A data-driven multi-objective optimization control method for municipal solid waste incineration process

Abstract

The application provides a kind of data-driven urban solid waste incineration process multi-objective optimization control method, first, the MSWI process whole process model facing pollutant emission is established, including furnace temperature, boiler steam flow and flue gas oxygen content controlled object model and CO, CO2 and NOx pollutant emission index model, then, using single neuron adaptive PID establishes the multi-input multi-output loop controller facing key controlled variable, realizes the stable control to combustion process, finally, combined with the knowledge of field experts, the adaptive solution of key controlled variable optimization set value based on multi-objective particle swarm optimization algorithm is realized.The data-driven urban solid waste incineration process multi-objective optimization control method provided by the application can realize the multi-objective optimization control of urban solid waste incineration process, guide the existing running factory, reduce the pollutant emission concentration and improve the combustion efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of municipal solid waste treatment, in particular to a data-driven multi-objective optimization control method for municipal solid waste incineration process. BACKGROUND

[0002] Municipal solid waste (MSW) is a major factor leading to the gradual worsening of urban pollution problems. With the increasing emphasis on garbage classification by the country, the heat value of MSW is gradually increasing, making MSWI technology an important means of treating MSW. Compared with other treatment methods, MSWI technology, which has the characteristics of harmless, reduction and resource, can save a large amount of land resources and has obvious social, economic and environmental benefits. Due to the complexity of MSW composition and low heat value in China, the imported MSWI technology from abroad is difficult to effectively support the normal operation of the plant, and currently relies on manual operation by experts in the field. However, the difference in expert experience and the delay in problem solving make it difficult for MSWI power plants in China to maintain stable operating conditions, resulting in short-term over-standard emissions of pollutants, insufficient MSW combustion, and other problems. Therefore, how to reduce the concentration of the above-mentioned pollutants and improve the efficiency of MSW combustion is a problem that needs to be solved. In particular, the main pollutants such as dioxin, which causes the "not in my backyard" effect of incineration plants, NOx, which causes acid rain, and CO2, only need to reduce the concentration of emissions.

[0003] Generally, establishing an accurate controlled object model is the basis for intelligent optimization control research. Due to the complexity of the mechanism, the frequent fluctuation of the working condition, and the uncertainty of the disturbance, it is difficult to construct an accurate mathematical model of the MSWI process. Currently, researchers mostly use industrial process data to construct the controlled object model. However, in existing research, modeling is mostly focused on a single controlled variable, ignoring the existence of multiple controlled variables and their coupling in the MSWI process. Achieving stable control of key controlled variables in the MSWI process is a prerequisite for solving the optimization of their set values. Currently, the control of the MSWI process mostly uses PID controllers and model predictive control, and there is no related research on multi-loop controllers for key controlled variables of the MSWI process that conforms to industrial reality. For the optimization problem of the MSWI process, the furnace temperature, boiler steam flow, and flue gas oxygen content as key controlled variables of the MSWI process, there is no related research on optimization settings. Currently, there is no report on the optimization control of the MSWI process with the above-mentioned pollutant emission concentration and MSW combustion efficiency as the target. Therefore, it is necessary to design a data-driven multi-objective optimization control method for municipal solid waste incineration process. SUMMARY

[0004] The application aims to provide a data-driven multi-objective optimization control method for municipal solid waste incineration process, which can realize multi-objective optimization control of municipal solid waste incineration process, guide the existing running factory, reduce the pollutant emission concentration and improve the combustion efficiency.

[0005] To achieve the above object, the application provides the following scheme.

[0006] A data-driven multi-objective optimization control method for municipal solid waste incineration process comprises the following steps.

[0007] Step 1: analyzing the influence factors for multi-objective optimization;

[0008] Step 2: detecting a multi-objective optimization control model according to the influence factors.

[0009] Optionally, in step 1, the influence factors for multi-objective optimization are analyzed, and specifically, the influence factors for multi-objective optimization are analyzed.

[0010] The constraint conditions of MSWI are as follows.

[0011]

[0012] In the formula, t is an environmental or time variable, and x is a set value of a process operation variable.

[0013] To ensure that the environmental indicators meet the standards, there are:

[0014] min F1 m (t,x), m=1,2,.... (2)

[0015] To improve the product indicators, there are:

[0016] max F2 n (t,x), n=1,2,... (3)

[0017] To improve the economic indicators, there are:

[0018] max F3 l (t,x), l=1,2,... (4)

[0019] The optimization objectives are homogenized, and the optimization objectives of MSWI are expressed as:

[0020]

[0021]

[0022] By solving x, the optimized operation of the MSWI process is realized, wherein x includes the furnace temperature, the boiler steam flow and the oxygen content of flue gas.

[0023] Optionally, in step 1, the multi-objective optimization control model is detected according to the influencing factors, and specifically includes the following steps:

[0024] Step 301: Establish a full-process model oriented to pollutant emission reduction as a controlled object model;

[0025] Step 302: Establish a single neuron adaptive PID multi-input multi-output loop controller to stably control the controlled object model;

[0026] Step 303: Establish a controlled variable optimization setting solving model based on multi-objective PSO to realize adaptive solving of the optimized setting value of the key controlled variable, and send the solving result to the single neuron adaptive PID multi-input multi-output loop controller to stably control the controlled object model.

[0027] Optionally, in step 301, the full-process model oriented to pollutant emission reduction is established, specifically as follows:

[0028] The full-process model oriented to pollutant emission reduction includes a series controlled object model and a parallel pollutant index model, a single neuron adaptive PID multi-input multi-output loop controller is connected to the series controlled object model, the series controlled object model is connected to the parallel pollutant index model, and the parallel pollutant index model is connected to a controlled variable optimization setting solving model based on multi-objective PSO;

[0029] The series controlled object model includes a furnace temperature model, a boiler steam flow model and a flue gas oxygen content model, wherein the furnace temperature model, the boiler steam flow model and the flue gas oxygen content model are all constructed by a linear regression decision tree algorithm based on Tikhonov regularization, a single neuron adaptive PID multi-input multi-output loop controller is connected to the furnace temperature model, the boiler steam flow model and the flue gas oxygen content model, u PriAir , u Feeder and u Dry are input to the furnace temperature model, u FeederWater is input to the boiler steam flow model, and u SecAir is input to the flue gas oxygen content model, the furnace temperature model is connected to the boiler steam flow model and the flue gas oxygen content model, and the furnace temperature model, the boiler steam flow model and the flue gas oxygen content model are connected to the parallel pollutant index model, wherein u FeederWater , u PriAir , u Feeder , u Dry and u SecAir represent the outputs of the feed water amount, the primary air amount, the feeder speed, the dry grate speed and the secondary air amount, respectively;

[0030] The parallel pollution index model comprises a CO model, a CO2 model and a NOx model, wherein the CO model, the CO2 model and the NOx model are all constructed through a linear regression decision tree algorithm based on Tikhonov regularization, the furnace temperature model, the boiler steam flow model and the flue gas oxygen content model are connected with the CO model, the CO2 model and the NOx model respectively, and the CO model, the CO2 model and the NOx model are connected with the controlled variable optimization setting solving model based on multi-objective PSO.

[0031] Optionally, in step 302, a single neuron adaptive PID multi-input multi-output loop controller is established, specifically:

[0032] The single neuron adaptive PID multi-input multi-output loop controller comprises a furnace temperature controller taking the feeder speed, the dry grate speed and the primary air volume as the operation variables, a boiler steam flow controller taking the feed water volume as the operation variable and a flue gas oxygen content controller taking the secondary air volume as the operation variable, wherein the furnace temperature controller, the boiler steam flow controller and the flue gas oxygen content controller are all realized through a single neuron adaptive PID algorithm, the controlled variable optimization setting solving model based on multi-objective PSO is connected with the furnace temperature controller, the boiler steam flow controller and the flue gas oxygen content controller respectively, the furnace temperature controller is connected with the furnace temperature model, the boiler steam flow controller is connected with the boiler steam flow model, and the flue gas oxygen content controller is connected with the flue gas oxygen content model.

[0033] Optionally, in step 303, the controlled variable optimization setting solving model based on multi-objective PSO is established, specifically:

[0034] The output values of the CO model, the CO2 model and the NOx model, i.e. the sum of the emission values of CO, CO2 and NOx, are taken as the comprehensive pollution emission concentration to be minimized, and the MSW combustion efficiency is maximized to establish the controlled variable optimization setting solving model based on multi-objective PSO as follows:

[0035]

[0036] s.t.900℃≤r FT ≤950℃

[0037] 76t / h≤r BSF ≤78t / h

[0038] 6.5%≤r OX ≤8.5% (6)

[0039] In the formula, and respectively represent the CO, CO2 and NOx index model outputs, r FT , r BSF and r OXRespectively represent the set value of the furnace temperature, the boiler steam flow and the flue gas oxygen content in the multi-loop control system, and the controlled variables, i.e., the optimized set value of the furnace temperature, the boiler steam flow and the flue gas oxygen content, are solved based on an improved multi-objective particle swarm optimization algorithm, wherein the improved multi-objective particle swarm optimization algorithm comprises population initialization, updating of particle individual optimum, global optimum and archive, updating of particle speed and position, variation of population, detection of boundary limit, calculation of fitness value, judgment of termination condition and judgment of output optimal set value according to expert experience.

[0040] According to the specific embodiments of the present application, the following technical effects are disclosed: the data-driven multi-objective optimization control method for municipal solid waste incineration process provided by the present application, according to the MSWI process analysis for multi-objective optimization, a multi-objective optimization model is established, which takes the key controlled variables as the decision variables, minimizes the comprehensive pollutant emission concentration and maximizes the MSW combustion efficiency, and a corresponding multi-objective optimization control strategy is proposed, specifically, first, the MSWI process whole-process model facing the pollutant emission is established, including the controlled object model of the furnace temperature, the boiler steam flow and the flue gas oxygen content, and the CO, CO2 and NOx pollutant emission index model; then, the single neuron adaptive PID is used to establish the multi-input and multi-output loop controller facing the key controlled variables, to realize the stable control of the combustion process; finally, combined with the field expert knowledge, the adaptive solution of the optimized set value of the key controlled variables based on the multi-objective particle swarm optimization algorithm is realized, based on the field data of a certain MSWI power plant, it is verified that the proposed method can obtain the optimized set value of the three key controlled variables, reduce the pollutant emission concentration and improve the combustion efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0042] Figure 1 It is a process flow diagram of MSWI power plant;

[0043] Figure 2 It is a process flow diagram of the data-driven multi-objective optimization control method for municipal solid waste incineration process in the embodiments of the present application;

[0044] Figure 3 It is a structure schematic diagram of the multi-objective optimization control model;

[0045] Figure 4 It is a structure schematic diagram of the input and output relationship of the combustion process model;

[0046] Figure 5 SNAP ID controller structure schematic;

[0047] Figure 6a FT model test set fit curve schematic;

[0048] Figure 6b BSF model test set fit curve schematic;

[0049] Figure 6c OX model test set fit curve schematic;

[0050] Figure 7a NOx model test set fit curve schematic;

[0051] Figure 7b CO model test set fit curve schematic;

[0052] Figure 7c CO2 model test set fit curve schematic;

[0053] Figure 8a Furnace temperature tracking curve schematic;

[0054] Figure 8b Boiler steam flow tracking curve schematic;

[0055] Figure 8c Flue gas oxygen content tracking curve schematic;

[0056] Figure 9a Furnace temperature error curve schematic;

[0057] Figure 9b Boiler steam flow error curve schematic;

[0058] Figure 9c Flue gas oxygen content error curve schematic;

[0059] Figure 10a Furnace temperature IAE index change curve schematic;

[0060] Figure 10b Boiler steam flow IAE index change curve schematic;

[0061] Figure 10c Flue gas oxygen content IAE index change curve schematic;

[0062] Figure 11a Furnace temperature ISE index change curve schematic;

[0063] Figure 11b Boiler steam flow ISE index change curve schematic;

[0064] Figure 11c ISE index change curve of flue gas oxygen content;

[0065] Figure 12a MOPSO algorithm solving result schematic view;

[0066] Figure 12b MOPSO algorithm solving result schematic view. DETAILED DESCRIPTION

[0067] The purpose of the present application is to provide a data-driven municipal solid waste incineration process multi-objective optimization control method, which can realize municipal solid waste incineration process multi-objective optimization control, guide the existing running factory, reduce the pollutant emission concentration, and improve the combustion efficiency.

[0068] In order to make the above-mentioned purpose, characteristics and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below in combination with the drawings and specific embodiments.

[0069] The process flow of MSWI power plant is shown in the figure. Figure 1

[0070] As shown in the figure, the data-driven municipal solid waste incineration process multi-objective optimization control method provided by the embodiment of the present application comprises the following steps: Figure 2

[0071] Step 1: analyze the influence factors for multi-objective optimization;

[0072] Step 2: detect the multi-objective optimization control model according to the influence factors.

[0073] In step 1, the influence factors for multi-objective optimization are analyzed, specifically:

[0074] The constraint conditions of MSWI are:

[0075]

[0076] In the formula, t is the environmental or time variable, and x is the set value of the process operation variable;

[0077] In order to ensure that the environmental indicators (reduce the emission concentration of solid, liquid and gaseous pollutants and greenhouse gases, etc.) meet the standards, there are:

[0078] min F1 m (t,x), m=1,2,.... (2)

[0079] In order to improve the product indicators (reduce the furnace slag loss on ignition rate and fly ash production, improve the combustion efficiency and organic matter removal rate, etc.), there are:

[0080] ​​

[0081] To improve economic indicators (reduce MSW treatment costs, increase on-grid power generation, etc.), there are:

[0082]

[0083] The optimization objectives are homogenized, and the optimization objectives of MSWI are expressed as:

[0084]

[0085]

[0086] By solving x, the MSWI process optimization operation is realized, wherein x includes the furnace temperature, the boiler steam flow and the flue gas oxygen content.

[0087] Analysis shows that in order to realize the emission control of main pollutants (CO, CO2 and NOx) and maximize the MSW combustion efficiency, the optimal set value of key controlled variables (furnace temperature, boiler steam flow and flue gas oxygen content) needs to be solved;

[0088] As shown in Figure 3 Step 1, a multi-objective optimization control model is detected according to the influencing factors, which specifically includes the following steps:

[0089] Step 301: Establish a full-process model oriented to pollutant emission reduction as a controlled object model;

[0090] Step 302: Establish a single neuron adaptive PID multi-input multi-output loop controller to stably control the controlled object model;

[0091] Step 303: Establish a controlled variable optimization setting solving model based on multi-objective PSO, which is used to realize adaptive solving of the optimal set value of the key controlled variable, and send the solving result to the single neuron adaptive PID multi-input multi-output loop controller to stably control the controlled object model.

[0092] The functions of the above three modules are as follows:

[0093] 1) Controlled variable optimization setting solving based on multi-objective PSO: This module uses multi-objective PSO algorithm to solve the Pareto front of multi-objective optimization model, and outputs the optimal set value combined with the experience of field experts;

[0094] 2) Single neuron adaptive PID multi-input multi-output loop control: Based on the characteristic analysis in the foregoing, this module establishes a multi-loop controller of key controlled variables of MSWI process, and realizes effective tracking of the controlled object by adjusting the controller output;

[0095] 3) Full process model oriented to pollutant emission reduction: This module serves as the basis for intelligent optimization algorithm research, providing pollutant emission index model output and multi-input multi-output loop controlled object model for multi-objective optimization model.

[0096] For Figure 3 are explained, wherein γ emission represents the tail gas emission model output, and represent the NOx model, CO model and CO2 model output respectively; CE represents the combustion efficiency; r FT , r BSF and r OX represent the FT, BSF and OX particle dimensions respectively; and represent the FT, BSF and OX optimal set values respectively; and represent the FT model, BSF model and OX model output respectively; e FT , e BSF and e OX represent the FT, BSF and OX error values respectively; and represent the FT controller error state transition output; and represent the BSF controller error state transition output; and represent the OX controller error state transition output; u FeederWater , u PriAir , u Feeder , u Dry and u SecAir represent the feed water amount, primary air amount, feeder speed, drying grate speed and secondary air amount output respectively.

[0097] In step 301, a full process model oriented to pollutant emission reduction is established, specifically:

[0098] In order to eliminate different noises existing in process data, the present application adopts wavelet denoising data to establish a full process model oriented to pollutant emission reduction, including a series controlled object model and a parallel pollutant index model, a single neuron adaptive PID multi-input multi-output loop controller connecting the series controlled object model, the series controlled object model connecting the parallel pollutant index model, and the parallel pollutant index model connecting a controlled variable optimization set solving model based on multi-objective PSO;

[0099] As Figure 4As shown, the series controlled object model includes a furnace temperature model, a boiler steam flow model, and a flue gas oxygen content model. These models are all constructed using a Tikhonov-regularized linear regression decision tree algorithm. A single-neuron adaptive PID multi-input multi-output loop controller connects the furnace temperature model, boiler steam flow model, and flue gas oxygen content model, inputting u to the furnace temperature model. PriAir u Feeder and u Dry Input u into the boiler steam flow model FeederWater Input u into the flue gas oxygen content model SecAir The furnace temperature model is connected to the boiler steam flow model and the flue gas oxygen content model. The furnace temperature model, boiler steam flow model, and flue gas oxygen content model are connected to a parallel pollutant index model. Among them, u FeederWater u PriAir u Feeder u Dry and u SecAir These represent the output of water supply, primary air volume, feeder speed, drying grate speed, and secondary air volume, respectively.

[0100] In this invention, the furnace temperature, boiler steam flow rate, and flue gas oxygen content are respectively denoted as f. FT (·), f BSF (·) and f OX (·), all are constructed using the Tikhonov-regularized linear regression decision tree (TR-LRDT) algorithm. Its advantage is strong interpretability, with f FT Taking (·) as an example, the modeling process is described as follows:

[0101] In conjunction with industrial practice, f FT The input of (·) is denoted as u. FT =[u PriAir ,u Feeder ,u Dry ]

[0102] First, the optimal splitting variables and splitting points are found by iterating through the data until the number of leaf node samples is less than an empirically set threshold θ. FT The following criteria are used to divide the input feature space into K FT each region get:

[0103]

[0104] In the formula, and Representing regions and True values ​​of the middle sample; and respectively represent the sample mean values in the regions and respectively;

[0105] The CART tree's leaf node prediction value only considers the sample mean value, ignoring the relationship between the process variable and the true value. The present application improves it to calculate the prediction output of the CART tree's leaf node by using the linear regression method, as follows:

[0106]

[0107] In the formula, and respectively represent the prediction output, input and weight vector of the leaf node;

[0108] Considering the fact that the number of samples is greater than the dimension of the characteristics, obtaining the weight vector can be attributed to solving a kind of overdetermined matrix equation. In order to ensure the convergence of the weight vector, the regularization least square loss function is used here, as follows:

[0109]

[0110] In the formula, λ FT represents the positive regularization term coefficient with a positive value;

[0111] Further, the gradient of the above loss function with respect to can be expressed as:

[0112]

[0113] Let

[0114]

[0115] Finally, according to the weight vector obtained from the above formula, the leaf node prediction value is calculated based on formula (10), considering all leaf nodes, so the furnace temperature model can be expressed as:

[0116]

[0117] In the formula, I(·) is the indicator function, which is 1 when exists, otherwise 0.

[0118] ​The parallel pollutant index model includes a CO model, a CO2 model, and a NOx model. The CO, CO2, and NOx models are all constructed using a linear regression decision tree algorithm based on Tikhonov regularization. The furnace temperature model, boiler steam flow model, and flue gas oxygen content model are connected to the CO, CO2, and NOx models, respectively. These models are then connected to a multi-objective PSO-based controlled variable optimization solution model. The CO, CO2, and NOx models can be represented as f... CO (·) and f NOx (·),as follows:

[0119]

[0120]

[0121]

[0122] In step 302, a single-neuron adaptive PID multi-input multi-output loop control model is established, specifically as follows:

[0123] Based on the experience of domain experts and the control characteristics analysis above, a single-neuron adaptive PID multi-input multi-output loop controller is constructed. The single-neuron adaptive PID multi-input multi-output loop controller includes a furnace temperature controller with feeder speed, dryer grate speed and primary air volume as operating variables, a boiler steam flow controller with feedwater volume as operating variable, and a flue gas oxygen content controller with secondary air volume as operating variable. Among them, the furnace temperature controller, boiler steam flow controller and flue gas oxygen content controller are all implemented by the single neuron adaptive PID (SNA-PID) algorithm. The solution model based on the controlled variable optimization setting of multi-objective PSO is connected to the furnace temperature controller, boiler steam flow controller and flue gas oxygen content controller respectively. The furnace temperature controller is connected to the furnace temperature model, the boiler steam flow controller is connected to the boiler steam flow model, and the flue gas oxygen content controller is connected to the flue gas oxygen content model.

[0124] The j-th operation variable of FT Taking an example (j=1,2,3) as an example, the structure of the SNA-PID controller is as follows: Figure 5 As shown, an explanation will be provided;

[0125] First, the error signal undergoes a state transition, which can be represented as:

[0126]

[0127]

[0128] where ε(k) represents the error value; and respectively represent the furnace temperature set value and the feedback value of the controlled object; k represents the iteration number;

[0129] Therefore, the SNA-PID controller output for the jth operating variable can be represented as:

[0130]

[0131] where represents the increment of the jth operating variable; represents the neuron gain coefficient of the jth operating variable; θ i j (k) represents the i th neuron weighting coefficient of the jth operating variable at the k th moment, which is represented as:

[0132]

[0133] The SNA-PID controller can realize the set value tracking under the supervised Hebb learning algorithm by continuously adjusting the neuron weighting coefficient online, as follows:

[0134]

[0135] where and respectively represent the weight coefficients of the proportional, integral, and differential neurons in the jth operating variable, and respectively represent the learning coefficients of the proportional, integral, and differential neurons in the jth operating variable.

[0136] In step 303, a controlled variable optimization set solving model based on multi-objective PSO is established, specifically:

[0137] As can be seen from the foregoing analysis, CO, CO2, and NOx are all pollutants whose emission concentrations need to be minimized. The sum of the output values of the CO model, the CO2 model, and the NOx model, i.e., the emission values of CO, CO2, and NOx, is taken as the comprehensive pollutant emission concentration that needs to be minimized, and the MSW combustion efficiency is maximized to establish a controlled variable optimization set solving model based on multi-objective PSO as follows:

[0138]

[0139] s.t. 900℃≤r FT ≤950℃

[0140] 76t / h≤r BSF ≤78t / h

[0141] 6.5% < r < 7.5% OX ≤8.5% (20)

[0142] wherein, and denote the CO, CO2 and NOx index model outputs, respectively, and r FT , r BSF and r OX denote the set points of the furnace temperature, the boiler steam flow and the flue gas oxygen content, respectively, in a multi-loop control system, the optimized set points of the furnace temperature, the boiler steam flow and the flue gas oxygen content being solved based on an improved multi-objective particle swarm optimization algorithm, wherein the improved multi-objective particle swarm optimization algorithm comprises population initialization, updating of particle individual optimum, global optimum and archive, updating of particle speed and position, mutation of population, detection of boundary limit, calculation of fitness value, judgment of termination condition and judgment of output of optimal set point according to expert experience;

[0143] The improved multi-objective particle swarm optimization algorithm is described in detail as follows:

[0144] In the population initialization, firstly, the related parameters such as the number of particles N Pos , the number of archive particles , the number of iterations N iter , the speed updating weight ω, the number of grids N grid , the mutation rate and the upper and lower limits of the set point decision variables are set; then, the population is randomly generated based on the set parameters, and the particle fitness value is calculated; then, the global optimum individual is determined based on the grid method according to the particle fitness value; finally, the above population is stored in the external archive A;

[0145] After the population initialization is completed, the algorithm enters the iteration optimization stage. Firstly, the particle speed and position are updated according to the individual optimum and the global optimum after the population initialization, and the formula is as follows:

[0146]

[0147] wherein, and are the speed and position of the d-th dimension of the i-th particle; c1 and c2 are learning factors, generally set as c1 = c2 = 2; r1 and r2 are random numbers between 0 and 1; and are the positions of the d-th dimension of the individual optimum and the global optimum of the i-th particle;

[0148] Next, in order to ensure that the algorithm has stronger global search ability in the optimization process and avoid falling into local optimum, the MOPSO algorithm introduces mutation operation in the iteration process: 1) according to the total number of population particles, the particles are divided into three parts; 2) the number of required mutation particles is determined according to the mutation rate, wherein the first part of particles does not take mutation operation, the mutation rate of the second part of particles is fixed, and the mutation rate of the third part of particles is calculated as follows:

[0149]

[0150] In the formula, is the mutation rate of the third part of particles; n iter and N iter respectively represent the current iteration number and the maximum iteration number; D represents the particle dimension. 3) After determining the number of required mutation particles of each part of population particles, random particles in the corresponding part are selected for mutation operation, as follows:

[0151]

[0152] In the formula, v and x respectively represent particle speed information and position information; v max and v min respectively represent the upper and lower limits of the particle speed; x max and x min respectively represent the upper and lower limits of the particle position; μ mut represents the particle mutation rate, φ mut (·) is the process of randomly selecting particles for mutation operation;

[0153] Then, the position information of the population after updating and mutation is detected for boundary limitation, and the limitation of the decision variable in the application is as follows:

[0154] 900℃≤r FT ≤950℃

[0155] 76t / h≤r BSF ≤78t / h

[0156] 6.5%≤r OX ≤8.5% (24)

[0157] In order to avoid particles exceeding the boundary from again exceeding the boundary in the next iteration process, the speed information of the particle is inverted to change the flight direction of the particle so as to make it away from the boundary, thereby ensuring effective iteration;

[0158] Next, the fitness values of the particles in the current population are recalculated, and the external archive A is updated and maintained. The application deletes the dominated solutions to ensure the uniqueness of the external archive A by judging the dominated relationship of the particles in the external archive A, and the dominated relationship is judged as follows:

[0159]

[0160] As can be seen from the above formula, when the combustion efficiency of the i-th particle is greater than that of the j-th particle and the overall pollutant emission concentration is less than that, it means that the j-th particle is dominated by the i-th particle, so the j-th particle in the external file is deleted.

[0161] Meanwhile, to improve the optimization speed of the MOPSO algorithm, an upper limit needs to be set for the total number of particles in external file A. If the number of particles in external file A exceeds the limit Then: 1) Determine the number of particles N to be deleted. Del 2) Sort the particles in external file A according to their fitness values ​​and calculate the crowding distance using the following formula:

[0162]

[0163] In the formula, Ω represents the crowding distance of particles in the external archive; i Indicates the first i 3) Delete N particles based on crowding distance. Del One particle;

[0164] Then, based on external file A, the individual optimal and global optimal of the particles are updated. The individual optimal is updated as follows: determine the dominance relationship between the current particle and the individual optimal particle; if the current particle dominates the individual optimal, then update the individual optimal; if there is no dominance relationship between the two, then randomly determine whether to update the individual optimal.

[0165] The global optimum update method based on the grid method is as follows: 1) Divide the minimum and maximum fitness values ​​of particles in external file A into N equal parts for each target. grid , forming N grid ×N grid 1) Determine the mesh location and the position where particle A from the external file falls into the mesh; 2) Evaluate the mesh quality according to the following formula:

[0166]

[0167] In the formula, Q grid (z) represents the quality of the z-th grid, where z = 1, 2, ..., N grid ×N grid num(z) represents the number of particles in the z-th grid. 3) Select the particle with the higher grid quality as the global optimum. If multiple particles exist in the same grid or multiple grids have the same quality, randomly select the particle that meets the conditions as the global optimum.

[0168] Further, by continuously repeating the above steps, the present application introduces the termination condition of the difference of the comprehensive pollutant emission concentration as the target on the basis of the MOPSO algorithm, when the absolute value of the difference between the mean value of the comprehensive pollutant emission concentration of the particles stored in the external file and the mean value of the comprehensive pollutant emission concentration of the particles in the external file is less than the set value e stop , the algorithm stops iteration, and the Pareto frontier is output;

[0169] Finally, based on the Pareto frontier and expert experience, the set value meeting the requirements of economic indicators, safety performance and other related requirements in the actual industrial field is selected as the optimal set value as the multi-input multi-output loop control set value. The present application only gives the basic expert experience as follows:

[0170] Rule 1: When the CO concentration increases, the Pareto solution with higher combustion efficiency is preferred

[0171] Rule 2: When the NOx concentration increases, the Pareto solution with smaller pollutant emission index is preferred

[0172] Rule 3: When the CO2 concentration increases, the MSW combustion efficiency is improved, and in the current case, any solution in the Pareto solution set can be selected according to the current incineration demand.

[0173] The present application provides an embodiment, which specifically adopts the process data of 8 hours of continuous operation from 16:00 to 24:00 on a certain day in a certain month in 2021 of a certain power plant to verify the accuracy of the multi-objective optimization control method. In order to verify the effectiveness of the proposed model, CART tree, RF and BPNN are compared with TR-LRDT algorithm. The root mean square error (RMSE) and the mean absolute error (MAE) are used to evaluate the performance of the built model, as follows:

[0174]

[0175]

[0176] In the formula, y i and are the model prediction value, the true value and the mean value respectively, and N is the sample number. The model of the series connection controlled object is verified, and the model parameter settings are shown in Table 1;

[0177] Table 1 Model parameter settings

[0178]

[0179] In the formula, θ TR-LRDT , θ CART and θ RF, respectively, represent the minimum sample number in TR-LRDT, CART and RF algorithms; λ represents the regularization term coefficient; Tn represents the number of decision trees in RF algorithm; Hidden, epoch, error and η represent the number of hidden layer neurons, iteration number, convergence error and learning rate in BPNN algorithm, respectively, the fitting curves of the test set of each model are shown in Figures 6a-6c Table 2 shows the performance evaluation comparison results.

[0180] Table 2 shows the performance evaluation comparison results.

[0181]

[0182]

[0183] The model of the pollutant index is verified, and the model parameter settings are shown in Table 3.

[0184] Table 3 shows the model parameter settings.

[0185]

[0186] In the formula, θ TR-LRDT , θ CART and θ RF respectively represent the minimum sample number in TR-LRDT, CART and RF algorithms; λ represents the regularization term coefficient; Tn represents the number of decision trees in RF algorithm; Hidden, epoch, error and η represent the number of hidden layer neurons, iteration number, convergence error and learning rate in BPNN algorithm, respectively, the fitting curves of the test set of each model are shown in Figures 7a-7c Table 4 shows the performance evaluation comparison results.

[0187] Table 4 shows the performance evaluation comparison results.

[0188]

[0189]

[0190] From the modeling comparison results, the TR-LRDT algorithm has better fitting effect and superior modeling accuracy compared with other algorithms, because the TR-LRDT algorithm introduces weights in the tree structure model, corrects the defects of the mean output of the leaf node of the tree structure model, and thus enhances the smoothing performance of the model.

[0191] The SNA-PID controller related parameter settings are shown in Table 5, and the PID controller related parameters are shown in Table 6.

[0192] Table 5 shows the SNA-PID controller parameter settings.

[0193]

[0194] Table 6 PID controller parameter settings

[0195]

[0196] The controller tracking results and error curves are shown in Figs. 3 and 4, respectively. Figures 8a-8c Figures 9a-9c As can be seen from Figs. 3 and 4, in the case of varying set value, both controllers can achieve effective tracking of the set value, but the SNA-PID controller has a faster adjustment time than the PID controller. Figures 8a-8c Figures 9a-9c

[0197] The controller performance is evaluated by integral square error (ISE), integral absolute error (IAE) and maximum deviation (DEV max ), and the calculation formulas are as follows:

[0198]

[0199]

[0200] DEV max = max{|e(t)|} (32)

[0201] In the formulas, t0 and t f represent the control start and end times; the index variation curves are shown in Figs. 5 and 6, and the statistical results are shown in Table 7. Figures 10a-10c Figures 11a-11c

[0202] Table 7 Controller index statistical results

[0203]

[0204] As can be seen from the index variation curves and statistical results, the SNA-PID controller has a lower performance in the DEV max index, but has a higher precision in the IAE and ISE indexes. The reason is that the SNA-PID controller continuously adjusts its output by error and controller parameters, thereby obtaining a faster adjustment speed; when facing the situation of set value variation, since the error is large at this time, the SNA-PID controller makes the controller output amplitude large without changing the parameters, thereby causing the DEV max index to perform worse than the PID controller;

[0205] ​​​​​The results of multi-objective optimization are as follows: firstly, the emission concentration of pollutants is normalized and summed as the comprehensive pollutant emission concentration, then the IMOPSO algorithm is used to solve the set value of the controlled variable optimization, and the related parameters are as follows: the population number is 200, the upper limit of the external archive particle is 200, the iteration number is 500, the inertia factor is 0.9, the learning factor c1=c2=2, the grid number is 20, the mutation rate is 0.5, the maximum particle speed is [1, 0.1, 0.1], the iteration termination condition e stop is 0.0001, and the Pareto front is shown in Figure 12a and Figure 12b It can be seen that 1) for the Pareto front, the solution results of the two are similar, indicating that the IMOPSO algorithm can solve the multi-objective optimization problem of the MSWI process; 2) for the population distribution, compared with the IMOPSO algorithm, the population of the MOPSO algorithm is more concentrated near the Pareto front after iteration, because with the increase of the iteration number, the population gradually tends to the global optimum, which further proves the effectiveness of the Pareto front obtained by the two algorithms. However, for optimization problems, researchers are more concerned about the Pareo front, therefore, the IMOPSO algorithm can accurately solve the Pareto front while reducing the iteration time, and has stronger engineering application significance.

[0206] The data-driven multi-objective optimization control method for municipal solid waste incineration process provided by the application is based on the analysis of the MSWI process for multi-objective optimization, a multi-objective optimization model is established with the key controlled variables as the decision variables, the minimum comprehensive pollutant emission concentration and the maximum MSW combustion efficiency, and a corresponding multi-objective optimization control strategy is proposed, specifically, firstly, a full-process model of the MSWI process for pollutant emission is established, including the controlled object models of furnace temperature, boiler steam flow and flue gas oxygen content, and the pollutant emission index models of CO, CO2 and NOx; then, a single neuron adaptive PID is used to establish a multi-input and multi-output loop controller for the key controlled variables, to realize stable control of the combustion process; finally, based on the field data of a certain MSWI power plant, the adaptive solution of the optimized set value of the key controlled variables based on the multi-objective particle swarm optimization algorithm is realized, and the method is verified to be able to obtain the optimized set value of the three key controlled variables, reduce the pollutant emission concentration and improve the combustion efficiency.

[0207] The principles and implementation manners of the present application are described by using specific examples in the present application, and the above examples are only used for helping to understand the method of the present application and its core idea; meanwhile, for the general technical personnel in the art, according to the idea of the present application, the specific implementation manners and application ranges will be changed. In conclusion, the content of the present specification should not be understood as the limitation of the present application.

Claims

1. A data-driven multi-objective optimization control method for urban solid waste incineration processes, characterized in that, Includes the following steps: Step 1: Analyze the influencing factors of multi-objective optimization; Step 2: Detect multi-objective optimization control model based on influencing factors; Step 2 involves detecting influencing factors and establishing a multi-objective optimization control model, specifically including the following steps: Step 301: Establish a full-process model for pollutant emission reduction as the controlled object model; Step 302: Establish a single-neuron adaptive PID multi-input multi-output loop controller to perform stable control on the controlled object model; Step 303: Establish a control variable optimization setting solution model based on multi-objective PSO to achieve adaptive solution of the optimization setting value of key control variables, and send the solution results to the multi-input multi-output loop controller of single neuron adaptive PID to perform stable control of the controlled object model; In step 301, a full-process model for pollutant emission reduction is established, specifically as follows: A full-process model for pollutant emission reduction is established, including a series controlled object model and a parallel pollutant index model. A single-neuron adaptive PID multi-input multi-output loop controller is connected to the series controlled object model, the series controlled object model is connected to the parallel pollutant index model, and the parallel pollutant index model is connected to the controlled variable optimization setting solution model based on multi-objective PSO. The series-controlled object model includes a furnace temperature model, a boiler steam flow model, and a flue gas oxygen content model. These models are all constructed using a Tikhonov-regularized linear regression decision tree algorithm. A single-neuron adaptive PID multi-input multi-output loop controller connects these models, providing input to the furnace temperature model. , and Input the boiler steam flow model Input the oxygen content of the flue gas into the model The furnace temperature model is connected to the boiler steam flow model and the flue gas oxygen content model. The furnace temperature model, boiler steam flow model, and flue gas oxygen content model are connected to a parallel pollutant index model. , , , and These represent the output of water supply, primary air volume, feeder speed, drying grate speed, and secondary air volume, respectively. The parallel pollutant index models include a CO model, a CO2 model, and a NOx model. The CO, CO2, and NOx models are all constructed using a linear regression decision tree algorithm based on Tikhonov regularization. The furnace temperature model, boiler steam flow model, and flue gas oxygen content model are connected to the CO, CO2, and NOx models, respectively. The CO, CO2, and NOx models are connected to a solution model based on the optimization setting of controlled variables for multi-objective PSO.

2. The data-driven multi-objective optimization control method for urban solid waste incineration process according to claim 1, characterized in that, Step 1 analyzes the influencing factors for multi-objective optimization, specifically: The constraints of MSWI are: （1）； In the formula, t is the time variable, and x is the set value of the process operation variable; To ensure that environmental protection standards are met, the following measures are taken: （2）； To improve product specifications, the following measures are taken: （3）； To improve economic indicators, the following measures are taken: （4）； To align the optimization objectives, the optimization objective of MSWI is represented as follows: （5）； By solving for x, the MSWI process can be optimized, where x includes furnace temperature, boiler steam flow rate, and flue gas oxygen content.

3. The data-driven multi-objective optimization control method for urban solid waste incineration process according to claim 1, characterized in that, In step 302, a single-neuron adaptive PID multiple-input multiple-output loop controller is established, specifically as follows: The single-neuron adaptive PID multi-input multi-output loop controller includes a furnace temperature controller with feeder speed, dryer grate speed, and primary air volume as operating variables, a boiler steam flow controller with feedwater flow as operating variable, and a flue gas oxygen content controller with secondary air volume as operating variable. The furnace temperature controller, boiler steam flow controller, and flue gas oxygen content controller are all implemented using a single-neuron adaptive PID algorithm. The controlled variable optimization setting solution model based on multi-objective PSO is connected to the furnace temperature controller, boiler steam flow controller, and flue gas oxygen content controller, respectively. The furnace temperature controller is connected to the furnace temperature model, the boiler steam flow controller is connected to the boiler steam flow model, and the flue gas oxygen content controller is connected to the flue gas oxygen content model.

4. The data-driven multi-objective optimization control method for urban solid waste incineration process according to claim 3, characterized in that, In step 303, a solution model for the controlled variable optimization based on multi-objective PSO is established, specifically as follows: The output values ​​of the CO model, CO2 model, and NOx model, i.e., the sum of the emissions of CO, CO2, and NOx, are used as the comprehensive pollutant emission concentration to be minimized, while maximizing the MSW combustion efficiency. The controlled variable optimization model based on multi-objective PSO is then established as follows: （6）； In the formula, , and These represent the model outputs for CO, CO2, and NOx indicators, respectively. , and These represent the setpoints for furnace temperature, boiler steam flow rate, and flue gas oxygen content in a multi-loop control system. The controlled variables, namely the optimized setpoints for furnace temperature, boiler steam flow rate, and flue gas oxygen content, are optimized based on an improved multi-objective particle swarm optimization algorithm. The improved multi-objective particle swarm optimization algorithm includes population initialization, updating individual particle optimality, global optimality, and profiles, updating particle velocity and position, mutating the population, detecting boundary constraints, calculating fitness values, determining termination conditions, and outputting the optimal setpoints based on expert experience.

Images