An intelligent emission reduction method and system based on big data

Abstract

The present invention relates to the field of cement industry production control technology, and specifically to a big data-based intelligent emission reduction method and system. The method comprises: constructing a pollutant gas concentration model and a fuel utilization thermal efficiency model respectively; constructing a multi-objective function; optimizing and solving the multi-objective function using an improved multi-objective hybrid genetic particle swarm algorithm to obtain an optimal solution set; constructing a raw material calcination temperature prediction model; establishing a control objective function based on the optimal solution set and the raw material calcination temperature prediction model, and obtaining the optimal temperature corresponding to the minimum value of the control objective function; inputting parameter data collected in real time during the cement production process into the raw material calcination temperature prediction model to obtain a predicted temperature value, calculating the error between the predicted temperature value and the corresponding optimal temperature, and inputting the error into the optimal temperature at the next moment through feedback correction. That is, the solution of the present invention can achieve the purpose of intelligent emission reduction by accurately controlling the temperature during the cement production process.

Description

Technical Field

[0001] The present invention relates to the technical field of cement industry production control, and more specifically, to a big data-based intelligent emission reduction method and system. Background Art

[0002] With the continuous improvement of global awareness of environmental protection and the advancement of sustainable development goals, the pollutant gases generated in the cement production process account for a certain proportion of global emissions. Therefore, intelligent emission reduction of cement is particularly important.

[0003] Currently, intelligent cement emission reduction primarily involves improving and optimizing the cement production process to reduce energy consumption and pollutant emissions. The proper operation of the precalciner is a critical step in the cement production process, and precalciner temperature is a key indicator affecting the decomposition rate of cement raw materials. Therefore, maintaining a normal and stable precalciner temperature is crucial to ensuring cement quality. Therefore, it is particularly important to stably and accurately control the precalciner temperature to ensure normal operation.

[0004] In the related art, for example, the patent application document with publication number CN117406601A discloses an energy-saving and emission-reducing cement decomposition furnace alternative fuel combustion optimization control method. This method uses a multi-objective particle swarm algorithm to solve the optimal temperature of the decomposition furnace, maintain the normal operation of the decomposition furnace under the condition of increasing the amount of garbage fed to the decomposition furnace, and limit NOx emissions to achieve energy conservation and emission reduction.

[0005] However, the multi-objective particle swarm algorithm used in the above method is prone to local optimal problems and low diversity of solution sets. That is, it does not take into account that the particle swarm in multi-objective optimization may quickly converge to a small area, resulting in a lack of diversity in the solution set. Ultimately, a set of local non-inferior similar solution sets will be obtained, making it impossible to achieve the optimal set temperature value when controlling the temperature of the cement decomposition furnace. Summary of the Invention

[0006] The purpose of the present invention is to propose an intelligent emission reduction method and system based on big data to solve the problem that the existing technology cannot accurately control the temperature in the cement production process and cannot achieve the purpose of intelligent emission reduction; to this end, the present invention provides solutions in the following two aspects.

[0007] In a first aspect, a smart emission reduction method based on big data includes:

[0008] Obtain parameter data of the cement production process in the current time period; the parameter data includes but is not limited to calcination temperature, oxygen content, secondary air temperature, and rotary kiln speed;

[0009] Input parameter data into a pre-built raw material calcination temperature prediction model to obtain a predicted temperature value;

[0010] Calculate the error between the predicted temperature value and the pre-acquired optimal temperature, input the error into the optimal temperature of the next time period through feedback correction, and perform temperature control in the cement production process;

[0011] The process of obtaining the optimal temperature is as follows:

[0012] Using historical parameter data from the cement production process, we construct pollution gas concentration models and fuel utilization thermal efficiency models.

[0013] A multi-objective function is constructed based on the pollutant gas concentration model and the fuel utilization thermal efficiency model; an improved multi-objective hybrid genetic particle swarm algorithm is used to optimize and solve the multi-objective function to obtain the optimal solution set;

[0014] A control objective function is established according to the optimal solution set and the raw material calcination temperature prediction model, and an optimal temperature corresponding to the minimum value of the control objective function is obtained.

[0015] The above scheme constructs a multi-objective function and uses an improved multi-objective hybrid genetic particle swarm algorithm to optimize and solve the multi-objective function to obtain the optimal solution set in the cement production process. Subsequently, a temperature prediction model is constructed to obtain the error between the predicted temperature and the optimal temperature (control temperature), and the temperature control of the next time period of cement production is carried out, thereby realizing precise temperature control and achieving the purpose of intelligent emission reduction in the cement production process.

[0016] Optionally, the process of optimizing and solving the multi-objective function using the improved multi-objective hybrid genetic particle swarm algorithm to obtain the optimal solution set is as follows:

[0017] Step 1: Set the key parameters of the particle swarm and genetic hybrid algorithm, including the maximum number of iterations, mutation probability, crossover probability, particle swarm size, and population size; let each particle in the population represent a data sample, where the data sample includes thermal efficiency, pollutant gas concentration, calcination temperature, oxygen content, total fuel input, fuel ratio, combustion time, secondary air temperature, and rotary kiln speed; initialize the position, velocity, individual extreme value, and group extreme value of the particles in the population;

[0018] Step 2: In the search space of size N, the particles are divided based on their positions to generate n initial sub-populations. The initial sub-population size is

[0019] Step 3: Calculate the multi-objective dominance of each initial subpopulation, stratify the individuals within each initial subpopulation, and perform selection, crossover, and mutation operations on each individual. Update the speed and position of each individual according to the speed and position update formula of the particle swarm optimization algorithm. Repeat multiple iterations. If the maximum number of iterations is reached, the optimization ends and the optimal solution of multiple iterations is output.

[0020] Step 4: Select any one of the initial subpopulations and its nearest neighboring initial subpopulation, and determine the diversity deficiency of the two initial subpopulations based on the optimal solution. If at least one of the initial subpopulations has diversity deficiency, calculate the individual migration rate of each individual in the two initial subpopulations. In response to the individual migration rate being greater than a set threshold, migrate the corresponding individual to the other initial subpopulation, thereby obtaining two updated populations, and further obtaining all updated populations.

[0021] Step 5: Repeat steps 3 and 4 for all updated populations, and finally obtain the optimal solution set consisting of the optimal solutions of n populations.

[0022] The above scheme utilizes the diversity of the population during the population iteration process to migrate individuals in the population, so that the diversity between different populations meets the needs and avoids the problem of obtaining a single or several local optimal solutions.

[0023] Optionally, the individual mobility P i for:

[0024]

[0025] in, is the size of the population, α and β are weights, D i is the average multi-objective dominance value of individual i, D s is the sum of the average multi-objective dominance of all individuals in the population, represents the selection probability of individual i in the roulette wheel selection operator, F i is the individual contribution score of individual i to population diversity, F s is the sum of the contribution scores of all individuals in the population, Indicates the importance of individual i in the population.

[0026] The above scheme provides a way to accurately calculate individual migration rates.

[0027] Optionally, the individual contribution score is:

[0028]

[0029] in, is the size of the population, x iis the solution of the i-th individual in the population, x j is the solution of the jth individual in the population, d(x i ,x j ) is x i and x j The distance between them, D(x i ,x j ) is x i and x j The multi-objective dominance between them, μ is the parameter that adjusts the exponential decay rate, and σ is the standard deviation of the distance between individuals in the population;

[0030] The multi-objective dominance D(x i ,x j )for:

[0031]

[0032] Among them, w m is the weight of the mth objective function, d m (x i ,x j ) is the mth objective function on x i Relative to x j The target dominance contribution, f m (x i ) is the objective function value, for each objective function f m (x) <f m (x j ) represents the solution x i Better than x on target j .

[0033] By evaluating the diversity of the population and analyzing the optimal solution, we can provide data support for subsequent individual migration rates.

[0034] Optionally, the multi-objective function is:

[0035] Among them, f1 is the objective function of fuel utilization thermal efficiency, y max is the maximum predicted value of thermal efficiency at the end of the tth iteration, y0 is the optimal value of thermal efficiency, n i is the first input variable for predicting fuel utilization thermal efficiency; f2 is the objective function of pollutant gas concentration, y′ min is the lowest predicted value of pollutant gas at the end of the tth iteration, y0′ is the standard value of pollutant gas emission, u iIt is the second input variable for pollutant gas concentration prediction; the first input variable is the calcination temperature, oxygen content, secondary air temperature, and rotary kiln speed in the current time period; the second input variable is the calcination temperature, oxygen content, total fuel input, fuel ratio, and combustion time in the current time period.

[0036] Comprehensive analysis of the objective function in the cement production process from multiple perspectives will facilitate subsequent acquisition of the optimal solution to achieve the goal of intelligent emission reduction.

[0037] Optionally, the pollutant gas concentration model adopts an RNN network model; the fuel utilization thermal efficiency model is an LSTM network model.

[0038] Optionally, the raw material calcination temperature prediction model is a support vector machine model based on a Gaussian kernel function.

[0039] Optionally, the control objective function is: the minimum square error between the temperature prediction value and the optimal temperature in the optimal solution set.

[0040] Optionally, the optimal temperature value is: y′(k+1)=y(k+1)+rδ(k); wherein y′(k+1) is the adjusted optimal temperature value at time k+1; y(k+1) is the optimal temperature at time k+1 obtained by the optimization control algorithm, δ(k) is the error, and r is the feedback coefficient, and the value range is (0, 1).

[0041] In a second aspect, the present invention provides an intelligent emission reduction system based on big data, characterized by comprising:

[0042] processor;

[0043] A memory stores computer instructions for intelligent emission reduction based on big data. When the computer instructions are executed by the processor, the system executes the above-mentioned intelligent emission reduction method based on big data.

[0044] The beneficial effects of the present invention are:

[0045] The solution of the present invention can improve the multi-objective hybrid genetic particle swarm algorithm and optimize various parameters in the cement production process to achieve the optimal solution set. It avoids the problem that the multi-objective particle swarm optimization algorithm explores the space of nonlinear relationships by moving multiple particles in the search space, and relies on increasing the number of particles to slow down the trend of local convergence, which will lead to overfitting and reduce the ability to find the global optimal solution. The optimal parameters in the cement production process are obtained to achieve the purpose of intelligent emission reduction. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an illustrative and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0047] Figure 1 The following is a schematic diagram showing a flow chart of the steps of a smart emission reduction method based on big data in this embodiment;

[0048] Figure 2 The structural block diagram of a big data-based intelligent emission reduction system in this embodiment is schematically shown. DETAILED DESCRIPTION

[0049] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0050] like Figure 1 As shown, an intelligent emission reduction method based on big data in this embodiment includes the following steps:

[0051] Step S1: Acquire parameter data of the cement production process in the current time period. Specifically, the parameter data in this embodiment include but are not limited to calcination temperature, oxygen content, secondary air temperature, and rotary kiln speed.

[0052] Parameter data may also include combustion heat utilization efficiency, total fuel input, fuel ratio, combustion time, concentration of pollutant gases, etc.

[0053] Among them, the calcination temperature is measured by a temperature sensor installed in the rotary kiln; the total input amount of fuel and the fuel ratio are obtained based on the content of multiple fuels input; the above-mentioned combustion heat utilization efficiency is based on utilized heat / generated heat, and the heat is calculated using a physical formula. The heat generated by the combustion of heat-generating fuel is Q=mq, where m is the fuel mass, q is the combustion value, and utilized heat is the heat absorbed by the calcination of raw materials; the above-mentioned pollutant gas concentration is measured by a gas detector.

[0054] Furthermore, in this embodiment, data cleaning is performed on the acquired parameter data to obtain cleaned parameter data in order to remove noise.

[0055] Step S2: Input the parameter data into a pre-built raw material calcination temperature prediction model to obtain a predicted temperature value; calculate the error between the predicted temperature value and the pre-obtained optimal temperature, input the error into the optimal temperature of the next time period through feedback correction, obtain the optimal temperature value, and perform temperature control in the cement production process.

[0056] The raw material calcination temperature prediction model in this embodiment is a support vector machine model based on a Gaussian kernel function.

[0057] Among them, the training process of the raw material calcination temperature prediction model is:

[0058] The parameter data in multiple set time periods in the historical parameter data are input into the raw material calcining temperature prediction model, the calcining temperatures in the rotary kiln corresponding to the multiple time periods are output, and the raw material calcining temperature prediction model is trained.

[0059] In this embodiment, the process of obtaining the optimal temperature includes steps S21 to S25, which are specifically as follows:

[0060] In step S21 , historical parameter data of the cement production process are used to construct a pollution gas concentration model and a fuel utilization thermal efficiency model.

[0061] Among them, historical parameter data include calcination temperature, oxygen content, combustion heat utilization efficiency, total fuel input, fuel ratio, combustion time, concentration of pollutant gas and secondary air temperature.

[0062] In this embodiment, a fuel utilization thermal efficiency model is constructed and trained using historical parameter data to obtain a trained fuel utilization thermal efficiency model. This model is then used to predict the fuel utilization thermal efficiency during the current cement production process. The fuel utilization thermal efficiency model in this embodiment is an RNN network model, but other network models, such as a CNN network model, may also be used.

[0063] Taking the RNN network model as an example, its training process is as follows:

[0064] Obtaining a training data set. The training data set in this embodiment is obtained from historical parameter data, that is, the calcination temperature, oxygen content, secondary air temperature, and rotary kiln speed in each historical time period in the historical parameter data of the cement production process are obtained as input data, and the fuel utilization thermal efficiency in the corresponding time period is obtained as output data.

[0065] The training data set is input into the RNN network model, and the RNN network model is trained to obtain a trained RNN network model.

[0066] During the training process, the loss function used by the RNN network model is the mean square error loss function.

[0067] In this embodiment, a pollutant gas concentration model is constructed and trained using historical parameter data to obtain a trained pollutant gas concentration model, which is used to predict the pollutant gas concentration that may be generated in the current cement production process.

[0068] The pollutant gas concentration model LSTM network model in this embodiment can of course also adopt other network models, such as RNN network model.

[0069] Taking the LSTM network model as an example, its training process is:

[0070] Obtaining a training data set. The training data set in this embodiment is obtained from historical parameter data, that is, the calcination temperature, oxygen content, total fuel input, fuel ratio, and combustion time in each historical time period in the historical data of the cement production process are obtained as input data, and the pollutant gas concentration in the corresponding time period is obtained as output data.

[0071] The training data set is input into the LSTM network model, and the LSTM network model is trained to obtain a trained LSTM network model.

[0072] During the training process, the loss function used by the LSTM network model is the mean square error loss function.

[0073] Step S22: construct a multi-objective function based on the pollutant gas concentration model and the fuel utilization thermal efficiency model; use an improved multi-objective hybrid genetic particle swarm algorithm to optimize and solve the multi-objective function to obtain an optimal solution set.

[0074] The multi-objective function in this embodiment is:

[0075] The multi-objective function is:

[0076] Among them, f1 is the fuel utilization thermal efficiency optimization objective function, y max,t is the maximum predicted value of thermal efficiency at the end of the tth iteration, y0 is the optimal value of thermal efficiency, n i is the input variable for predicting the thermal efficiency of fuel utilization; f2 is the objective function for optimizing the concentration of pollutants, y′ min,t is the lowest predicted value of pollutant gas at the end of the tth iteration, y0′ is the standard value of pollutant gas emission, u i It is the input variable for predicting the concentration of pollutant gas.

[0077] The above-mentioned optimal value of thermal efficiency is determined based on historical experience. For example, the highest thermal efficiency in historical experience can be used as the optimal value of thermal efficiency, and the pollutant gas emission standard value is determined based on the existing emission standard.

[0078] The above objective function is constructed to improve the thermal efficiency of fuel utilization while keeping the fuel usage as low as possible, while ensuring low emission requirements of pollutant gases.

[0079] It should be noted that complex nonlinear relationships exist between characteristics such as raw material calcination temperature, combustion thermal efficiency, combustion time, pollutant gas concentration, air volume, and oxygen concentration, which may contain multiple local optimal solutions. The multi-objective particle swarm optimization algorithm explores the space of nonlinear relationships by moving multiple particles in the search space. Relying on increasing the number of particles to slow the trend of local convergence will also lead to overfitting and reduce the ability to find the global optimal solution. In addition, only a single or a few local optimal solutions can be obtained in multi-peak optimization. Therefore, in this embodiment, an improved multi-objective hybrid genetic particle swarm algorithm is used to optimize and solve the multi-objective function. The improved multi-objective hybrid genetic particle swarm algorithm in this embodiment is MO-HPSOGA.

[0080] In this embodiment, the specific process of obtaining the optimal solution includes steps 221 to 225:

[0081] Step S221, set the key parameters of the particle swarm and genetic hybrid algorithm, including the maximum number of iterations, mutation probability, crossover probability, particle swarm size and population size; let each particle in the population represent a data sample, where the data sample includes thermal efficiency, pollutant gas concentration, calcination temperature, oxygen content, total fuel input, fuel ratio, combustion time, secondary air temperature, and rotary kiln speed; initialize the position, velocity, individual extreme value and group extreme value of the particles in the population.

[0082] Step S222: In the search space of size N, the particles are divided based on their positions to generate n initial sub-populations. The size of the initial sub-population is Where N is greater than n, and n is greater than 1.

[0083] In step S223, the multi-objective dominance of each initial subpopulation is calculated, the individuals within each initial subpopulation are stratified, and selection, crossover, and mutation operations are performed on each individual. The speed and position of each individual are updated according to the speed and position update formula of the particle swarm optimization algorithm; and multiple iterations are performed. If the maximum number of iterations is reached, the optimization ends and the optimal solution of multiple iterations is output.

[0084] In multi-objective optimization, if a solution x is not inferior to another solution y in all objectives and is better than y in at least one objective, then x is said to dominate y. The initial connection strength between each population and its neighboring populations is defined as P0 (the default is 1), and the multi-objective dominance within the population is calculated. Each population and the population are stratified, and each population independently performs genetic operations of selection, crossover, and mutation to screen out excellent sub-region solutions (local optimal solutions). The specific multi-objective dominance D(x i ,x j ) is calculated as:

[0085]

[0086] Among them, w m is the weight of the mth objective function, d m (x i ,x j ) is the mth objective function on x i Relative to x j The target dominance contribution, f m (x i ) is the objective function value, for each objective function f m (x) <f m (x j ) represents the solution x i Better than x on target j , x i is the solution of the i-th individual in the population, x j is the solution of the jth individual in the population.

[0087] Among them, D(x i ,x j ) Characterization and solution j Compared to x i The degree of quality.

[0088] Step S224: select any one of the initial subpopulations and its nearest neighboring initial subpopulation, and determine the diversity deficiency of the two initial subpopulations based on the optimal solution. If at least one of the initial subpopulations has diversity deficiency, calculate the individual migration rate of each individual in the two initial subpopulations. In response to the individual migration rate being greater than a set threshold, migrate the corresponding individual to the other initial subpopulation to obtain two updated populations, and then obtain all updated populations.

[0089] In this embodiment, the nearest neighbor initial subpopulation selected for any initial subpopulation is the closest population in the search space.

[0090] In the above, the method for determining the lack of diversity in the two initial subpopulations based on the optimal solution is as follows: if a population fails to find a new local optimal solution for several consecutive generations, it indicates a lack of diversity; otherwise, it indicates rich diversity. The consecutive generations actually refer to multiple consecutive iterations, which can be 10, 20, or more. The specific number of iterations can be determined based on the actual situation.

[0091] Therefore, there are two cases of diversity deficiency of the two initial subpopulations based on the optimal solution. The first case is that at least one initial subpopulation has insufficient diversity; the second case is that both initial subpopulations have sufficient diversity.

[0092] For the first type, when population diversity is lacking, it is necessary to increase the connection strength between the population and other populations (individual migration). In this embodiment, when population diversity is lacking, the individual contribution score of each individual in the two selected initial subpopulations is calculated, and the individual migration rate between populations is calculated based on the individual contribution score and the improved adaptive hybrid selection operator.

[0093] The individual contribution score of an individual to population diversity in this embodiment is:

[0094]

[0095] in, is the size of the population, x i is the solution of the i-th individual in the population, x j is the solution of the jth individual in the population, d(x i ,x j ) is x i and x j The distance between them, D(x i ,x j ) is x i and x j is the multi-objective dominance between them, μ is the parameter that adjusts the exponential decay rate, and σ is the standard deviation of the distance between individuals in the population.

[0096] Among them, F i is the diversity contribution score of individual i to the population. Individuals with different dominance have different contributions to diversity. When the distance between individuals is large, the contribution of distant individuals to the diversity metric gradually decreases through the exponential decay function, which can better adapt to the actual distribution of the population.

[0097] Individual migration rate P of individuals selected to migrate i Expressed as:

[0098]

[0099] in, is the size of the population, α and β are adjustment coefficients to balance the importance of multi-objective dominance and diversity, i.e., weights, P i is the probability of individual i being selected, individuals with high diversity and dominance are more likely to be selected, D i is the average multi-objective dominance value of individual i (the average dominance of the individual and the rest of the population), D s is the sum of the average multi-objective dominance of all individuals in the population, represents the selection probability of individual i in the roulette wheel selection operator, F i is the individual contribution score of individual i to population diversity, F a is the sum of the individual contribution scores of all individuals in the population, Indicates the importance of individual i in the population.

[0100] If the mean multi-objective dominance is high, increase α; if diversity is low, increase β. After obtaining the individual migration rate, when the individual migration rate exceeds a set threshold, the individual can be migrated to another population. It should be noted that during the migration process, it is necessary to increase the diversity of the scarce population while maintaining the population diversity above the threshold to avoid premature convergence.

[0101] It should be noted that when both initial sub-populations are deficient, and there is still a lack of diversity after migration, one can replace any of the initial sub-populations and re-migrate individuals between different populations.

[0102] For the second type, when the diversity of both populations is not lacking, no migration between individuals in the two initial subpopulations is performed.

[0103] The reason for calculating the migration rate of individuals in the above is that the population distribution is random, and the global solutions within some populations are similar. Therefore, when the population diversity is scarce, the use of multi-objective dominance stratification cannot accurately obtain the regional optimal solution. The new individuals generated by genetic operations may be similar to each other, and then there will be a problem of no new local optimal solution for several consecutive generations. At this time, it is necessary to migrate between populations and individuals within the population to avoid falling into the problem of local optimal solutions.

[0104] Of course, as another implementation, in this embodiment, diversity can also be determined directly based on the diversity score of each population (the sum of the individual contribution scores of all individuals in the population), that is, by comparing the diversity score with a threshold to determine whether the population is lacking in diversity. Of course, this can also be determined based on actual circumstances.

[0105] In step S225, steps 3 and 4 are repeated for all updated populations, ultimately obtaining an optimal solution set consisting of the optimal solutions of n populations. In this embodiment, after obtaining all updated populations, iterations are continued for each individual in every two updated populations until the maximum number of iterations is reached, resulting in an optimal solution set consisting of the optimal solutions of n populations.

[0106] At the same time, in this embodiment, the population needs to be updated iteratively multiple times to avoid falling into the problem of local optimal solution.

[0107] The improved multi-objective hybrid genetic particle swarm algorithm mentioned above solves the problems of single local optimal solution and solution set diversity, performs accurate global optimization solution, and the solution set constitutes the Pareto frontier. The global optimal solution is selected according to the emission reduction focus target.

[0108] Step S23: Establish a control objective function based on the optimal solution set and the raw material calcination temperature prediction model, and obtain the optimal temperature corresponding to the minimum value of the control objective function. In this embodiment, the control objective function is the minimum square error between the temperature prediction value output by the raw material calcination temperature prediction model and the optimal temperature in the optimal solution set. The optimal temperature corresponding to the minimum value of the control objective function is selected as the optimal temperature for the current time period.

[0109] In this embodiment, the optimal temperature value is: y′(k+1)=y(k+1)+rδ(k); wherein y′(k+1) is the adjusted optimal temperature value at time k+1; y(k+1) is the optimal temperature at time k+1 obtained by the optimization control algorithm, δ(k) is the error, and r is the feedback coefficient, with a value range of (0, 1).

[0110] The solution of the present invention can improve the multi-objective hybrid genetic particle swarm algorithm and optimize various parameters in the cement production process to achieve the optimal solution set. It avoids the problem that the multi-objective particle swarm optimization algorithm explores the space of nonlinear relationships by moving multiple particles in the search space, and relies on increasing the number of particles to slow down the trend of local convergence, which will lead to overfitting and reduce the ability to find the global optimal solution. The optimal parameters in the cement production process are obtained to achieve the purpose of intelligent emission reduction.

[0111] The present invention also provides an intelligent emission reduction system based on big data. Figure 2 As shown, the system includes a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, an intelligent emission reduction system based on big data according to the present invention is implemented.

[0112] The intelligent emission reduction system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

[0113] In the present invention, the aforementioned memory can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, the computer-readable storage medium can be any suitable magnetic storage medium or magneto-optical storage medium, such as resistive random access memory (RRAM), dynamic random access memory (DRAM), static random access memory (SRAM), enhanced dynamic random access memory (EDRAM), high-bandwidth memory (HBM), hybrid memory cube (HMC), etc., or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium can be part of, accessible to, or connectable to a device. Any application or module described in the present invention can be implemented using computer-readable / executable instructions that can be stored or otherwise retained by such a computer-readable medium.

[0114] In the description of this specification, “a plurality of” means at least two, for example, two, three or more, etc., unless otherwise clearly defined.

[0115] While several embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous modifications, variations, and alternatives will occur to those skilled in the art without departing from the concept and spirit of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be employed in practicing the present invention.

Claims

1. An intelligent emission reduction method based on big data, characterized in that: include: Obtain parameter data of cement production process in the current time period; The parameter data include but are not limited to calcination temperature, oxygen content, secondary air temperature and rotary kiln speed; Input parameter data into a pre-built raw material calcination temperature prediction model to obtain a predicted temperature value; Calculate the error between the predicted temperature value and the pre-acquired optimal temperature, input the error into the optimal temperature of the next time period through feedback correction, and perform temperature control in the cement production process; The process of obtaining the optimal temperature is as follows: Using historical parameter data from the cement production process, we construct pollution gas concentration models and fuel utilization thermal efficiency models. A multi-objective function is constructed based on the pollutant gas concentration model and the fuel utilization thermal efficiency model; Step 1, key parameters of the particle swarm and genetic hybrid algorithm are set, including the maximum number of iterations, mutation probability, crossover probability, particle swarm size, and population size; each particle in the population is made to represent a data sample, where the data sample includes thermal efficiency, pollutant gas concentration, calcination temperature, oxygen content, total fuel input, fuel ratio, combustion time, secondary air temperature, and rotary kiln speed; the position, velocity, individual extreme value, and group extreme value of the particles in the population are initialized; Step 2: In the search space of size N, the particles are divided based on their positions to generate n initial sub-populations. The initial sub-population size is ; Step 3: Calculate the multi-objective dominance of each initial subpopulation, stratify the individuals within each initial subpopulation, and perform selection, crossover, and mutation operations on each individual. Update the speed and position of each individual according to the speed and position update formula of the particle swarm optimization algorithm. Repeat multiple iterations. If the maximum number of iterations is reached, the optimization ends and the optimal solution of multiple iterations is output. Step 4: Select any one of the initial subpopulations and its nearest neighboring initial subpopulation, and determine the diversity deficiency of the two initial subpopulations based on the optimal solution. If at least one of the initial subpopulations has diversity deficiency, calculate the individual migration rate of each individual in the two initial subpopulations. In response to the individual migration rate being greater than a set threshold, migrate the corresponding individual to the other initial subpopulation, thereby obtaining two updated populations, and further obtaining all updated populations. Step 5: Repeat steps 3 and 4 for all updated populations, and finally obtain the optimal solution set consisting of the optimal solutions of n populations; Individual contribution scores are: in, is the size of the population, Is the first in the population Individual solutions, Is the first in the population Individual solutions, yes and The distance between yes and The multi-objective dominance between is a parameter that adjusts the exponential decay rate, is the standard deviation of the distances between individuals in the population; The multi-objective dominance for: in, It is The weight of the objective function, It is On the objective function Relative to The target dominance contribution of is the objective function value, for each objective function Time table interpretation Outperforms in target ; A control objective function is established according to the optimal solution set and the raw material calcination temperature prediction model, and an optimal temperature corresponding to the minimum value of the control objective function is obtained.

2. The intelligent emission reduction method based on big data according to claim 1, characterized in that: The individual mobility for: in, is the size of the population, and is the weight, For individuals The average multi-objective dominance value of is the sum of the average multi-objective dominance of all individuals in the population, represents the individual in the roulette wheel selection operator The probability of selection, For individuals The individual contribution score to population diversity, is the sum of the contribution scores of all individuals in the population, Represents individuals in a population the importance of.

3. The intelligent emission reduction method based on big data according to claim 1, characterized in that: The multi-objective function is: ; in, is the objective function of fuel utilization thermal efficiency, is the maximum predicted value at the end of the tth iteration using thermal efficiency, To utilize the optimal value of thermal efficiency, It is the first input variable when predicting the thermal efficiency of fuel utilization; is the objective function of pollutant gas concentration, is the lowest predicted value of the polluted gas at the end of the tth iteration, is the standard value of pollutant gas emission, It is the second input variable for pollutant gas concentration prediction; the first input variable is the calcination temperature, oxygen content, secondary air temperature, and rotary kiln speed in the current time period; the second input variable is the calcination temperature, oxygen content, total fuel input, fuel ratio, and combustion time in the current time period.

4. The intelligent emission reduction method based on big data according to claim 1, characterized in that: The pollutant gas concentration model adopts an RNN network model; the fuel utilization thermal efficiency model adopts an LSTM network model.

5. The intelligent emission reduction method based on big data according to claim 1, characterized in that: The raw material calcination temperature prediction model is a support vector machine model based on Gaussian kernel function.

6. The intelligent emission reduction method based on big data according to claim 1, characterized in that: The control objective function is: the minimum square error between the temperature prediction value and the optimal temperature in the optimal solution set.

7. The intelligent emission reduction method based on big data according to claim 1, characterized in that: The optimal temperature value is: ;in, is the adjusted optimal temperature value at time k+1; is the optimal temperature at time k+1 obtained by the optimization control algorithm, is the error, r is the feedback coefficient, and its value range is (0, 1).

8. An intelligent emission reduction system based on big data, characterized in that: include: processor; A memory storing computer instructions for intelligent emission reduction based on big data, wherein when the computer instructions are executed by the processor, the system executes an intelligent emission reduction method based on big data according to any one of claims 1 to 7.

Images