuron node in the second hidden layer, and i4 ranges from 1 to 15; is the weight coefficient of the i4th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer, =1~80; is the bias term of the i4th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer; is the output value of the kth neuron in the first hidden layer; is the drying process item, is the drying process coefficient, which is determined by expert experience, design documents, experimental research, etc. S3.

2. Construct a pyrolysis process group. This group uses the kinetic characteristics of organic matter pyrolysis to reflect the influence of temperature on volatile analysis, describe the compositional changes of pyrolysis products, and extract key features of the pyrolysis stage. The 20 neuron nodes in the pyrolysis process group are calculated as follows: in, is the output value of the i5th neuron node in the second hidden layer, and i5 ranges from 16 to 35; is the weight coefficient of the i5th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer; is the bias term of the i5th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer; is the pyrolysis process term, is the pyrolysis process coefficient, which is determined by expert experience, design documents, experimental research, etc. S3.

3. Construct a combustion process group to characterize the oxidation of fixed carbon and volatile matter, reflect the effect of oxygen concentration on combustion, and describe the process characteristics of the final oxidation stage. The calculation method for the 15 neuron nodes in the combustion process group is as follows: in, is the output value of the i6th neuron node in the second hidden layer, and the value of i6 ranges from 36 to 50; is the weight coefficient of the i6th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer; is the bias term of the i6th neuron node in the second hidden layer corresponding to the kth neuron in the first hidden layer; is the combustion process term, is the combustion process coefficient, It is the reference oxygen concentration, determined by expert experience, design documents, and experimental research methods.

5. The method for predicting and abnormality prediction of solid waste heat treatment based on deep learning according to claim 4 is characterized in that: Step S4 divides the output layer into conversion efficiency group, product distribution group, and energy conversion group based on the evaluation requirements of the solid waste thermal treatment process; the conversion efficiency group indicators reflect the completeness and adequacy of material conversion, the product distribution group indicators reveal the migration patterns and environmental impacts of materials, and the energy conversion group indicators reflect the level of energy utilization; The conversion efficiency group indicators include Y1~Y6, where Y1 is water conversion rate, Y2 is volatile matter conversion rate, Y3 is fixed carbon conversion rate, Y4 is carbon conversion rate, Y5 is sulfur conversion rate, and Y6 is chlorine conversion rate; Product distribution group indicators include material distribution ratio indicator Y 71 ~ Y 75 and residue distribution index Y8~ Y9, Y 71 is the mass distribution ratio of water entering the flue gas after drying, Y 72 Y is the mass distribution ratio of volatile matter entering the flue gas after pyrolysis, 73 Y is the distribution ratio of carbon into flue gas after being converted into carbon dioxide, 74 Y is the distribution ratio of sulfur into flue gas after being converted into sulfur dioxide, 75 Y is the distribution ratio of chlorine into the flue gas after conversion to hydrogen chloride; Y8 is the mass distribution ratio of the fixed carbon remaining in the slag after combustion; Y9 is the mass distribution ratio of the fly ash components into the fly ash after combustion of fixed carbon; Energy conversion group indicators include Y 10 ~Y 12 , Y 10 is the thermal efficiency, Y 11 is the combustion efficiency, Y 12 is the heat loss rate; The specific implementation method of step S4 includes the following steps: S4.

1. The calculation method for constructing the output values ​​of different indicators in the conversion efficiency group is as follows: in, are the output values ​​of different indicators in the conversion efficiency group, and s1 are the different indicators in the conversion efficiency group; is the weight coefficient of the output values ​​of different indicators in the conversion efficiency group corresponding to the vth neuron in the second hidden layer; is the bias term for the output values ​​of different indicators in the conversion efficiency group corresponding to the vth neuron in the second hidden layer; is the output value of the vth neuron in the second hidden layer; When s1=1, v∈{1,2,…,15}; when s1=2, v∈{16,17,…,35}; when s1=3~6, v∈{36,37,…,50}; S4.

2. The calculation method for the output values ​​of different indicators in the product distribution group is as follows: in, are the output values ​​of different indicators in the product distribution group, and s2 are the different indicators in the product distribution group; is the weight coefficient of the output values ​​of different indicators in the product distribution group corresponding to the vth neuron in the second hidden layer; is the bias term for the output values ​​of different indicators in the product distribution group corresponding to the vth neuron in the second hidden layer; When s2=71, v∈{1,2,…,15}; when s2=72, v∈{16,17,…,35}; when s2=73,74,75,8,9, v∈{36,37,…,50}; S4.

3. The calculation method for constructing the output values ​​of different indicators in the energy conversion group is as follows: in, are the output values ​​of different indicators in the energy conversion group, and s3 are the different indicators in the energy conversion group; is the weight coefficient of the output value of different indicators in the energy conversion group corresponding to the vth neuron in the second hidden layer; is the bias term for the output values ​​of different indicators in the energy conversion group corresponding to the vth neuron in the second hidden layer; When s3=10,12, v∈{1,2,…,50}; when s3=11, v∈{36,37,…,50}.

6. The method for predicting and abnormality prediction of solid waste heat treatment based on deep learning according to claim 5 is characterized in that: The loss function of step S5 takes into account the prediction loss and the heat treatment loss. The specific implementation method includes the following steps: S5.

1. Construct the predicted loss based on the conversion efficiency predicted loss, product distribution predicted loss, and energy conversion predicted loss. The expression is: in, Predict losses for conversion efficiency; Predict losses for product distribution; Energy conversion prediction losses; in, is the mean square error calculation function; Output values ​​for different indicators in the conversion efficiency group predicted by the model; Output values ​​for different indicators in the actual measured conversion efficiency group; in, Output values ​​for different indicators in the product distribution group predicted by the model; Output values ​​for different indicators in the actual measured product distribution group; in, Output values ​​of different indicators in the energy conversion group predicted by the model; Output values ​​of different indicators in the energy conversion group actually measured; S5.

2. Constructing heat treatment losses based on material conversion constraints, component conversion constraints, and heat conversion constraints , the expression is: in, is the material conversion constraint, The weight corresponding to the material conversion constraint; is the component transformation constraint, is the weight corresponding to the component transformation constraint; is the heat conversion constraint, is the weight corresponding to the heat conversion constraint; Material conversion constraints are established as: in, is the volatile matter conversion item; Allocate items for fixed carbon; The component conversion constraints are established as: in, For the sulfur element transformation constraints, The conversion constraint for chlorine element; The heat conversion constraint is established as: ； S5.

3. Constructing Total Losses Based on Predicted Losses and Heat Treatment Losses ; in, is the weight corresponding to the prediction loss, is the weight corresponding to the heat treatment loss, determined by expert experience.

7. The method for predicting and abnormality prediction of solid waste heat treatment based on deep learning according to claim 6 is characterized in that: In step S7, the model training adopts the back-propagation algorithm. During the training process, the model calculates the predicted value through forward propagation, substitutes the predicted value and the true value into the loss function to calculate the loss, and then updates the network parameters through back-propagation. To prevent overfitting, an early stopping strategy is adopted, and training is stopped when the validation set loss does not decrease for 10 consecutive epochs. The output result is restored to the actual physical quantity through denormalization.

8. The method for predicting and abnormality prediction of solid waste heat treatment based on deep learning according to claim 7 is characterized in that: The specific implementation method of step S8 includes the following steps: S8.

1. During the identification of abnormalities in solid waste thermal treatment, the material conversion abnormality reflects the degree of independent deviation of each process parameter, which helps to accurately locate the abnormal link and take targeted adjustments. The calculation formula is: in, Output values ​​for different indicators predicted by the model, include 、 、 ; For The corresponding standard values ​​are determined by expert experience, design documents, and experimental research methods; The allowable fluctuation range corresponding to the material conversion abnormality formula is determined by expert experience, design documents, experimental research, etc. for and The corresponding allocation weight coefficient; S8.

2. Material distribution abnormality includes flue gas distribution abnormality and solid distribution abnormality; Considering the distribution deviation of each component in the flue gas during the process of water vaporization, volatile matter analysis, carbon oxidation, sulfur oxidation and hydrogen chloride generation, the flue gas distribution abnormality is calculated. , the expression is: in, is the distribution weight coefficient, s2=71, 72, 73, 74, 75; For The corresponding standard values ​​are determined by expert experience, design documents, experimental research, etc.; Assign the corresponding allowable fluctuation range in the abnormality degree formula to the flue gas; Considering the distribution equilibrium state of inorganic matter in slag and fly ash, combined with the influence of process parameters such as combustion temperature, residence time and gas-solid separation efficiency on slag-to-fly ash ratio, the solid distribution abnormality is calculated. , the expression is: in, 、 Assign weight coefficients to slag and fly ash respectively; 、 are the allowable fluctuation ranges corresponding to the slag and fly ash distribution weight coefficients in the solid distribution anomaly formula respectively; S8.

3. Energy conversion anomaly includes energy anomaly and energy balance abnormality ; Considering the impact of factors such as solid waste calorific value fluctuations, combustion air distribution, furnace temperature distribution and heat exchange efficiency on system thermal efficiency, combustion efficiency and heat loss, calculate the energy anomaly , the expression is: in, is the energy index weight, s3=10, 11, 12; is the corresponding allowable fluctuation range in the energy anomaly formula; Considering the balance between the effective heat output and heat loss of the system, combined with the influence of factors such as furnace wall insulation performance and cooling system efficiency, calculate the energy balance anomaly , the expression is: in, The permissible error in energy balance is determined by expert experience, design documents, experimental studies, etc. S8.

4. Construct a comprehensive anomaly index ZH; in, 、 、 、 、 Respectively 、 、 、 、 The corresponding process weight coefficient is determined by expert experience, design documents, experimental research, etc. S8.5 abnormal diagnosis based on the comprehensive abnormality index obtained in step S4; The abnormality level is divided into 4 levels, namely normal operation, slight abnormality, moderate abnormality and severe abnormality. The critical comprehensive abnormality indexes corresponding to slight abnormality, moderate abnormality and severe abnormality are 、 、 ; then we get For normal operation; It is a slight abnormality; Moderate abnormality; A serious abnormality.

Images