Adaptive control method and system for pyrolysis temperature and oxygen concentration in rotary kiln

By arranging sensors on the rotary kiln to collect data and using an adaptive control method based on LSTM and deep Q-network models, the problem of multi-parameter coupled control in rotary kiln pyrolysis was solved, achieving more precise temperature and oxygen concentration management, and improving incineration efficiency and environmental protection.

CN118642375BActive Publication Date: 2024-11-12NANTONG INST OF TECH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411110730.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-11-12
Estimated Expiration
2044-08-14

AI Technical Summary

Technical Problem

In the existing technology, the control of rotary kiln pyrolysis temperature and oxygen concentration mainly relies on the traditional PID control method, which lacks overall optimization of complex coupled multi-parameter systems, resulting in poor control performance.

Method used

An adaptive control method is adopted, which collects waste and flue gas parameters by arranging sensors at different positions of the rotary kiln, uses an LSTM network to learn the temporal correlation of historical data, and combines a deep Q-network model to optimize fuel and air volume control. A comprehensive reward function is constructed to achieve multi-objective adaptive adjustment.

Benefits of technology

It achieves precise temperature control and oxygen concentration management in the rotary kiln pyrolysis process, improves combustion completeness, reduces harmful gas emissions, and enhances the robustness and efficiency of control.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118642375B_ABST
    Figure CN118642375B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for adaptively controlling the pyrolysis temperature and oxygen concentration of a rotary kiln, which relates to the field of automated control technology. The method and system obtain the fuel quantity control quantity and the air quantity control quantity by obtaining the waste physical property parameters, temperature parameters, oxygen concentration parameters, water vapor concentration parameters and product component parameters; train a temperature adaptive control model; calculate the secondary air volume control quantity, and predict the fuel quantity control quantity and the air quantity control quantity; calculate the temperature distribution uniformity reward, the product component qualification rate reward, the energy consumption reward, and obtain a comprehensive reward function; design the state, action and reward function of an intelligent agent, and when the mean value of the comprehensive reward function is greater than a strategy switching threshold, use the secondary air volume control quantity and the predicted fuel quantity control quantity and air quantity control quantity as control strategies, and when the mean value of the comprehensive reward function is less than or equal to the strategy switching threshold, use the intelligent agent to generate a control strategy to optimize the pyrolysis effect.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of automation control technology, and in particular to a method and system for adaptively controlling the pyrolysis temperature and oxygen concentration of a rotary kiln. Background Art

[0002] The rotary kiln plays a key role in the pyrolysis of low-rank coal. Low-rank coal pyrolysis is the process of heating low-rank coal to a certain temperature under oxygen-deficient or oxygen-free conditions to cause a series of physical changes and chemical reactions to produce products such as coke, tar and coal gas. The rotary kiln is one of the key equipment for achieving low-rank coal pyrolysis, which uses the rotation and axial tilt of the furnace body to achieve material transportation and heating.

[0003] At present, the control of rotary kiln pyrolysis temperature and oxygen concentration mainly adopts the traditional PID control method, which adjusts the fuel amount or air amount according to the temperature at the kiln head or kiln tail to achieve temperature feedback control.

[0004] A Chinese patent with publication number CN112947094A discloses a method for self-adjusting PID parameters of temperature control of a cement rotary kiln, the method comprising: (1) obtaining and setting initial parameters of an incremental PID control module by reinforcement learning; (2) collecting the ideal temperature of the cement rotary kiln, the deviation between the actual temperature and the ideal temperature, and the actual temperature in real time as input values ​​of a BP neural network module; (3) inputting the data collected in real time in step (2) into the BP neural network module, performing online network operation, the network outputting PID parameters and then obtaining control parameters, the cement rotary kiln model module using the control parameters to control the opening of the coal feeding valve of the cement rotary kiln in real time, and then controlling the internal working temperature of the cement rotary kiln; repeating steps (2) and (3), and continuously updating the network weights and thresholds of the BP neural network module until the internal temperature of the cement rotary kiln reaches the set conditions.

[0005] Rotary kiln pyrolysis is a multi-parameter complex coupling system. Temperature, oxygen concentration, negative pressure, material ratio, residence time and many other factors interact with each other. Traditional PID control is mostly aimed at a single parameter and lacks overall optimization. It is urgent to establish a multi-parameter coupled pyrolysis process mechanism model to reveal the quantitative correlation of key parameters. Summary of the invention

[0006] The present application aims to solve at least one of the technical problems in the related art to a certain extent. To this end, one purpose of the present application is to propose a method and system for adaptively controlling the pyrolysis temperature and oxygen concentration of a rotary kiln, thereby optimizing the pyrolysis effect.

[0007] One aspect of the present application provides an adaptive control method for pyrolysis temperature and oxygen concentration of a rotary kiln, comprising:

[0008] Step S100: sending waste into the feed end of the rotary kiln, obtaining waste physical property parameters, collecting temperature parameters, oxygen concentration parameters, water vapor concentration parameters and kiln product component parameters, and obtaining fuel quantity control amount and air quantity control amount in the rotary kiln pyrolysis process;

[0009] The specific method of feeding the waste into the feed end of the rotary kiln, obtaining the physical property parameters of the waste, collecting the temperature parameters, oxygen concentration parameters, water vapor concentration parameters and product component parameters in the kiln, and obtaining the fuel quantity control amount and air quantity control amount in the rotary kiln pyrolysis process is:

[0010] Step S110: feeding the waste to be processed into the feeding end of the rotary kiln through a feeding device, installing a sensor at the feeding device to collect physical property parameters of the waste;

[0011] Step S120: The temperature parameters include: kiln body temperature parameters, waste temperature parameters, and flue gas temperature parameters. Temperature sensors are installed at different positions of the rotary kiln to collect the kiln body temperature parameters of the rotary kiln. An infrared thermometer is installed in the rotary kiln to collect the waste temperature parameters. A thermocouple is installed at the outlet of the rotary kiln to collect the flue gas temperature parameters in the kiln.

[0012] Step S130: The oxygen concentration parameters include: oxygen concentration parameters at the feed end and oxygen concentration parameters at the discharge end. Oxygen concentration sensors are installed at the feed end and the discharge end of the rotary kiln to collect oxygen concentration parameters at the feed end and the discharge end.

[0013] Step S140: arranging a humidity sensor in the rotary kiln to collect water vapor concentration parameters in the rotary kiln;

[0014] Step S150: placing a near infrared spectrometer at the outlet of the rotary kiln to collect component parameters of the products in the kiln;

[0015] Step S160: During the pyrolysis process of the rotary kiln, a control amount generated by the system is obtained, wherein the control amount includes a fuel amount control amount and an air amount control amount;

[0016] Step S200: using waste physical property parameters, temperature parameters and water vapor concentration parameters as input data, and using fuel quantity control amount and air quantity control amount as output data to train a temperature adaptive control model;

[0017] The specific method of training the temperature adaptive control model by taking the waste physical property parameters, temperature parameters and water vapor concentration parameters as input data and the fuel quantity control amount and air quantity control amount as output data is as follows:

[0018] Step S210: Obtain waste physical property parameters, temperature parameters and water vapor concentration parameters, as well as fuel quantity control amount and air quantity control amount in the historical pyrolysis process, construct training samples for training the temperature adaptive control model, preset the prediction time step and the sliding window length, use the waste physical property parameters, temperature parameters and water vapor concentration parameters of the sliding window length as input data, and use the fuel quantity control amount and air quantity control amount of the prediction time step as output data;

[0019] Step S220: Using the LSTM model as the initial model, a temperature adaptive control model is established, including: an input layer, an LSTM layer, a fully connected layer and an output layer, the input layer is used to input input data, the LSTM layer is used to learn the time dependency between the input data and the output data, the fully connected layer is used to map the output of the LSTM layer to the predicted output data through the fully connected layer, and the output layer is used to output the output data;

[0020] Step S230: taking the mean square error between the predicted output data and the actual output data as the loss function of the temperature adaptive control model, and stopping the training when the loss function value between the predicted output data and the actual output data reaches the minimum, thereby obtaining a trained temperature adaptive control model;

[0021] Step S300: Calculate the secondary air volume control amount using the oxygen concentration parameter, the water vapor concentration parameter and the product composition parameter, and predict the fuel volume control amount and the air volume control amount using the temperature adaptive control model;

[0022] The specific method for calculating the secondary air volume control amount using the oxygen concentration parameter, the water vapor concentration parameter and the product component parameter is:

[0023] Step S310: Calculate the oxygen concentration difference according to the oxygen concentration parameter at the feed end and the oxygen concentration parameter at the discharge end. ;

[0024] The oxygen concentration difference The calculation formula is: ,in, is the oxygen concentration parameter at the feed end, is the oxygen concentration parameter at the discharge end;

[0025] Step S320: Calculate the amount of oxygen required for complete combustion of each mole of waste according to the product component parameter n1 of the waste in the rotary kiln. ;

[0026] The amount of oxygen required for complete combustion of each mole of waste is calculated based on the product component parameter n1 of the waste in the rotary kiln. The specific method is:

[0027] The reaction formula for complete combustion of the waste is: ,in, is the chemical formula of the waste;

[0028] Waste in rotary kiln The product component parameter is n1, then the amount of oxygen required for complete combustion is =n1×( );

[0029] Step S330: Obtaining the kiln volume of the rotary kiln , according to the oxygen concentration difference Calculate secondary air volume requirements;

[0030] The calculation formula for the secondary air volume requirement is: , where 22.4 is the value of the molar volume of oxygen;

[0031] Step S340: Calculating a correction parameter of the secondary air volume according to the water vapor concentration parameter;

[0032] The calculation formula of the correction parameter of the secondary air volume is: ,in, is the water vapor concentration parameter, is the correction coefficient of the effect of water vapor concentration parameter on secondary air volume;

[0033] Step S350: The oxygen amount Combined with the correction parameters of the secondary air volume, the actual secondary air volume control amount is calculated ;

[0034] The calculation formula of the secondary air volume control amount is: ;

[0035] Step S400: Calculate the temperature distribution uniformity reward according to the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters, calculate the product component qualification rate reward according to the product component parameters, obtain the actual fuel amount, air amount, and product output, and calculate the energy consumption reward in combination with the secondary air volume control amount to obtain a comprehensive reward function;

[0036] The specific method of calculating the temperature distribution uniformity reward according to the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters, calculating the product component qualification rate reward according to the product component parameters, obtaining the actual fuel amount, air amount, and product output, and calculating the energy consumption reward in combination with the secondary air volume control amount, and obtaining the comprehensive reward function is as follows:

[0037] Step S410: Calculate the standard deviation of the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters at different locations collected by the sensors. ;

[0038] The calculation formula of the standard deviation of the kiln temperature parameter is: ,in, is the jth kiln temperature parameter, is the average value of kiln temperature parameters, n1 is the total number of kiln temperature parameters;

[0039] The calculation formula for the standard deviation of the waste temperature parameter is: ,in, is the bth waste temperature parameter, is the average value of the waste temperature parameter, n2 is the total number of waste temperature parameters;

[0040] The calculation formula of the standard deviation of the flue gas temperature parameter is: ,in, is the cth flue gas temperature parameter, is the average value of the flue gas temperature parameters, n3 is the total number of flue gas temperature parameters;

[0041] Step S420: Calculate the temperature distribution uniformity reward based on the standard deviation of the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters ;

[0042] The calculation formula for the temperature distribution uniformity bonus is: , where α, β, and γ are weight parameters of kiln temperature parameter, waste temperature parameter, and flue gas temperature parameter respectively;

[0043] Step S430: Obtain the target value of product components under ideal conditions, and calculate the product component qualification rate reward according to the actual product component parameters in the kiln ;

[0044] The calculation formula for the product component qualification rate reward is: ,in, is the actual product component parameter, is the target value of the product component under ideal conditions, and δ is the non-zero threshold;

[0045] Step S440: Obtaining the actual amount of fuel used from the control system and air volume , record the product output per unit time P, combined with the secondary air volume control , calculate the energy consumption of the product , the inverse of the product energy consumption As a reward for energy consumption ;

[0046] The calculation formula for the energy consumption of the product is: ;

[0047] The calculation formula for the energy consumption reward is: ;

[0048] Step S450: Reward according to temperature distribution uniformity , Product component qualification rate reward and energy consumption rewards The weighted sum is used to obtain the comprehensive reward function;

[0049] Step S500: Design the state, action and reward function of the agent, build a deep Q network model, calculate the mean of its comprehensive reward function for each control cycle, and when the mean of the comprehensive reward function is greater than the strategy switching threshold, use the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount as the control strategy; when the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, use the action generated by the agent as the control strategy;

[0050] The specific method of designing the state, action and reward function of the intelligent agent and constructing the deep Q network model is:

[0051] Step S510: defining the temperature parameter, the oxygen concentration parameter, the water vapor concentration parameter and the product component parameter as a state s in the state space;

[0052] Step S520: defining the fuel quantity control instruction af, the air quantity control instruction aa, and the secondary air quantity control instruction as as an action space, wherein each action a in the action space corresponds to a different adjustment combination of the fuel valve, the combustion-supporting air valve, and the smoke exhaust valve;

[0053] Step S530: defining the comprehensive reward function as a reward function r;

[0054] Step S540: constructing a deep Q network model, including an input layer, a convolution layer, a pooling layer, an LSTM layer and a fully connected layer, using a dual network structure, including an online network and a target network, the online network is used to predict and update the predicted Q value of the next time step, and the target network is used to calculate the target Q value;

[0055] Step S550: train the deep Q network model using the experience replay method, calculate the predicted Q value of the online network and the target Q value of the target network, and minimize the mean square error between the predicted Q value and the target Q value as the training goal;

[0056] Step S560: During the training process, the agent selects actions based on the ε-greedy strategy, randomly selects actions with a probability of ε, selects actions with the largest predicted Q value with a probability of 1-ε, and ε decreases in an exponential decay manner. The comprehensive reward function of each training is recorded, and the training is completed when the value of the comprehensive reward function reaches the reward threshold;

[0057] For each control cycle, the mean of the comprehensive reward function is calculated. When the mean of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount are used as the control strategy. When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the specific method of using the action generated by the intelligent agent as the control strategy is:

[0058] Step S570: setting a strategy switching threshold, at the end of each control cycle, calculating the mean of the comprehensive reward function of all time steps in the control cycle, and when the mean of the comprehensive reward function is greater than the strategy switching threshold, using the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount as the control strategy;

[0059] Step S580: When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the action generated by the intelligent agent is used as the control strategy, and the current state s is input into the trained deep Q network model. The predicted Q value of each action is obtained by the online network, and the action with the largest predicted Q value is selected to generate the corresponding fuel quantity control instruction af, air quantity control instruction aa, and secondary air quantity control instruction as. The control instructions are output to the fuel valve, combustion-supporting air valve, and smoke exhaust valve to generate corresponding control signals to adjust their openings.

[0060] One aspect of the present application provides an adaptive control system for pyrolysis temperature and oxygen concentration of a rotary kiln, comprising:

[0061] The pyrolysis data acquisition module is used to feed waste into the feed end of the rotary kiln, obtain waste physical property parameters, collect temperature parameters, oxygen concentration parameters, water vapor concentration parameters and product component parameters in the kiln, and obtain the fuel quantity control amount and air quantity control amount in the rotary kiln pyrolysis process;

[0062] A temperature model training module is used to train a temperature adaptive control model by taking waste physical property parameters, temperature parameters and water vapor concentration parameters as input data and taking fuel quantity control quantity and air quantity control quantity as output data;

[0063] A secondary air volume calculation module is used to calculate the secondary air volume control amount using oxygen concentration parameters, water vapor concentration parameters and product component parameters, and to predict the fuel volume control amount and air volume control amount using a temperature adaptive control model;

[0064] The reward function calculation module is used to calculate the temperature distribution uniformity reward based on the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters, calculate the product component qualification rate reward based on the product component parameters, obtain the actual fuel volume, air volume, and product output, and calculate the energy consumption reward in combination with the secondary air volume control amount to obtain a comprehensive reward function;

[0065] The control strategy switching module is used to design the state, action and reward function of the intelligent agent and build a deep Q network model. For each control cycle, the mean of its comprehensive reward function is calculated. When the mean of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount are used as the control strategy. When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the action generated by the intelligent agent is used as the control strategy.

[0066] One aspect of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for adaptively controlling the pyrolysis temperature and oxygen concentration of a rotary kiln are implemented.

[0067] One aspect of the present application provides a readable storage medium storing a computer program, wherein the computer program is suitable for being loaded by a processor to execute the steps in a method for adaptively controlling the pyrolysis temperature and oxygen concentration of a rotary kiln.

[0068] The adaptive control method and system of rotary kiln pyrolysis temperature and oxygen concentration proposed in this application have the following advantages over the prior art:

[0069] By arranging multiple sensors at different locations in the incineration system, all-round perception and data collection of waste properties, incineration process, flue gas emissions, etc. are achieved. Compared with traditional single parameter control, this application has richer perception dimensions and more complete information utilization.

[0070] The temperature adaptive control model uses waste properties, historical temperature and water vapor concentration as input to predict future fuel quantity and primary air volume control, and uses LSTM network to learn the time series correlation in historical data. Compared with simple PID control, this model can adaptively adjust the combustion control parameters according to the dynamic changes of feed, achieving more accurate temperature control.

[0071] By correcting the secondary air volume control amount based on the actual volume and water vapor concentration in the kiln, the blindness of empirical settings is overcome, and measures are taken according to the materials, which helps to improve the completeness of incineration and reduce the emission of harmful gases.

[0072] This application quantifies and weighs multiple objectives of the pyrolysis process, and innovatively integrates multiple sub-objectives into a scalar comprehensive reward function to facilitate the training of the optimization algorithm.

[0073] This application designs a strategy switching mechanism based on the mean of the reward function. While exerting its adaptive optimization capability, this mechanism also sets a "safety threshold" with the reward mean, avoiding blind strategy updates and improving the robustness of control. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 A method flow chart of the adaptive control method of the pyrolysis temperature and oxygen concentration of the rotary kiln provided in the present application;

[0075] Figure 2 Functional module diagram of the adaptive control system of rotary kiln pyrolysis temperature and oxygen concentration provided by the present application;

[0076] Figure 3 A schematic diagram of the structure of an electronic device provided in this application;

[0077] Figure 4 It is a schematic diagram of the structure of a readable storage medium provided by this application. DETAILED DESCRIPTION

[0078] In order to better understand the present application, various aspects of the present application will be described in more detail with reference to the accompanying drawings. It should be understood that these detailed descriptions are only descriptions of exemplary embodiments of the present application and do not limit the scope of the present application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0079] In the accompanying drawings, the size, dimensions and shapes of the elements have been slightly adjusted for ease of illustration. The drawings are for illustration only and are not strictly drawn to scale. As used herein, the terms "substantially", "approximately" and similar terms are used as terms of approximation, not as terms of degree, and are intended to illustrate the inherent deviations in measurements or calculations that will be recognized by those of ordinary skill in the art. In addition, in this application, the order in which the steps are described does not necessarily represent the order in which these processes occur in actual operation, unless otherwise specified or can be derived from the context.

[0080] It should also be understood that expressions such as "including", "comprising", "having", "containing" and / or "comprising" are open rather than closed expressions in this specification, which indicate the presence of the stated features, elements and / or components, but do not exclude the presence of one or more other features, elements, components and / or combinations thereof. In addition, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features rather than just the individual elements in the list. In addition, when describing embodiments of the present application, "may" is used to mean "one or more embodiments of the present application". And, the term "exemplary" is intended to refer to an example or illustration.

[0081] Unless otherwise defined, all words (including engineering terms and scientific and technological terms) used in this article have the same meaning as those commonly understood by ordinary technicians in the field to which this application belongs. It should also be understood that unless there is a clear explanation in this application, words defined in commonly used dictionaries should be interpreted as having the same meaning as their meaning in the context of the relevant technology, and should not be interpreted in an idealized or overly formal sense.

[0082] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0083] Example 1

[0084] like Figure 1 As shown, the adaptive control method of the rotary kiln pyrolysis temperature and oxygen concentration provided by the present application includes:

[0085] Step S100: sending waste into the feed end of the rotary kiln, obtaining waste physical property parameters, collecting temperature parameters, oxygen concentration parameters, water vapor concentration parameters and kiln product component parameters, and obtaining the fuel quantity control amount and air quantity control amount in the rotary kiln pyrolysis process.

[0086] The specific method of feeding the waste into the feed end of the rotary kiln, obtaining the physical property parameters of the waste, collecting the temperature parameters, oxygen concentration parameters, water vapor concentration parameters and product component parameters in the kiln, and obtaining the fuel quantity control amount and air quantity control amount in the rotary kiln pyrolysis process is:

[0087] Step S110: feeding the waste to be processed into the feeding end of the rotary kiln through a feeding device, installing a sensor at the feeding device to collect physical property parameters of the waste;

[0088] The physical property parameters include specific heat capacity, thermal conductivity, and density;

[0089] Step S120: The temperature parameters include: kiln body temperature parameters, waste temperature parameters, and flue gas temperature parameters. Temperature sensors are installed at different positions of the rotary kiln to collect the kiln body temperature parameters of the rotary kiln. An infrared thermometer is installed in the rotary kiln to collect the waste temperature parameters. A thermocouple is installed at the outlet of the rotary kiln to collect the flue gas temperature parameters in the kiln.

[0090] Step S130: The oxygen concentration parameters include: oxygen concentration parameters at the feed end and oxygen concentration parameters at the discharge end. Oxygen concentration sensors are installed at the feed end and the discharge end of the rotary kiln to collect oxygen concentration parameters at the feed end and the discharge end.

[0091] Step S140: arranging a humidity sensor in the rotary kiln to collect water vapor concentration parameters in the rotary kiln;

[0092] Step S150: placing a near infrared spectrometer at the outlet of the rotary kiln to collect component parameters of the products in the kiln;

[0093] Step S160: During the pyrolysis process of the rotary kiln, a control amount generated by the system is obtained, wherein the control amount includes a fuel amount control amount and an air amount control amount;

[0094] The physical properties of the waste are collected through sensors at the feeding device, and sensors are arranged inside and outside the rotary kiln to collect temperature, oxygen concentration, water vapor concentration and product component parameters to obtain comprehensive status information and provide data support for subsequent modeling and control.

[0095] Step S200: Using waste physical property parameters, temperature parameters and water vapor concentration parameters as input data, and using fuel quantity control amount and air quantity control amount as output data, to train a temperature adaptive control model.

[0096] The specific method of training the temperature adaptive control model by taking the waste physical property parameters, temperature parameters and water vapor concentration parameters as input data and the fuel quantity control amount and air quantity control amount as output data is as follows:

[0097] Step S210: Obtain waste physical property parameters, temperature parameters and water vapor concentration parameters, as well as fuel quantity control amount and air quantity control amount in the historical pyrolysis process, construct training samples for training the temperature adaptive control model, preset the prediction time step and the sliding window length, use the waste physical property parameters, temperature parameters and water vapor concentration parameters of the sliding window length as input data, and use the fuel quantity control amount and air quantity control amount of the prediction time step as output data;

[0098] Step S220: Using the LSTM model as the initial model, a temperature adaptive control model is established, including: an input layer, an LSTM layer, a fully connected layer and an output layer, the input layer is used to input input data, the LSTM layer is used to learn the time dependency between the input data and the output data, the fully connected layer is used to map the output of the LSTM layer to the predicted output data through the fully connected layer, and the output layer is used to output the output data;

[0099] Step S230: taking the mean square error between the predicted output data and the actual output data as the loss function of the temperature adaptive control model, and stopping the training when the loss function value between the predicted output data and the actual output data reaches the minimum, thereby obtaining a trained temperature adaptive control model;

[0100] The calculation formula of the loss function of the temperature adaptive control model is: , where N is the total number of training samples, is the actual fuel quantity control value of the i-th training sample, is the predicted fuel quantity control amount of the i-th training sample, is the actual air volume control value of the i-th training sample, is the predicted air volume control amount of the i-th training sample.

[0101] Step S300: Calculate the secondary air volume control amount using the oxygen concentration parameter, the water vapor concentration parameter and the product composition parameter, and predict the fuel volume control amount and the air volume control amount using the temperature adaptive control model.

[0102] The specific method for calculating the secondary air volume control amount using the oxygen concentration parameter, the water vapor concentration parameter and the product component parameter is:

[0103] Step S310: Calculate the oxygen concentration difference according to the oxygen concentration parameter at the feed end and the oxygen concentration parameter at the discharge end. ;

[0104] The oxygen concentration difference The calculation formula is: ,in, is the oxygen concentration parameter at the feed end, is the oxygen concentration parameter at the discharge end;

[0105] Step S320: Calculate the amount of oxygen required for complete combustion of each mole of waste according to the product component parameter n1 of the waste in the rotary kiln. ;

[0106] The amount of oxygen required for complete combustion of each mole of waste is calculated based on the product component parameter n1 of the waste in the rotary kiln. The specific method is:

[0107] The reaction formula for complete combustion of the waste is: ,in, is the chemical formula of the waste;

[0108] Waste in rotary kiln The product component parameter is n1, then the amount of oxygen required for complete combustion is =n1×( );

[0109] Waste from the product components , its product component parameter is n1 mole;

[0110] Step S330: Obtaining the kiln volume of the rotary kiln , according to the oxygen concentration difference Calculate secondary air volume requirements;

[0111] The calculation formula for the secondary air volume requirement is: , where 22.4 is the value of the molar volume of oxygen;

[0112] The molar volume of oxygen refers to the volume occupied by 1 mole of pure oxygen under standard conditions, generally 0°C and 1 atmosphere, which is about 22.4 L / mole;

[0113] Step S340: Calculating a correction parameter of the secondary air volume according to the water vapor concentration parameter;

[0114] The calculation formula of the correction parameter of the secondary air volume is: ,in, is the water vapor concentration parameter, is the correction coefficient of the effect of water vapor concentration parameter on secondary air volume;

[0115] The correction coefficient of the effect of the water vapor concentration parameter on the secondary air volume is set by those skilled in the art based on experience;

[0116] Step S350: The oxygen amount Combined with the correction parameters of the secondary air volume, the actual secondary air volume control amount is calculated ;

[0117] The calculation formula of the secondary air volume control amount is: .

[0118] Step S400: Calculate the temperature distribution uniformity reward based on the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters, calculate the product component qualification rate reward based on the product component parameters, obtain the actual fuel volume, air volume, and product output, and calculate the energy consumption reward in combination with the secondary air volume control amount to obtain a comprehensive reward function.

[0119] The specific method of calculating the temperature distribution uniformity reward according to the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters, calculating the product component qualification rate reward according to the product component parameters, obtaining the actual fuel amount, air amount, and product output, and calculating the energy consumption reward in combination with the secondary air volume control amount, and obtaining the comprehensive reward function is as follows:

[0120] Step S410: Calculate the standard deviation of the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters at different locations collected by the sensors. ;

[0121] The calculation formula of the standard deviation of the kiln temperature parameter is: ,in, is the jth kiln temperature parameter, is the average value of kiln temperature parameters, n1 is the total number of kiln temperature parameters;

[0122] The calculation formula for the standard deviation of the waste temperature parameter is: ,in, is the bth waste temperature parameter, is the average value of the waste temperature parameter, n2 is the total number of waste temperature parameters;

[0123] The calculation formula of the standard deviation of the flue gas temperature parameter is: ,in, is the cth flue gas temperature parameter, is the average value of the flue gas temperature parameters, n3 is the total number of flue gas temperature parameters;

[0124] Step S420: Calculate the temperature distribution uniformity reward based on the standard deviation of the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters ;

[0125] The calculation formula for the temperature distribution uniformity bonus is: , where α, β, and γ are weight parameters of kiln temperature parameter, waste temperature parameter, and flue gas temperature parameter respectively;

[0126] The weight parameters of the kiln temperature parameter, the waste temperature parameter, and the flue gas temperature parameter are set by those skilled in the art according to actual needs;

[0127] Step S430: Obtain the target value of product components under ideal conditions, and calculate the product component qualification rate reward according to the actual product component parameters in the kiln ;

[0128] The calculation formula for the product component qualification rate reward is: ,in, is the actual product component parameter, is the target value of the product component under ideal conditions, and δ is the non-zero threshold;

[0129] The non-zero threshold is a value that is not 0 and is set by those skilled in the art according to actual needs, and is used to determine the pass rate of the actual product component parameters according to the difference between the actual product component parameters and the product component target value under ideal conditions;

[0130] Step S440: Obtaining the actual amount of fuel used from the control system and air volume , record the product output per unit time P, combined with the secondary air volume control , calculate the energy consumption of the product , the inverse of the product energy consumption As a reward for energy consumption ;

[0131] The calculation formula for the energy consumption of the product is: ;

[0132] The calculation formula for the energy consumption reward is: ;

[0133] Step S450: Reward according to temperature distribution uniformity , Product component qualification rate reward and energy consumption rewards The weighted sum is used to obtain the comprehensive reward function;

[0134] The calculation formula of the comprehensive reward function is: ,in, , , Rewards for uniformity of temperature distribution , Product component qualification rate reward and energy consumption rewards The weight coefficient of

[0135] The temperature distribution uniformity reward , Product component qualification rate reward and energy consumption rewards The weight coefficient of is set by those skilled in the art according to actual needs, and .

[0136] Step S500: Design the state, action and reward function of the intelligent agent, build a deep Q network model, and calculate the mean of its comprehensive reward function for each control cycle. When the mean of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount are used as the control strategy. When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the action generated by the intelligent agent is used as the control strategy.

[0137] The state, action and reward function of the design agent are constructed, and a deep Q network model is constructed. For each control cycle, the mean value of the comprehensive reward function is calculated. When the mean value of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount are used as the control strategy. When the mean value of the comprehensive reward function is less than or equal to the strategy switching threshold, the specific method of using the action generated by the agent as the control strategy is:

[0138] Step S510: defining the temperature parameter, the oxygen concentration parameter, the water vapor concentration parameter and the product component parameter as a state s in the state space;

[0139] Step S520: defining the fuel quantity control instruction af, the air quantity control instruction aa, and the secondary air quantity control instruction as as an action space, wherein each action a in the action space corresponds to a different adjustment combination of the fuel valve, the combustion-supporting air valve, and the smoke exhaust valve;

[0140] Step S530: defining the comprehensive reward function as a reward function r;

[0141] Step S540: constructing a deep Q network model, including an input layer, a convolution layer, a pooling layer, an LSTM layer and a fully connected layer, using a dual network structure, including an online network and a target network, the online network is used to predict and update the predicted Q value of the next time step, and the target network is used to calculate the target Q value;

[0142] The input layer is used to receive the state vector of the state space;

[0143] The convolution layer uses one-dimensional convolution, the convolution kernel size is 3, the number of convolution kernels is 32, and the activation function is ReLU, which is used to extract state feature representation;

[0144] The pooling layer adopts maximum pooling with a pooling kernel size of 2 and a step size of 2 to reduce the feature dimension;

[0145] The LSTM layer adopts a two-layer LSTM structure, and the number of hidden units in each LSTM layer is 64, which is used to model the temporal relationship between states;

[0146] The fully connected layer is used to output the Q value Q(s, a) of each action;

[0147] The online network and the target network adopt the same network structure;

[0148] The online network takes the state s of the current time step as input and outputs the predicted Q value corresponding to each action a;

[0149] Step S550: train the deep Q network model using the experience replay method, calculate the predicted Q value of the online network and the target Q value of the target network, and minimize the mean square error between the predicted Q value and the target Q value as the training goal;

[0150] The specific method of using the experience replay method to train the deep Q network model is:

[0151] Initialize the experience replay buffer D with a capacity of ND;

[0152] In each training, the agent interacts with the environment, selects actions according to the ε-greedy strategy, obtains the state and reward of the next moment, and stores the transfer sample (s, a, r, s') in the experience replay buffer D, where s' represents the state of the next time step;

[0153] Randomly sample a batch of transfer samples (s, a, r, s') from the experience replay buffer D;

[0154] For each transfer sample, calculate the predicted Q value of the online network and the target Q value of the target network;

[0155] The mean square error between the predicted Q value and the target Q value is used as the loss function of the deep Q network model, and the value minimizing the loss function is used as the training goal;

[0156] Copy the parameters of the online network to the target network every C time steps;

[0157] Step S560: During the training process, the agent selects actions based on the ε-greedy strategy, randomly selects actions with a probability of ε, selects actions with the largest predicted Q value with a probability of 1-ε, and ε decreases in an exponential decay manner. The comprehensive reward function of each training is recorded, and the training is completed when the value of the comprehensive reward function reaches the reward threshold;

[0158] The calculation formula for the exponential decay of ε is: ,in, is the final exploration rate, is the initial exploration rate, t is the number of trainings, and τ is the decay time constant;

[0159] Preferably, the initial exploration rate is 0.9 or 1, and the final exploration rate is 0.01 or 0.05;

[0160] The value of the decay time constant is set by those skilled in the art according to requirements;

[0161] The agent selects actions based on the ε-greedy strategy, randomly selects actions with a probability of ε, and selects the action with the largest predicted Q value with a probability of 1-ε. This means that when the random number is less than ε, the agent randomly selects an action a, and when the random number is greater than or equal to ε, the agent selects the action with the largest predicted Q value;

[0162] Step S570: setting a strategy switching threshold, at the end of each control cycle, calculating the mean of the comprehensive reward function of all time steps in the control cycle, and when the mean of the comprehensive reward function is greater than the strategy switching threshold, using the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount as the control strategy;

[0163] Step S580: When the mean value of the comprehensive reward function is less than or equal to the strategy switching threshold, the action generated by the agent is used as the control strategy, the current state s is input into the trained deep Q network model, the predicted Q value of each action is obtained by the online network, the action with the largest predicted Q value is selected, and the corresponding fuel quantity control instruction af, air quantity control instruction aa, and secondary air quantity control instruction as are generated, and the control instructions are output to the fuel valve, the combustion-supporting air valve, and the smoke exhaust valve to generate corresponding control signals to adjust their openings;

[0164] The strategy switching threshold and reward threshold are set by those skilled in the art according to requirements.

[0165] Example 2

[0166] like Figure 2 As shown, the adaptive control system of the pyrolysis temperature and oxygen concentration of the rotary kiln provided in this application includes:

[0167] The pyrolysis data acquisition module is used to feed waste into the feed end of the rotary kiln, obtain waste physical property parameters, collect temperature parameters, oxygen concentration parameters, water vapor concentration parameters and product component parameters in the kiln, and obtain the fuel quantity control amount and air quantity control amount in the rotary kiln pyrolysis process;

[0168] A temperature model training module is used to train a temperature adaptive control model by taking waste physical property parameters, temperature parameters and water vapor concentration parameters as input data and taking fuel quantity control quantity and air quantity control quantity as output data;

[0169] A secondary air volume calculation module is used to calculate the secondary air volume control amount using oxygen concentration parameters, water vapor concentration parameters and product component parameters, and to predict the fuel volume control amount and air volume control amount using a temperature adaptive control model;

[0170] The reward function calculation module is used to calculate the temperature distribution uniformity reward based on the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters, calculate the product component qualification rate reward based on the product component parameters, obtain the actual fuel volume, air volume, and product output, and calculate the energy consumption reward in combination with the secondary air volume control amount to obtain a comprehensive reward function;

[0171] The control strategy switching module is used to design the state, action and reward function of the intelligent agent and build a deep Q network model. For each control cycle, the mean of its comprehensive reward function is calculated. When the mean of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount are used as the control strategy. When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the action generated by the intelligent agent is used as the control strategy.

[0172] Example 3

[0173] Figure 3 Schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3 As shown, according to another aspect of the present application, an electronic device is also provided. The electronic device may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are run by one or more processors, the self-adaptive control method of the pyrolysis temperature and oxygen concentration of the rotary kiln as described above can be executed.

[0174] The method or system according to the embodiment of the present application can also be Figure 3 The electronic device architecture shown in FIG. Figure 3As shown, the electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, and the like. A storage device in the electronic device, such as a ROM or a hard disk, may store the adaptive control method for the pyrolysis temperature and oxygen concentration of a rotary kiln provided in the present application. The adaptive control method for the pyrolysis temperature and oxygen concentration of a rotary kiln may, for example, include: feeding waste into the feed end of the rotary kiln, obtaining physical property parameters of the waste, collecting temperature parameters, oxygen concentration parameters, water vapor concentration parameters, and product component parameters in the kiln, and obtaining the fuel quantity control amount and air quantity control amount in the pyrolysis process of the rotary kiln; using the physical property parameters of the waste, temperature parameters, and water vapor concentration parameters as input data, and using the fuel quantity control amount and the air quantity control amount as output data, to train a temperature adaptive control model; using the oxygen concentration parameters, water vapor concentration parameters, and product component parameters to calculate the secondary air volume control amount, and using the temperature adaptive control model to predict the fuel quantity control amount and the air volume control amount; according to the kiln body temperature parameters, The temperature distribution uniformity reward is calculated based on the waste temperature parameters and flue gas temperature parameters, and the product component qualification rate reward is calculated based on the product component parameters. The actual fuel volume, air volume and product output are obtained, and the energy consumption reward is calculated in combination with the secondary air volume control amount to obtain a comprehensive reward function; the state, action and reward function of the intelligent agent are designed, and a deep Q network model is constructed. For each control cycle, the mean of its comprehensive reward function is calculated. When the mean of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount are used as the control strategy. When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the action generated by the intelligent agent is used as the control strategy. Furthermore, the electronic device may also include a user interface. Of course, Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown.

[0175] Example 4

[0176] Figure 4 Schematic diagram of a readable storage medium structure provided by an embodiment of the present application. Figure 4 As shown, it is a readable storage medium according to an embodiment of the present application. Computer-readable instructions are stored on the computer-readable storage medium. When the computer-readable instructions are executed by the processor, the adaptive control method of the pyrolysis temperature and oxygen concentration of the rotary kiln according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache). Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0177] In addition, according to the implementation mode of the present application, the process described in the above reference flow chart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, such as: sending waste into the feed end of the rotary kiln, obtaining waste physical property parameters, collecting temperature parameters, oxygen concentration parameters, water vapor concentration parameters and product component parameters in the kiln, and obtaining the fuel quantity control amount and air quantity control amount in the rotary kiln pyrolysis process; using waste physical property parameters, temperature parameters and water vapor concentration parameters as input data, and using fuel quantity control amount and air quantity control amount as output data, to train a temperature adaptive control model; using oxygen concentration parameters, water vapor concentration parameters and product component parameters to calculate the secondary air volume control amount, and using temperature adaptive control model to calculate the secondary air volume control amount; ... The adaptive control model predicts the fuel quantity control quantity and the air quantity control quantity; calculates the temperature distribution uniformity reward according to the kiln temperature parameters, the waste temperature parameters, and the flue gas temperature parameters, calculates the product component qualification rate reward according to the product component parameters, obtains the actual fuel quantity, air quantity, and product output, and calculates the energy consumption reward in combination with the secondary air volume control quantity to obtain a comprehensive reward function; designs the state, action, and reward function of the intelligent agent, constructs a deep Q network model, and calculates the mean of its comprehensive reward function for each control cycle. When the mean of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control quantity and the predicted fuel quantity control quantity and air quantity control quantity are used as the control strategy. When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the action generated by the intelligent agent is used as the control strategy. When the computer program is executed by the central processing unit (CPU), the above functions defined in the method of this application are executed.

[0178] The methods, apparatuses, and devices of the present application may be implemented in many ways. For example, the methods, apparatuses, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0179] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0180] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. The adaptive control method of rotary kiln pyrolysis temperature and oxygen concentration is characterized in that: include: Send the waste into the feed end of the rotary kiln, obtain the physical property parameters of the waste, collect the temperature parameters, oxygen concentration parameters, water vapor concentration parameters and product component parameters in the kiln, and obtain the fuel quantity control amount and air quantity control amount in the rotary kiln pyrolysis process; The waste physical property parameters, temperature parameters and water vapor concentration parameters are used as input data, and the fuel quantity control quantity and air quantity control quantity are used as output data to train the temperature adaptive control model; The secondary air volume control amount is calculated using oxygen concentration parameters, water vapor concentration parameters and product composition parameters, and the fuel volume control amount and air volume control amount are predicted using the temperature adaptive control model; The temperature distribution uniformity reward is calculated based on the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters. The product component qualification rate reward is calculated based on the product component parameters. The actual fuel volume, air volume, and product output are obtained. Combined with the secondary air volume control amount, the energy consumption reward is calculated to obtain a comprehensive reward function. Design the state, action and reward function of the agent, build a deep Q network model, and calculate the mean of its comprehensive reward function for each control cycle. When the mean of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount are used as the control strategy. When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the action generated by the agent is used as the control strategy. The specific method for calculating the secondary air volume control amount using the oxygen concentration parameter, the water vapor concentration parameter and the product component parameter is: Calculate the oxygen concentration difference based on the oxygen concentration parameters at the feed end and the oxygen concentration parameters at the discharge end. ; The oxygen concentration difference The calculation formula is: ,in, is the oxygen concentration parameter at the feed end, is the oxygen concentration parameter at the discharge end; Based on the product component parameter n1 of the waste in the rotary kiln, the amount of oxygen required for complete combustion of each mole of waste is calculated. ; Get the kiln volume of the rotary kiln , according to the oxygen concentration difference Calculate secondary air volume requirements; The calculation formula for the secondary air volume requirement is: , where 22.4 is the value of the molar volume of oxygen; Calculate the correction parameter of the secondary air volume according to the water vapor concentration parameter; The calculation formula of the correction parameter of the secondary air volume is: ,in, is the water vapor concentration parameter, is the correction coefficient of the effect of water vapor concentration parameter on secondary air volume; The amount of oxygen Combined with the correction parameters of the secondary air volume, the actual secondary air volume control amount is calculated ; The calculation formula of the secondary air volume control amount is: ; The specific method of calculating the temperature distribution uniformity reward according to the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters, calculating the product component qualification rate reward according to the product component parameters, obtaining the actual fuel amount, air amount, and product output, and calculating the energy consumption reward in combination with the secondary air volume control amount, and obtaining the comprehensive reward function is as follows: According to the kiln temperature parameters, waste temperature parameters and flue gas temperature parameters collected by sensors at different locations, the standard deviation of the kiln temperature parameters, waste temperature parameters and flue gas temperature parameters is calculated. ; Calculate the temperature distribution uniformity reward based on the standard deviation of kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters ; The calculation formula for the temperature distribution uniformity bonus is: , where α, β, and γ are weight parameters of kiln temperature parameter, waste temperature parameter, and flue gas temperature parameter respectively; Obtain the target value of product components under ideal conditions, and calculate the product component qualification rate reward based on the actual product component parameters in the kiln ; The calculation formula for the product component qualification rate reward is: ,in, is the actual product component parameter, is the target value of the product component under ideal conditions, and δ is the non-zero threshold; Get the actual amount of fuel used from the control system and air volume , record the product output per unit time P, combined with the secondary air volume control , calculate the energy consumption of the product , the inverse of the product energy consumption As a reward for energy consumption ; The calculation formula for the energy consumption of the product is: ; The calculation formula for the energy consumption reward is: ; Reward based on temperature distribution uniformity , Product component qualification rate reward and energy consumption rewards The weighted sum is used to obtain the comprehensive reward function; The specific method of designing the state, action and reward function of the intelligent agent and constructing the deep Q network model is: The temperature parameter, oxygen concentration parameter, water vapor concentration parameter and product component parameter are defined as the state s of the state space; The fuel quantity control command af, the air quantity control command aa, and the secondary air quantity control command as are defined as an action space, and each action a in the action space corresponds to a different adjustment combination of the fuel valve, the combustion-supporting air valve, and the smoke exhaust valve; Define the comprehensive reward function as the reward function r; Construct a deep Q network model, including input layer, convolution layer, pooling layer, LSTM layer and fully connected layer, using a dual network structure, including an online network and a target network. The online network is used to predict and update the predicted Q value of the next time step, and the target network is used to calculate the target Q value. The deep Q network model is trained using the experience replay method to calculate the predicted Q value of the online network and the target Q value of the target network, and the training goal is to minimize the mean square error between the predicted Q value and the target Q value. During the training process, the agent selects actions based on the ε-greedy strategy, randomly selects actions with a probability of ε, and selects the action with the largest predicted Q value with a probability of 1-ε. ε decreases in an exponential decay manner. The comprehensive reward function of each training is recorded, and the training is completed when the value of the comprehensive reward function reaches the reward threshold; For each control cycle, the mean of the comprehensive reward function is calculated. When the mean of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount are used as the control strategy. When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the specific method of using the action generated by the intelligent agent as the control strategy is: A strategy switching threshold is set, and at the end of each control cycle, the mean value of the comprehensive reward function of all time steps in the control cycle is calculated. When the mean value of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount are used as the control strategy; When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the action generated by the intelligent agent is used as the control strategy, and the current state s is input into the trained deep Q network model. The predicted Q value of each action is obtained by the online network, and the action with the largest predicted Q value is selected to generate the corresponding fuel quantity control instruction af, air quantity control instruction aa, and secondary air quantity control instruction as. The control instructions are output to the fuel valve, combustion-supporting air valve, and smoke exhaust valve to generate corresponding control signals and adjust their openings.

2. The adaptive control method for pyrolysis temperature and oxygen concentration of a rotary kiln according to claim 1, characterized in that: The specific method of feeding the waste into the feed end of the rotary kiln, obtaining the physical property parameters of the waste, collecting the temperature parameters, oxygen concentration parameters, water vapor concentration parameters and product component parameters in the kiln, and obtaining the fuel quantity control amount and air quantity control amount in the rotary kiln pyrolysis process is: The waste to be treated is fed into the feeding end of the rotary kiln through a feeding device, and a sensor is installed at the feeding device to collect the physical property parameters of the waste; The temperature parameters include: kiln body temperature parameters, waste temperature parameters, flue gas temperature parameters, temperature sensors are installed at different positions of the rotary kiln to collect the kiln body temperature parameters of the rotary kiln, infrared thermometers are installed in the rotary kiln to collect waste temperature parameters, and thermocouples are installed at the outlet of the rotary kiln to collect the flue gas temperature parameters in the kiln; The oxygen concentration parameters include: oxygen concentration parameters at the feed end and oxygen concentration parameters at the discharge end. Oxygen concentration sensors are installed at the feed end and the discharge end of the rotary kiln to collect oxygen concentration parameters at the feed end and the discharge end. Humidity sensors are arranged in the rotary kiln to collect water vapor concentration parameters in the rotary kiln; A near-infrared spectrometer is placed at the outlet of the rotary kiln to collect the composition parameters of the products in the kiln; During the pyrolysis process of the rotary kiln, a control quantity generated by the system is obtained, wherein the control quantity includes a fuel quantity control quantity and an air quantity control quantity.

3. The adaptive control method for pyrolysis temperature and oxygen concentration of a rotary kiln according to claim 2, characterized in that: The specific method of training the temperature adaptive control model by taking the waste physical property parameters, temperature parameters and water vapor concentration parameters as input data and the fuel quantity control amount and air quantity control amount as output data is as follows: Obtain waste physical property parameters, temperature parameters and water vapor concentration parameters, as well as fuel quantity control amount and air quantity control amount in the historical pyrolysis process, construct training samples for training the temperature adaptive control model, preset the prediction time step and sliding window length, use the waste physical property parameters, temperature parameters and water vapor concentration parameters of the sliding window length as input data, and use the fuel quantity control amount and air quantity control amount of the prediction time step as output data; The LSTM model is used as the initial model to establish a temperature adaptive control model, including: input layer, LSTM layer, fully connected layer and output layer. The input layer is used to input input data, the LSTM layer is used to learn the time dependency between input data and output data, the fully connected layer is used to map the output of the LSTM layer to the predicted output data through the fully connected layer, and the output layer is used to output the output data. The mean square error between the predicted output data and the actual output data is used as the loss function of the temperature adaptive control model. When the loss function value between the predicted output data and the actual output data reaches the minimum, the training is stopped to obtain the trained temperature adaptive control model.

4. The adaptive control method for pyrolysis temperature and oxygen concentration of a rotary kiln according to claim 3, characterized in that: The amount of oxygen required for complete combustion of each mole of waste is calculated based on the product component parameter n1 of the waste in the rotary kiln. The specific method is: The reaction formula for complete combustion of the waste is: ,in, is the chemical formula of the waste; Waste in rotary kiln The product component parameter is n1, then the amount of oxygen required for complete combustion is =n1×( ).

5. An adaptive control system for pyrolysis temperature and oxygen concentration of a rotary kiln, which is used to implement the adaptive control method for pyrolysis temperature and oxygen concentration of a rotary kiln according to any one of claims 1 to 4, characterized in that: include: The pyrolysis data acquisition module is used to feed waste into the feed end of the rotary kiln, obtain waste physical property parameters, collect temperature parameters, oxygen concentration parameters, water vapor concentration parameters and product component parameters in the kiln, and obtain the fuel quantity control amount and air quantity control amount in the rotary kiln pyrolysis process; A temperature model training module is used to train a temperature adaptive control model by taking waste physical property parameters, temperature parameters and water vapor concentration parameters as input data and taking fuel quantity control quantity and air quantity control quantity as output data; A secondary air volume calculation module is used to calculate the secondary air volume control amount using oxygen concentration parameters, water vapor concentration parameters and product component parameters, and to predict the fuel volume control amount and air volume control amount using a temperature adaptive control model; The reward function calculation module is used to calculate the temperature distribution uniformity reward based on the kiln temperature parameters, waste temperature parameters, and flue gas temperature parameters, calculate the product component qualification rate reward based on the product component parameters, obtain the actual fuel volume, air volume, and product output, and calculate the energy consumption reward in combination with the secondary air volume control amount to obtain a comprehensive reward function; The control strategy switching module is used to design the state, action and reward function of the intelligent agent and build a deep Q network model. For each control cycle, the mean of its comprehensive reward function is calculated. When the mean of the comprehensive reward function is greater than the strategy switching threshold, the secondary air volume control amount and the predicted fuel volume control amount and air volume control amount are used as the control strategy. When the mean of the comprehensive reward function is less than or equal to the strategy switching threshold, the action generated by the intelligent agent is used as the control strategy.

6. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the adaptive control method of the pyrolysis temperature and oxygen concentration of the rotary kiln as described in any one of claims 1 to 4 are implemented.

7. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the steps in the adaptive control method of the pyrolysis temperature and oxygen concentration of a rotary kiln as described in any one of claims 1-4.