Intelligent optimized operation method for urban solid waste incineration process

Abstract

The invention provides an intelligent optimization operation method for an urban solid waste incineration process, and relates to the technical field of urban solid waste treatment and intelligent optimization. The method comprises the following steps: acquiring environmental data in the incinerator and concentration values of various incineration exhaust gases before a preset duration; iteratively optimizing the solution of the decision variable so that the value of the objective function obtained according to the solution of the decision variable is optimized towards a decreasing direction; the solution of the target decision variable is determined, and the set values of the flue gas oxygen content and the hearth temperature in the current urban solid waste incineration process are obtained; the step of calculating the value of the objective function comprises the substeps of inputting the solution of the decision variable, the current environmental data in the incinerator and the concentration values of various incineration exhaust gases before a preset time length into a pre-trained operation index model, outputting the predicted concentration values of the various incineration exhaust gases, calculating the combustion efficiency and the flue gas pollutant concentration, and calculating the value of the objective function according to the combustion efficiency and the flue gas pollutant concentration. And obtaining the value of the target function. And emission of flue gas pollutants is reduced while the combustion efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of urban solid waste treatment and intelligent optimization technology, and particularly relates to an intelligent optimization operation method for the urban solid waste incineration process. Background Art

[0002] Driven by the acceleration of urbanization, urban solid waste is experiencing rapid growth. The incineration method has now become the mainstream way of solid waste treatment. During the incineration process, the combustion efficiency reflects the adequacy of solid waste incineration, and the pollutants emitted during the incineration process will cause great harm to the environment and the lives of people in the surrounding areas. Therefore, optimizing the urban solid waste incineration process to reduce the emissions of flue gas pollutants while improving the combustion efficiency has important theoretical significance and application value. Summary of the Invention

[0003] The present invention provides an intelligent optimization operation method for the urban solid waste incineration process, which is used to solve the defect in the prior art that it is difficult to reduce the emissions of flue gas pollutants while improving the combustion efficiency, and realizes reducing the emissions of flue gas pollutants while improving the combustion efficiency.

[0004] The present invention provides an intelligent optimization operation method for the urban solid waste incineration process, including the following steps.

[0005] During the urban solid waste incineration process, obtain the environmental data in the current incinerator and the concentration values of various incineration exhaust gases before a preset time period; Iteratively optimize the solution of the decision variable so that the value of the objective function obtained according to the solution of the decision variable is optimized in a decreasing direction; the decision variables include the oxygen content in the flue gas and the furnace temperature; Determine the solution of the target decision variable from the solutions of the groups of decision variables obtained in the last round of iteration, and determine the values of the oxygen content in the flue gas and the furnace temperature included in the solution of the target decision variable as the set values of the oxygen content in the flue gas and the furnace temperature in the current urban solid waste incineration process; Wherein, the calculation steps of the value of the objective function include: inputting the solution of the decision variable, the environmental data in the current incinerator, and the concentration values of various incineration exhaust gases before the preset time period into the operation index models of various incineration exhaust gases that have been pre-trained, outputting the predicted concentration values of the current various incineration exhaust gases, calculating the combustion efficiency and the flue gas pollutant concentration according to the predicted concentration values of the various incineration exhaust gases, and calculating the value of the objective function according to the combustion efficiency and the flue gas pollutant concentration.

[0006] According to the intelligent optimization operation method for the urban solid waste incineration process provided by the present invention, the training steps of the pre-trained operation index models of various incineration exhaust gases include: Construct a sample data set based on the historical data of the urban solid waste incineration process; the sample data set includes the sample oxygen content in the flue gas, the sample furnace temperature, the sample environmental data in the incinerator, and the sample concentration values of various incineration emission gases during the historical process of urban solid waste incineration. Iteratively input the sample oxygen content in the flue gas, the sample furnace temperature, the sample environmental data in the incinerator, and the sample concentration values of various incineration emission gases before the preset time duration at the moment into the operation index models of various incineration emission gases to be trained, and output the sample predicted concentration values of various incineration emission gases at the moment. According to the difference between the sample predicted concentration values of various incineration emission gases and the true sample concentration values at the moment in the sample data set, adjust the parameters of the operation index models of various incineration emission gases to be trained and enter the next round of iteration until the iteration stops, and obtain the trained operation index models of various incineration emission gases.

[0007] According to an intelligent optimization operation method for the urban solid waste incineration process provided by the present invention, the iteratively inputting the sample oxygen content in the flue gas, the sample furnace temperature, the sample environmental data in the incinerator, and the sample concentration values of various incineration emission gases before the preset time duration at the moment into the operation index models of various incineration emission gases to be trained, and outputting the sample predicted concentration values of various incineration emission gases at the moment includes: Iteratively use the sample oxygen content in the flue gas, the sample furnace temperature, the sample environmental data in the incinerator, and the sample concentration values of various incineration emission gases before the preset time duration at the moment as input variables, and sequentially input them into the input layers of each sub-module of the operation index models of various incineration emission gases to be trained; wherein, each sub-module is respectively used to output the sample predicted concentration value of an incineration emission gas. In each sub-module, perform normalization processing on the input variables through the input layer, input the normalized input variables into the hidden layer, perform spatial mapping transformation on the normalized input variables through the hidden layer, input the spatially mapped transformed input variables into the output layer, and perform weighted summation on the spatially mapped transformed input variables through the output layer to output the sample predicted concentration value of the incineration emission gas corresponding to the sub-module.

[0008] According to an intelligent optimization operation method for the urban solid waste incineration process provided by the present invention, the incineration emission gases include nitrogen oxides, carbon monoxide, carbon dioxide, sulfur dioxide, and hydrogen chloride. Calculating the combustion efficiency and the concentration of flue gas pollutants according to the predicted concentration values of various incineration emission gases includes: Calculate the combustion efficiency according to the predicted concentration values of the carbon dioxide and the carbon monoxide; Calculate the flue gas pollutant concentration according to the predicted concentration values of the nitrogen oxides, the carbon monoxide, the sulfur dioxide and the hydrogen chloride.

[0009] According to an intelligent optimization operation method for the urban solid waste incineration process provided by the present invention, iteratively optimizing the solution of the decision variable so that the value of the objective function obtained according to the solution of the decision variable is optimized in a decreasing direction, includes: Initialize the decision variable population as the original decision variable population in the first round of iteration; In each round of iteration, perform a mutation operation on each individual in the original decision variable population in this round of iteration to obtain a mutated population; Perform a crossover operation on the original decision variable population and the mutated population in this round of iteration to obtain a trial population; Merge the trial population and the original decision variable population to obtain a mixed population; Determine the original decision variable population in the next round of iteration according to the values of the objective functions respectively obtained from the solutions of the decision variables in the mixed population, and enter the next round of iteration.

[0010] According to an intelligent optimization operation method for the urban solid waste incineration process provided by the present invention, performing a mutation operation on each individual in the original decision variable population in this round of iteration to obtain a mutated population, includes: If the current iteration number is less than or equal to the target iteration number, perform a mutation operation on each individual in the original decision variable population in this round of iteration by using a single mutation strategy to obtain a mutated population; the target iteration number is obtained according to a preset ratio of the maximum iteration number; If the current iteration number is greater than the target iteration number, allocate multiple mutation strategies to each individual in the original decision variable population in this round of iteration according to their respective target ratios, and perform a mutation operation on the corresponding individual by using the allocated mutation strategy to obtain multiple mutated populations; Wherein, the target ratio is determined according to the performance of each mutation strategy in the previous round of iteration.

[0011] According to an intelligent optimization operation method for the urban solid waste incineration process provided by the present invention, the determining step of the target ratio includes: Determine the first fitness of the solutions of the decision variables in the mutated populations respectively obtained by each mutation strategy in the previous round of iteration, and the second fitness of the solutions of the decision variables in the original decision variable population in the previous round of iteration; the fitness is determined according to the value of the objective function; Determine the performance corresponding to each mutation strategy according to the distance between the first fitness corresponding to each mutation strategy and the second fitness respectively. Determine the target ratio corresponding to each mutation strategy according to the ratio of the performance corresponding to each mutation strategy to the total performance of all mutation strategies.

[0012] According to an intelligent optimization operation method for the urban solid waste incineration process provided by the present invention, determining the original population of decision variables in the next iteration according to the values of the objective function respectively obtained from the solutions of each decision variable in the mixed population includes: Divide the mixed population into multiple clusters; Determine the non-dominated solutions and dominated solutions in each cluster according to the values of the objective function corresponding to the solutions of the decision variables in each cluster, divide the non-dominated solutions into the elite archive, and divide the dominated solutions into the non-elite archive; Determine the original population of decision variables in the next iteration according to the elite archive.

[0013] According to an intelligent optimization operation method for the urban solid waste incineration process provided by the present invention, determining the original population of decision variables in the next iteration according to the elite archive includes: If the number of solutions of the decision variables in the elite archive does not reach the preset scale, quantify the distribution density of the solutions of the decision variables in the non-elite archive in the decision space, select the solutions of the decision variables in the sparse area according to the distribution density and supplement them to the elite archive, and select multiple solutions of the decision variables from the supplemented elite archive to form the original population of decision variables in the next iteration; If the number of solutions of the decision variables in the elite archive reaches the preset scale, directly select multiple solutions of the decision variables from the elite archive according to the values of the objective function corresponding to the solutions of each decision variable in the elite archive to form the original population of decision variables in the next iteration.

[0014] According to an intelligent optimization operation method for the urban solid waste incineration process provided by the present invention, determining the solution of the target decision variable from the solutions of each group of decision variables obtained in the last iteration includes: For each group of solutions of the decision variables obtained in the last iteration respectively, perform weighted summation on the first optimization objective value and the second optimization objective value corresponding to the solution of the decision variable according to the first preset weight and the second preset weight respectively, to obtain the weighted summation result corresponding to the solution of the decision variable; the first optimization objective value is determined according to the combustion efficiency corresponding to the solution of the decision variable; the second optimization objective value is determined according to the flue gas pollutant concentration corresponding to the solution of the decision variable; Determine the solution of the decision variable with the smallest weighted summation result as the solution of the target decision variable.

[0015] The intelligent optimization operation method for the urban solid waste incineration process provided by the present invention determines the predicted concentration values of various incineration emission gases through pre-trained operation index models of various incineration emission gases, calculates the combustion efficiency and flue gas pollutant concentrations based on the predicted concentration values of various incineration emission gases, calculates the value of the objective function based on the combustion efficiency and flue gas pollutant concentrations, and iteratively optimizes the solution of the decision variable so that the value of the objective function obtained according to the solution of the decision variable is optimized in the decreasing direction, thereby being able to accurately determine the set values of the flue gas oxygen content and the furnace temperature, and reducing the emission of flue gas pollutants while improving the combustion efficiency. Brief Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 is one of the flow schematic diagrams of the intelligent optimization operation method for the urban solid waste incineration process provided by the present invention.

[0018] Figure 2 is the structural schematic diagram of the operation index model of various incineration emission gases provided by the present invention.

[0019] Figure 3 is the schematic diagram of the test results of the operation index model of nitrogen oxides provided by the present invention.

[0020] Figure 4 is the schematic diagram of the test results of the operation index model of carbon monoxide provided by the present invention.

[0021] Figure 5 is the schematic diagram of the test results of the operation index model of carbon dioxide provided by the present invention.

[0022] Figure 6 is the schematic diagram of the test results of the operation index model of sulfur dioxide provided by the present invention.

[0023] Figure 7 is the schematic diagram of the test results of the operation index model of hydrogen chloride provided by the present invention.

[0024] Figure 8 is the flow schematic diagram of iteratively optimizing the solution of the decision variable provided by the present invention.

[0025] Figure 9It is a schematic diagram of the optimization result of the combustion efficiency provided by the present invention.

[0026] Figure 10 It is a schematic diagram of the optimization result of the concentration of flue gas pollutants provided by the present invention.

[0027] Figure 11 It is the second schematic diagram of the flow of the intelligent optimization operation method for the urban solid waste incineration process provided by the present invention. Detailed implementation manners

[0028] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0029] The following combines Figures 1 - 11 to describe the intelligent optimization operation method for the urban solid waste incineration process of the present invention.

[0030] Figure 1 It is the first schematic diagram of the flow of the intelligent optimization operation method for the urban solid waste incineration process provided by the present invention. As Figure 1 shown, the method includes the following: Step 102, during the urban solid waste incineration process, obtain the environmental data in the current incinerator and the concentration values of various incineration exhaust gases before a preset time period.

[0031] In one embodiment, the environmental data in the incinerator includes at least one of the primary air temperature, primary air pressure, secondary air temperature, and chimney flue gas temperature.

[0032] In one embodiment, the incineration exhaust gases include nitrogen oxides, carbon monoxide, carbon dioxide, sulfur dioxide, and hydrogen chloride.

[0033] In one embodiment, the preset time period can be set according to actual situations. For example: the preset time period can be 2 moments, that is, obtain the environmental data in the current (t moment) incinerator and the concentration values of various incineration exhaust gases 2 moments before (t - 2 moment).

[0034] In one embodiment, during the urban solid waste incineration process, steps 102 to 106 can be executed at every first preset time interval to obtain the set values of the oxygen content in the flue gas and the furnace temperature during the current urban solid waste incineration process, and set the current oxygen content in the flue gas and the furnace temperature. For example: every 5 minutes, execute steps 102 to 106 to obtain new set values of the oxygen content in the flue gas and the furnace temperature.

[0035] Step 104: Iteratively optimize the solution of the decision variables so that the value of the objective function obtained according to the solution of the decision variables is optimized in the decreasing direction; the decision variables include the oxygen content in the flue gas and the furnace temperature.

[0036] Among them, the calculation steps of the value of the objective function in Step 102 include: inputting the solution of the decision variables, the environmental data in the current incinerator, and the concentration values of various incineration emission gases before a preset time period into the pre-trained operation index models of various incineration emission gases, outputting the predicted concentration values of various current incineration emission gases, calculating the combustion efficiency and the flue gas pollutant concentration according to the predicted concentration values of various incineration emission gases, and calculating the value of the objective function according to the combustion efficiency and the flue gas pollutant concentration.

[0037] Among them, the pre-trained operation index models of various incineration emission gases include multiple sub-modules, and each sub-module is respectively used to output the predicted concentration value of an incineration emission gas. Specifically, each sub-module respectively outputs the predicted concentration value of the corresponding type of incineration emission gas according to the input solution of the decision variables, the environmental data in the current incinerator, and the concentration values of various incineration emission gases before a preset time period.

[0038] In one embodiment, the operation index models of various incineration emission gases can be composed of multiple independent RBF (Radial Basis Function) neural network sub-models.

[0039] It can be understood that since it is necessary to satisfy the maximum combustion of solid waste and at the same time reduce the concentration of flue gas pollutants emitted as much as possible, the collaborative optimization of the combustion efficiency and the flue gas pollutant concentration through the objective function belongs to a multi-objective optimization problem.

[0040] In one embodiment, the objective function may include the negative value of the combustion efficiency and the flue gas pollutant concentration. The objective function at time t can be expressed as . Among them, represents the combustion efficiency at time t. represents the flue gas pollutant concentration at time t.

[0041] Step 106: Determine the solution of the target decision variable from the solutions of the groups of decision variables obtained in the last round of iteration, and determine the values of the oxygen content in the flue gas and the furnace temperature included in the solution of the target decision variable as the set values of the oxygen content in the flue gas and the furnace temperature in the current municipal solid waste incineration process.

[0042] In one embodiment, when the preset maximum number of iterations is reached, output the set of solutions of the decision variables obtained in the last round of iteration, and determine the solution of the target decision variable from the set obtained in the last round of iteration.

[0043] The above intelligent optimization operation method for the urban solid waste incineration process determines the predicted concentration values of various incineration emission gases through pre-trained operation index models for various incineration emission gases. This model effectively solves the problems of complex mechanism and dynamic time-variation in the urban solid waste incineration process. Then, the combustion efficiency and flue gas pollutant concentration are calculated based on the predicted concentration values of various incineration emission gases, and the value of the objective function is calculated based on the combustion efficiency and flue gas pollutant concentration. The solution of the decision variable is iteratively optimized so that the value of the objective function obtained according to the solution of the decision variable is optimized in a decreasing direction, thereby accurately determining the set values of the flue gas oxygen content and furnace temperature, and reducing the emission of flue gas pollutants while improving the combustion efficiency. The method has high adaptability and effectively solves the problems of mutual conflict between optimization objectives and slow calculation process.

[0044] In one embodiment, the training steps of the pre-trained operation index models for various incineration emission gases include: constructing a sample data set based on the historical data of the urban solid waste incineration process; the sample data set includes the sample flue gas oxygen content, sample furnace temperature, sample environmental data in the incinerator, and sample concentration values of various incineration emission gases in the historical process of urban solid waste incineration; iteratively inputting the sample flue gas oxygen content, sample furnace temperature, sample environmental data in the incinerator, and sample concentration values of various incineration emission gases before the preset time period at the moment into the operation index models for various incineration emission gases to be trained, and outputting the sample predicted concentration values of various incineration emission gases at the moment; adjusting the parameters of the operation index models for various incineration emission gases to be trained according to the difference between the sample predicted concentration values of various incineration emission gases and the true sample concentration values at the moment in the sample data set, and entering the next round of iteration until the iteration stops, obtaining the trained operation index models for various incineration emission gases.

[0045] In one embodiment, data can be collected at the second preset time interval to obtain the sample flue gas oxygen content, sample furnace temperature, sample environmental data in the incinerator, and sample concentration values of various incineration emission gases at multiple moments in the historical process of urban solid waste incineration. For example, data can be collected every 30 seconds.

[0046] In the above embodiment, based on the historical data of the urban solid waste incineration process, a sample data set is constructed. The sample flue gas oxygen content, sample furnace temperature, sample environmental data in the incinerator, and sample concentration values of various incineration emission gases before the preset time period at the moment are iteratively input into the operation index models for various incineration emission gases to be trained, and the model training is iteratively performed according to the difference between the output sample predicted concentration values and the true sample concentration values in the sample data set, and an accurate operation index model for various incineration emission gases can be obtained. This model effectively solves the problems of complex mechanism and dynamic time-variation in the urban solid waste incineration process.

[0047] In one embodiment, the sample oxygen content of flue gas, the sample furnace temperature, the sample ambient data in the incinerator, and the sample concentration values of various incineration emission gases at a preset time duration before the time are iteratively input into the operation index models of various incineration emission gases to be trained, and the sample predicted concentration values of various incineration emission gases at the time are output, including: iteratively taking the sample oxygen content of flue gas, the sample furnace temperature, the sample ambient data in the incinerator, and the sample concentration values of various incineration emission gases at a preset time duration before the time as input variables, and sequentially inputting them into the input layers of each sub-module in the operation index models of various incineration emission gases to be trained; wherein each sub-module is respectively used to output the sample predicted concentration value of one incineration emission gas; in each sub-module, the input variables are normalized by the input layer, the normalized input variables are input into the hidden layer, the normalized input variables are subjected to spatial mapping transformation by the hidden layer, the input variables after the spatial mapping transformation are input into the output layer, and the input variables after the spatial mapping transformation are weighted and summed by the output layer to output the sample predicted concentration value of the incineration emission gas corresponding to the sub-module.

[0048] In one embodiment, the operation index models of various incineration emission gases can be composed of multiple independent RBF neural network sub-models. The topological structure of the RBF neural network model includes three layers: an input layer, a hidden layer, and an output layer. Each independent RBF neural network sub-model is respectively a sub-module in the operation index models of various incineration emission gases, and is used to output the sample predicted concentration value of one incineration emission gas. Figure 2 is a schematic structural diagram of the operation index models of various incineration emission gases.

[0049] In one embodiment, the sample ambient data in the incinerator includes at least one of the sample primary air temperature, the sample primary air pressure, the sample secondary air temperature, and the sample chimney flue gas temperature.

[0050] In one embodiment, the input variables of each sub-module include decision variables and auxiliary variables. The decision variables include the sample oxygen content of flue gas and the sample furnace temperature at the t-th moment, and the auxiliary variables include the sample primary air temperature at the t-th moment, the sample primary air pressure at the t-th moment, the sample secondary air temperature at the t-th moment, the sample chimney flue gas temperature at the t-th moment, and the sample concentration values of various incineration emission gases at the (t - τ)-th moment. Wherein, τ represents the preset time duration. The input variables can be expressed as . represents the serial number of the incineration emission gas.

[0051] In one embodiment, the normalization process of the input variables by the input layer can be expressed by the following formula: Among them, represents the input variable. represents the maximum value among the input variables. represents the minimum value among the input variables. represents the input variable after normalization processing.

[0052] In one embodiment, the number of neurons in the hidden layer depends on the need. The kernel function of the neurons in the hidden layer is a Gaussian function, which performs a spatial mapping transformation on the input information. The spatial mapping transformation of the input variable after normalization processing by the hidden layer can be expressed by the following formula: Among them, represents the input vector of the model (the input variable after normalization processing). represents the center vector of the i-th neuron in the hidden layer. represents the width value of the i-th neuron in the hidden layer. represents the output of the i-th hidden layer.

[0053] In one embodiment, the input variable after the spatial mapping transformation is weighted and summed through the output layer, and the sample predicted concentration value of the incineration emission gas corresponding to the output sub-module can be expressed by the following formula: Among them, represents the output of the sub-module (i.e., the RBF neural network). represents the input vector of the model (the input variable after normalization processing). represents the activation function of the i-th neuron in the hidden layer. represents the output of the i-th hidden layer. represents the connection weight value of the i-th neuron in the hidden layer. represents the total number of hidden layers.

[0054] In one embodiment, the data output by the output layer can be de-normalized to obtain the sample predicted concentration value of the corresponding incineration emission gas.

[0055] In one embodiment, the data from a municipal solid waste incineration plant in Beijing is used to verify the effectiveness of the method proposed in the present invention. 3000 groups of historical data are selected as the sample data set, among which the first 2000 groups of data are used as training samples, and the remaining 1000 groups of data are used as test samples. The model is trained according to the training samples to obtain the operation index models of various incineration emission gases after training. To verify the accuracy and reliability of the constructed operation index models of various incineration emission gases, an independent test sample data set is used for model verification: 1000 groups of test samples are respectively input into the sub-modules of the operation index models of various incineration emission gases, and the concentration fitting values of various incineration emission gases are output. The root mean square error (RMSE) and the coefficient of determination (R 2 ) are used to quantitatively evaluate the fitting accuracy of each sub-module. RMSE and R 2 are calculated as follows: Among them, represents the concentration fitting values of various incineration emission gases output by the operation index model of various incineration emission gases for the i-th test sample. represents the actual concentration values of various incineration emission gases in the test sample. represents the output average value of each test sample. m represents the total number of test samples. The lower the RMSE, the higher the fitting accuracy of the model. The closer R 2 is to 1, the better the fitting effect of the model.

[0056] Figure 3 , Figure 4 , Figure 5 , Figure 6 and Figure 7 respectively show the test results of the operation index models of nitrogen oxides, carbon monoxide, carbon dioxide, sulfur dioxide and hydrogen chloride. The x-axis in the figure represents the number of test samples, with the unit of number / sample, and the y-axis represents the concentration values of nitrogen oxides, carbon monoxide, carbon dioxide, sulfur dioxide and hydrogen chloride respectively, with the unit of mg / m 3 , the solid line represents the actual value in the test sample, and * represents the network prediction output value of the test sample. The root mean square error RMSE and the coefficient of determination R 2 are used to quantitatively evaluate the fitting accuracy of the operation index models of various incineration emission gases. For the operation index model of nitrogen oxides, the root mean square error RMSE obtained by testing with test samples is 15.223, and the coefficient of determination R2 is 0.9316; for the operation index model of carbon monoxide, the root mean square error RMSE obtained by testing with test samples is 0.7460, and the coefficient of determination R2 is 0.9052; for the operation index model of carbon dioxide, the root mean square error RMSE obtained by testing with test samples is 0.0975, and the coefficient of determination R2 is 0.9581; for the operation index model of sulfur dioxide, the root mean square error RMSE obtained by testing with test samples is 0.9470, and the coefficient of determination R2 is 0.8900; for the operation index model of hydrogen chloride, the root mean square error RMSE obtained by testing with test samples is 0.7689, and the coefficient of determination R2 is 0.8665.

[0057] In the above embodiments, a neural network model with multiple modules and multiple tasks is constructed for model training to obtain operation index models of various incineration emission gases. Each module can output the concentration value of one incineration emission gas respectively, so that the concentration values of various incineration emission gases can be accurately predicted through the operation index models of various incineration emission gases obtained by training. This model effectively solves the problems of complex mechanism and dynamic time-variation in the process of municipal solid waste incineration.

[0058] In one embodiment, the incineration emission gases include nitrogen oxides, carbon monoxide, carbon dioxide, sulfur dioxide and hydrogen chloride. Calculating the combustion efficiency and flue gas pollutant concentration according to the predicted concentration values of various incineration emission gases includes: calculating the combustion efficiency according to the predicted concentration values of carbon dioxide and carbon monoxide; calculating the flue gas pollutant concentration according to the predicted concentration values of nitrogen oxides, carbon monoxide, sulfur dioxide and hydrogen chloride.

[0059] In one embodiment, the predicted concentration values of carbon dioxide and carbon monoxide can be summed, and the combustion efficiency is determined according to the ratio of the predicted concentration value of carbon dioxide to the summation result.

[0060] In one embodiment, the calculation formula of the combustion efficiency is as follows: Wherein, represents the combustion efficiency at time t. represents the predicted concentration value of carbon dioxide at time t. represents the predicted concentration value of carbon monoxide at time t.

[0061] In one embodiment, the predicted concentration values of various flue gas pollutants in the incineration emission gases can be weighted and summed to obtain the flue gas pollutant concentration. The flue gas pollutants may include carbon monoxide, nitrogen oxides, sulfur dioxide and hydrogen chloride. The weights of various flue gas pollutants are determined by the dimension of the gas average value.

[0062] The calculation formula of the flue gas pollutant concentration is as follows: Wherein, represents the flue gas pollutant concentration at time t. represents the predicted concentration value of carbon monoxide at time t. represents the predicted concentration value of nitrogen oxides at time t. represents the predicted concentration value of sulfur dioxide at time t. represents the predicted concentration value of hydrogen chloride at time t. 、 、 、 respectively represent the weights of carbon monoxide, nitrogen oxides, sulfur dioxide, and hydrogen chloride.

[0063] In the above embodiments, according to the predicted concentration values of carbon dioxide and carbon monoxide, the combustion efficiency can be accurately calculated. According to the predicted concentration values of nitrogen oxides, carbon monoxide, sulfur dioxide, and hydrogen chloride, the flue gas pollutant concentration can be accurately calculated. Furthermore, according to the combustion efficiency and the flue gas pollutant concentration, the value of the objective function can be accurately calculated, thereby optimizing the decision variables.

[0064] In one embodiment, the solution of the decision variables is iteratively optimized so that the value of the objective function obtained according to the solution of the decision variables is optimized in a decreasing direction, including: initializing the decision variable population as the original decision variable population in the first round of iteration; in each round of iteration, performing a mutation operation on each individual in the original decision variable population in this round of iteration to obtain a mutant population; performing a crossover operation on the original decision variable population and the mutant population in this round of iteration to obtain a trial population; merging the trial population and the original decision variable population to obtain a mixed population; determining the original decision variable population in the next round of iteration according to the values of the objective function respectively obtained from the solutions of the decision variables in the mixed population, and entering the next round of iteration.

[0065] Among them, the decision variable population is a population containing solutions of multiple groups of decision variables. Each solution of the decision variables contains the values of the flue gas oxygen content and the furnace temperature.

[0066] In one embodiment, the population size Np, the mutation operator F, the crossover rate CR, and the maximum number of iterations G can be set. For example: the population size Np can be set to 100, the mutation operator F can be set to 0.5, the crossover rate CR can be set to 0.7, and the maximum number of iterations G can be set to 30.

[0067] In one embodiment, to maintain population diversity, the niche technology is used to initialize the decision variable population. The individuals (i.e., the solutions of the decision variables) in the decision variable population are randomly numbered according to the population size Np and organized in a ring topology. For each individual, the neighborhood radius R is defined, and the R individuals within its neighborhood form a neighborhood group, thereby allocating an independent neighborhood for each individual. The ring topology based on the index restricts the direct interaction between different neighborhood individuals, thereby forming multiple stable search spaces during the process of searching for the solutions of the decision variables and effectively maintaining population diversity.

[0068] In one embodiment, Np real values are used as the individuals in the initialized decision variable population. Each individual in the decision variable population is represented as: Among them, \(i\) represents the serial number of the individual in the decision variable population, \(N_p\) represents the population size, \(Gen\) represents the current iteration number, and \(G\) represents the maximum iteration number.

[0069] In one embodiment, the individuals (solutions of decision variables) in the decision variable population are determined by constraint conditions, and the constraint conditions are set based on actual industrial requirements. The bounds of the parameter variables are set as Then, the constraint conditions for the individuals in the decision variable population are: Among them, \(Gen = 0\) is set during initialization. represents a uniformly random number generated between \([0, 1]\). represents the upper bound of the \(j\)-th dimension decision variable, represents the lower bound of the \(j\)-th dimension decision variable. There are two defined decision variables (i.e., flue gas oxygen content and furnace temperature), so \(j = 1, 2\). \(i\) represents the serial number of the individual in the decision variable population. \(N_p\) represents the population size. represents the value of the \(j\)-th decision variable in the \(i\)-th individual in the decision variable population in the 0-th round of iteration (i.e., initialization).

[0070] In one embodiment, in each round of iteration, a mutation operation is performed on each individual in the original decision variable population of this round of iteration to obtain a mutant population. A crossover operation is performed on the original decision variable population and the mutant population of this round of iteration to obtain a trial population, and the individuals in the trial population are made to conform to the constraint conditions through boundary condition detection. The individuals in the trial population can be expressed by the following formula: Among them, represents the \(j\)-th estimated value of a random number generator generated between \([0, 1]\). \(CR\) represents the crossover operator. represents the value of the \(j\)-th decision variable after mutation in the solution of the \(i\)-th decision variable in the original decision variable population in the \(Gen\)-th round of iteration. represents the value of the \(j\)-th decision variable in the solution of the \(i\)-th decision variable in the original decision variable population in the \(Gen\)-th round of iteration. represents the value of the \(j\)-th decision variable in the solution of the \(i\)-th decision variable in the original decision variable population in the \((Gen + 1)\)-th round of iteration.

[0071] Such as Figure 8As shown in the figure, it is a schematic flow chart of iteratively optimizing the solution of decision variables. First, initialize the decision variable population to obtain the original decision variable population. When the maximum number of iterations has not been reached, in each iteration, perform a mutation operation on the original decision variable population. If the current iteration number Gen is less than or equal to 20% of the maximum iteration number G, then adopt a single mutation strategy (mutation strategy). If the current iteration number Gen is greater than 20% of the maximum iteration number G and less than the maximum iteration number G, then adopt an adaptive mutation strategy, that is, according to the performance of each mutation strategy in the previous iteration, allocate multiple mutation strategies to each individual in the original decision variable population. Then perform a crossover operation on the mutated population and the original decision variable population to obtain a trial population, combine the trial population and the original decision variable population to obtain a mixed population, and use an environmental selection mechanism to select the original decision variable population in the next iteration from the mixed population, and then enter the next iteration. Stop iterating until the maximum number of iterations is reached.

[0072] In the above embodiment, in each iteration, perform a mutation operation on the individuals in the original decision variable population, then perform a crossover operation on the original decision variable population and the mutated population to obtain a trial population, combine the trial population and the original decision variable population to obtain a mixed population, and determine the original decision variable population in the next iteration according to the values of the objective functions respectively obtained from the solutions of each decision variable in the mixed population, and enter the next iteration. By mutation and crossover, the population diversity is improved, thereby improving the accuracy of the solution of the optimized decision variables.

[0073] In one embodiment, perform a mutation operation on each individual in the original decision variable population in the current iteration to obtain a mutated population, including: if the current iteration number is less than or equal to the target iteration number, then adopt a single mutation strategy to perform a mutation operation on each individual in the original decision variable population in the current iteration to obtain a mutated population; the target iteration number is obtained according to a preset ratio of the maximum iteration number; if the current iteration number is greater than the target iteration number, then allocate multiple mutation strategies to each individual in the original decision variable population in the current iteration according to their respective target ratios, and perform a mutation operation on the corresponding individuals using the allocated mutation strategies to obtain multiple mutated populations; wherein, the target ratio is determined according to the performance of each mutation strategy in the previous iteration.

[0074] In one embodiment, a single mutation strategy can be the DE / rand / 1 mutation strategy. Multiple mutation strategies can include the DE / rand / 1 mutation strategy and the DE / best / 1 mutation strategy. That is, if the current iteration number is less than or equal to the target iteration number, the DE / rand / 1 mutation strategy is used to perform a mutation operation on each individual in the original population of decision variables in this round of iteration, obtaining a mutated population. If the current iteration number is greater than the target iteration number, the DE / rand / 1 mutation strategy and the DE / best / 1 mutation strategy are allocated to each individual in the original population of decision variables in this round of iteration according to their respective target ratios, and the allocated mutation strategies are used to perform mutation operations on the corresponding individuals, obtaining the mutated populations corresponding to the DE / rand / 1 mutation strategy and the DE / best / 1 mutation strategy respectively.

[0075] In one embodiment, the preset ratio can be set according to the actual situation. For example, the preset ratio can be 20%. That is, if the current iteration number is less than or equal to 20% of the maximum iteration number, a single mutation strategy is used; if the current iteration number is greater than 20% of the maximum iteration number, multiple mutation strategies are used.

[0076] In one embodiment, in the process of allocating multiple mutation strategies to each individual in the original population of decision variables in this round of iteration according to their respective target ratios, first, according to the target ratios of each mutation strategy, a randomly arranged index array is generated, and the mutation strategy of each individual is determined by the value in the index array. According to the index array, a mutation strategy is allocated to each individual in the population of decision variable elements and stored in a matrix.

[0077] In one embodiment, the individuals in the mutated population obtained by using the DE / rand / 1 mutation strategy are represented as follows: Among them, r1, r2, and r3 are randomly selected and different serial numbers, and F represents a mutation operator.

[0078] In one embodiment, the individuals in the mutated population obtained by using the DE / best / 1 mutation strategy are represented as follows: Among them, r1 and r2 are randomly selected and different serial numbers, and x best is the optimal solution found so far in the current iteration. F represents a mutation operator.

[0079] In the above embodiments, a single mutation strategy is used when the number of iterations is small, and multiple mutation strategies are used when the number of iterations is large. The usage ratio of each mutation strategy is determined according to the performance of each mutation strategy in the previous round of iteration, thereby improving the effect of the mutation operation and further improving the population diversity.

[0080] In one embodiment, the step of determining the target ratio includes: determining the first fitness of the solutions of the decision variables in the mutant populations obtained by each mutation strategy in the previous iteration, and the second fitness of the solutions of the decision variables in the original population of the decision variables in the previous iteration; the fitness is determined according to the value of the objective function; determining the performance corresponding to each mutation strategy according to the distance between the first fitness corresponding to each mutation strategy and the second fitness; determining the target ratio corresponding to each mutation strategy according to the ratio of the performance corresponding to each mutation strategy to the total performance of various mutation strategies.

[0081] In one embodiment, the distance between the fitnesses can be calculated using the following formula: where is the Pareto front generated by the algorithm, that is, the set of non-dominated solutions in the current population. is the true Pareto front. In practical applications, it is difficult to obtain the true Pareto front, so an approximate Pareto front is used as a reference. is the true Pareto front The number of points in. v is a point in which represents an ideal solution. u is a point in the Pareto front generated by the algorithm which represents a solution found by the algorithm in the current iteration. d(v, u) is the Euclidean distance between the fitnesses of point v and point u, to measure the distance between the fitness of the solution generated by the algorithm and the fitness of the points on the true Pareto front.

[0082] In one embodiment, the performance corresponding to each mutation strategy can be calculated using the following formula: where represents the performance of the DE / rand / 1 mutation strategy. represents the performance of the DE / best / 1 mutation strategy. and respectively represent the fitnesses (the first fitnesses) of the individuals in the mutant population obtained using the DE / rand / 1 mutation strategy and the fitnesses (the first fitnesses) of the individuals in the mutant population obtained using the DE / best / 1 mutation strategy. The DE / rand / 1 mutation strategy has good exploration search ability, and the DE / best / 1 mutation strategy is good at local search and has a faster convergence speed. Represents the fitness (second fitness) of an individual in the original population of decision variables. Represents the distance between the fitness of an individual in the mutant population obtained using the DE / rand / 1 mutation strategy and the fitness of an individual in the original population of decision variables. Represents the distance between the fitness of an individual in the mutant population obtained using the DE / best / 1 mutation strategy and the fitness of an individual in the original population of decision variables. rand is a small random number added to increase numerical stability.

[0083] In one embodiment, the total performance of various mutation strategies can be determined according to the sum of the performances of various mutation strategies. The target ratio corresponding to each mutation strategy is obtained according to the ratio of the performance corresponding to each mutation strategy to the total performance of various mutation strategies.

[0084] In one embodiment, the total performance of various mutation strategies is calculated using the following formula: Where, Represents a small constant to avoid division by zero error. Represents the total performance of various mutation strategies. Represents the serial number of the mutation strategy. Represents the performance of the i-th mutation strategy.

[0085] In one embodiment, the number of individuals using each mutation strategy in this round of iteration can be calculated according to the following formula: Where, Represents the number of individuals using the first mutation strategy (DE / rand / 1 mutation strategy). Represents the number of individuals using the second mutation strategy (DE / rand / 1 mutation strategy). Represents rounding. Represents the performance of the first mutation strategy. Represents the performance of the second mutation strategy. Represents the total performance of various mutation strategies. Represents the target ratio corresponding to the first mutation strategy. Represents the target ratio corresponding to the second mutation strategy. Represents the population size of the population of decision variable elements).

[0086] In the above embodiments, the performance of each mutation strategy is determined according to the distance between the fitness of the individuals in the mutated population obtained by each mutation strategy in the previous iteration and the fitness of the individuals in the original population of decision variables. The proportion of the number of individuals using each mutation strategy in this iteration is determined according to the performance of each mutation strategy, which can accurately adjust the mutation strategy, thereby better improving the population diversity.

[0087] In one embodiment, the original population of decision variables in the next iteration is determined according to the values of the objective functions respectively obtained from the solutions of the decision variables in the mixed population, including: dividing the mixed population into multiple clusters; determining the non-dominated solutions and dominated solutions in each cluster according to the values of the objective functions respectively corresponding to the solutions of the decision variables in each cluster, dividing the non-dominated solutions into the elite archive, and dividing the dominated solutions into the non-elite archive; determining the original population of decision variables in the next iteration according to the elite archive.

[0088] In one embodiment, the affinity propagation algorithm can be used to divide the mixed population into multiple clusters.

[0089] In the above embodiments, the mixed population is divided into multiple clusters, the non-dominated solutions and dominated solutions in each cluster are determined according to the values of the objective functions respectively corresponding to the solutions of the decision variables in each cluster, the non-dominated solutions are divided into the elite archive, and the dominated solutions are divided into the non-elite archive, which can accurately determine the original population of decision variables in the next iteration.

[0090] In one embodiment, determining the original population of decision variables in the next iteration according to the elite archive includes: if the number of solutions of the decision variables in the elite archive does not reach the preset scale, quantifying the distribution density of the solutions of the decision variables in the non-elite archive in the decision space, selecting the solutions of the decision variables in the sparse region according to the distribution density and supplementing them to the elite archive, and selecting multiple solutions of the decision variables from the supplemented elite archive to form the original population of decision variables in the next iteration; if the number of solutions of the decision variables in the elite archive reaches the preset scale, directly selecting multiple solutions of the decision variables from the elite archive according to the values of the objective functions respectively corresponding to the solutions of the decision variables in the elite archive to form the original population of decision variables in the next iteration.

[0091] In one embodiment, the harmonic average distance criterion (HAD) can be used to quantify the distribution density of the solutions of the decision variables in the non-elite archive in the decision space.

[0092] In one embodiment, the calculation formula for the distribution density is: where S represents the set of all solutions in the population. ∣S∣ represents the total number of solutions. Y represents another solution in the population. Represents the Euclidean distance between the solution x and the solution y.

[0093] In the above embodiments, if the number of solutions of the decision variables in the elite archive does not reach the preset scale, the distribution density of the solutions of the decision variables in the non-elite archive in the decision space is quantified, and the solutions of the decision variables in the sparse region are selected according to the distribution density and supplemented into the elite archive. Multiple solutions of the decision variables are selected from the supplemented elite archive to form the original population of decision variables in the next iteration; if the number of solutions of the decision variables in the elite archive reaches the preset scale, multiple solutions of the decision variables are directly selected from the elite archive according to the values of the objective functions corresponding to the solutions of each decision variable in the elite archive to form the original population of decision variables in the next iteration, so that the original population of decision variables in the next iteration can be accurately determined.

[0094] In one embodiment, determining the solution of the target decision variable from the solutions of each group of decision variables obtained in the last iteration includes: respectively for each group of solutions of the decision variables obtained in the last iteration, respectively according to the first preset weight and the second preset weight, performing weighted summation on the first optimization objective value and the second optimization objective value corresponding to the solution of the decision variable to obtain the weighted summation result corresponding to the solution of the decision variable; the first optimization objective value is determined according to the combustion efficiency corresponding to the solution of the decision variable; the second optimization objective value is determined according to the flue gas pollutant concentration corresponding to the solution of the decision variable; the solution of the decision variable with the smallest weighted summation result is determined as the solution of the target decision variable.

[0095] Among them, each group of solutions of the decision variables obtained in the last iteration refers to the set of solutions of the decision variables determined according to the elite archive in the last iteration.

[0096] In one embodiment, the following formula can be used to perform weighted summation on the first optimization objective value and the second optimization objective value corresponding to the solution of the decision variable to obtain the weighted summation result corresponding to the solution of the decision variable: Among them, are the corresponding preset weights. f ij is the jth optimization objective value of the solution of the i-th group of decision variables. n is the number of optimization objectives (n = 2 in the present invention). m is the number of solutions in the Pareto optimal solution set (that is, the set of solutions of the decision variables obtained in the last iteration).

[0097] It can be understood that when the weighted summation result reaches the minimum value, it indicates that the solution of the i-th decision variable is the optimal set point, the optimal set value of the decision variable x can be obtained, and then the optimal furnace temperature and flue gas oxygen content set values can be determined, realizing the improvement of combustion efficiency while reducing the flue gas pollutant emission concentration.

[0098] In one embodiment, it is set according to the actual situation and decision-making preferences. The weight of the first optimization objective (combustion efficiency) is set to 0.6, and the weight of the second optimization objective (flue gas pollutant concentration) is set to 0.4. During the actual incineration process, the decision variables need to be optimized every first preset time interval. In each optimization process, from the Pareto optimal solution set obtained in the last round of iteration, the solution with the minimum objective function value (the solution of the target decision variables) is calculated by the linear weighted method and determined as the optimal set values of the furnace temperature and flue gas oxygen content.

[0099] In the above embodiment, for each set of solutions of the decision variables obtained in the last round of iteration, according to the first preset weight and the second preset weight respectively, the first optimization objective value and the second optimization objective value corresponding to the solutions of the decision variables are weighted and summed to obtain the weighted sum result corresponding to the solutions of the decision variables, and the solution of the decision variable with the minimum weighted sum result is determined as the solution of the target decision variables, which can balance the weights between the two objectives of combustion efficiency and flue gas pollutant concentration and accurately determine the solution of the target decision variables.

[0100] As Figure 9 and Figure 10 shown, they are respectively the optimization result diagrams of combustion efficiency and flue gas pollutant concentration after being optimized based on the above method. In the figure, the x-axis represents the optimization duration with the unit of min, and the y-axis represents combustion efficiency and flue gas pollutant concentration respectively with the units of % and mg / m 3 , and the white columns represent the actual output values of combustion efficiency and flue gas pollutant concentration before optimization, while the black columns represent the actual output values of combustion efficiency and flue gas pollutant concentration after optimization. It can be seen from the figure that the combustion efficiency after optimization has a significant improvement compared with the actual combustion efficiency before optimization, and the average combustion efficiency after optimization can be increased by 7.46%. The flue gas pollutant concentration after optimization has a significant decrease compared with the actual flue gas pollutant concentration before optimization, and the average flue gas pollutant concentration after optimization can be reduced by 12.80%.

[0101] As Figure 11 shown, it is the overall flow schematic diagram of the intelligent optimization operation method for the urban solid waste incineration process in each embodiment of the present invention, including the following steps: Step 1, collect the historical data of the urban solid waste incineration process to construct a sample data set, and divide it into a training set and a test set.

[0102] Step 2, establish operation index models for nitrogen oxides, carbon monoxide, carbon dioxide, sulfur dioxide, and hydrogen chloride by using a multi-module multi-task neural network.

[0103] Step 3, design an optimization objective function according to the combustion efficiency and flue gas pollutant emission concentration of urban solid waste incineration.

[0104] Step 4: Use the optimized objective function designed as the objective function of the algorithm, and use the multi-objective adaptive differential evolution algorithm based on the environmental selection mechanism to obtain the Pareto optimal solution set of the decision variables.

[0105] Step 5: Use the linear weighted sum method to balance the weights between the objectives, and obtain the optimal setting values of the decision variables from the Pareto optimal solution set.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent optimization operation method for the urban solid waste incineration process, characterized in that Including: During the incineration of municipal solid waste, obtain the environmental data inside the current incinerator and the concentration values of various incineration emission gases before a preset duration; Iteratively optimize the solution of the decision variables so that the value of the objective function obtained according to the solution of the decision variables is optimized in a decreasing direction; the decision variables include the oxygen content in the flue gas and the furnace temperature; Determine the solution of the target decision variable from the solutions of the decision variables in the last round of iteration, and determine the values of the oxygen content in the flue gas and the furnace temperature included in the solution of the target decision variable as the set values of the oxygen content in the flue gas and the furnace temperature during the current municipal solid waste incineration process; Among them, the calculation steps of the value of the objective function include: input the solution of the decision variables, the environmental data inside the current incinerator, and the concentration values of various incineration emission gases before the preset duration into the operation index models of various incineration emission gases that have been pre-trained, output the predicted concentration values of various current incineration emission gases, calculate the combustion efficiency and the flue gas pollutant concentration according to the predicted concentration values of various incineration emission gases, and calculate the value of the objective function according to the combustion efficiency and the flue gas pollutant concentration.

2. The intelligent optimization operation method for the urban solid waste incineration process according to claim 1, characterized in that, The training steps of the pre-trained operation index models of various incineration emission gases include: Based on the historical data of the municipal solid waste incineration process, construct a sample data set; the sample data set contains the sample oxygen content in the flue gas, the sample furnace temperature, the sample environmental data inside the incinerator, and the sample concentration values of various incineration emission gases during the historical process of municipal solid waste incineration; Iteratively input the sample oxygen content in the flue gas, the sample furnace temperature, the sample environmental data inside the incinerator, and the sample concentration values of various incineration emission gases before the preset duration at different times into the operation index models of various incineration emission gases to be trained, and output the sample predicted concentration values of various incineration emission gases at that time; According to the difference between the sample predicted concentration values of various incineration emission gases and the true sample concentration values at that time in the sample data set, adjust the parameters of the operation index models of various incineration emission gases to be trained and enter the next round of iteration until the iteration stops, and obtain the trained operation index models of various incineration emission gases.

3. The intelligent optimization operation method for the urban solid waste incineration process according to claim 2, wherein The iteratively inputting the sample oxygen content in the flue gas, the sample furnace temperature, the sample environmental data inside the incinerator, and the sample concentration values of various incineration emission gases before the preset duration at different times into the operation index models of various incineration emission gases to be trained, and outputting the sample predicted concentration values of various incineration emission gases at that time includes: Iteratively input the sample oxygen content in the flue gas, the sample furnace temperature, the sample environmental data inside the incinerator, and the sample concentration values of various incineration emission gases before the preset duration at different times as input variables into the input layer of each sub-module of the operation index models of various incineration emission gases to be trained; among them, each sub-module is respectively used to output the sample predicted concentration value of an incineration emission gas. In each of the sub-modules, the input variables are normalized by the input layer, and the normalized input variables are input into the hidden layer. The hidden layer performs a spatial mapping transformation on the normalized input variables, and the input variables after the spatial mapping transformation are input into the output layer. The output layer performs a weighted summation on the input variables after the spatial mapping transformation, and outputs the sample predicted concentration value of the incineration emission gas corresponding to the sub-module.

4. The intelligent optimization operation method for the urban solid waste incineration process according to claim 1, wherein The incineration emission gas includes nitrogen oxides, carbon monoxide, carbon dioxide, sulfur dioxide, and hydrogen chloride; The calculating the combustion efficiency and the flue gas pollutant concentration according to the predicted concentration values of the various incineration emission gases includes: Calculating the combustion efficiency according to the predicted concentration values of the carbon dioxide and the carbon monoxide; Calculating the flue gas pollutant concentration according to the predicted concentration values of the nitrogen oxides, the carbon monoxide, the sulfur dioxide, and the hydrogen chloride.

5. The intelligent optimization operation method for the urban solid waste incineration process according to any one of claims 1 to 4, characterized in that, The iteratively optimizing the solution of the decision variables so that the value of the objective function obtained according to the solution of the decision variables is optimized in a decreasing direction includes: Initializing the decision variable population as the original decision variable population in the first round of iteration; In each round of iteration, performing a mutation operation on each individual in the original decision variable population in the current round of iteration to obtain a mutant population; Performing a crossover operation on the original decision variable population and the mutant population in the current round of iteration to obtain a trial population; Combining the trial population and the original decision variable population to obtain a mixed population; Determining the original decision variable population in the next round of iteration according to the values of the objective function respectively obtained from the solutions of the decision variables in the mixed population, and entering the next round of iteration.

6. The intelligent optimization operation method for the urban solid waste incineration process according to claim 5, characterized in that, The performing a mutation operation on each individual in the original decision variable population in the current round of iteration to obtain a mutant population includes: If the current iteration number is less than or equal to the target iteration number, performing a mutation operation on each individual in the original decision variable population in the current round of iteration by using a single mutation strategy to obtain a mutant population; the target iteration number is obtained according to a preset ratio of the maximum iteration number; If the current iteration number is greater than the target iteration number, distributing multiple mutation strategies to each individual in the original decision variable population in the current round of iteration according to their respective target ratios, and performing a mutation operation on the corresponding individuals by using the distributed mutation strategies to obtain multiple mutant populations; Wherein, the target ratio is determined according to the performance of each mutation strategy in the previous round of iteration.

7. The intelligent optimization operation method for the urban solid waste incineration process according to claim 6, characterized in that, The determining step of the target ratio includes: Determining the first fitness of the solutions of the decision variables in the mutant populations respectively obtained by each mutation strategy in the previous round of iteration, and the second fitness of the solutions of the decision variables in the original decision variable population in the previous round of iteration; the fitness is determined according to the value of the objective function; Determining the performance corresponding to each mutation strategy according to the distance between the first fitness corresponding to each mutation strategy and the second fitness; Determine the target ratio corresponding to each mutation strategy according to the ratio between the performance corresponding to each mutation strategy and the total performance of various mutation strategies.

8. The intelligent optimization operation method for the urban solid waste incineration process according to claim 5, characterized in that, Determining the original population of decision variables in the next iteration according to the values of the objective function obtained from the solutions of each decision variable in the mixed population includes: Dividing the mixed population into multiple clusters; Determining the non-dominated solutions and dominated solutions in each cluster according to the values of the objective function corresponding to the solutions of the decision variables in each cluster, dividing the non-dominated solutions into the elite archive, and dividing the dominated solutions into the non-elite archive; Determining the original population of decision variables in the next iteration according to the elite archive.

9. The intelligent optimization operation method for the urban solid waste incineration process according to claim 8, characterized in that The determining the original population of decision variables in the next iteration according to the elite archive includes: If the number of solutions of the decision variables in the elite archive does not reach the preset scale, quantify the distribution density of the solutions of the decision variables in the non-elite archive in the decision space, select the solutions of the decision variables in the sparse region according to the distribution density and supplement them to the elite archive, and select the solutions of multiple decision variables from the supplemented elite archive to form the original population of decision variables in the next iteration; If the number of solutions of the decision variables in the elite archive reaches the preset scale, directly select the solutions of multiple decision variables from the elite archive according to the values of the objective function corresponding to the solutions of each decision variable in the elite archive to form the original population of decision variables in the next iteration.

10. The intelligent optimization operation method for the urban solid waste incineration process according to any one of claims 1 to 4, characterized in that, The determining the solution of the target decision variable from the solutions of each group of decision variables obtained in the last iteration includes: For each group of solutions of decision variables obtained in the last iteration, respectively, according to the first preset weight and the second preset weight, perform weighted summation on the first optimization objective value and the second optimization objective value corresponding to the solution of the decision variable to obtain the weighted summation result corresponding to the solution of the decision variable; the first optimization objective value is determined according to the combustion efficiency corresponding to the solution of the decision variable; the second optimization objective value is determined according to the flue gas pollutant concentration corresponding to the solution of the decision variable; Determine the solution of the decision variable with the smallest weighted summation result as the solution of the target decision variable.

Images