Sintering process carbon consumption adaptive weighted width echo state learning system modeling method

Abstract

The invention provides a modeling method for a carbon consumption adaptive weighted width echo state learning system in a sintering process, and relates to the field of energy conservation and consumption reduction in production in a steel sintering process. Determining process parameters which directly influence the sintering carbon consumption: BRP, BRP temperature, BTP, BTP temperature, bellows negative pressure, vertical combustion speed, trolley speed, material layer thickness, return mine, coke powder ratio, FeT content, SiO2 content, CaO content and MgO content; then establishing a sintering carbon consumption intelligent prediction model by using a self-adaptive weighted width echo state learning system; and finally, dynamically predicting the sintering carbon consumption by taking the directly related process parameters influencing the sintering carbon consumption as the input and the sintering carbon consumption as the output according to the actual production data. The intelligent prediction method for the sintering carbon consumption has the beneficial effects that the accurate dynamic prediction of the carbon consumption can be realized, the utilization rate of the carbon is improved, and a foundation is laid for realizing green manufacturing and intelligent manufacturing in the iron and steel industry.

Description

Technical Field

[0001] The invention relates to the field of energy saving and consumption reduction in steel sintering process production, and in particular to a sintering process carbon consumption adaptive weighted width echo state learning system modeling method. Background Art

[0002] At present, resource and environmental issues are increasingly constraining the realization of global sustainable development. Against this background, all countries are actively pursuing "green", "low-carbon" and "sustainable development". With the conclusion of important international agreements such as the 2030 Agenda for Sustainable Development and the Paris Agreement, the global green development trend has become a development trend. As the lifeblood of national development, industry has always received great attention from our country. The steel industry is the mainstay of the national economy, the material guarantee for national survival and development, and an important symbol of the country's economic level and comprehensive national strength. Faced with the current shortage of resources and increasingly serious environmental pollution, the steel industry urgently needs to assume the heavy responsibility of energy conservation, emission reduction and intelligent manufacturing.

[0003] The sintering process is an important part of the iron and steel metallurgical process. The quality and output of the sintered ore produced by it not only directly affect the output, quality and energy consumption of the blast furnace ironmaking process, but also play a vital role in the blast furnace ironmaking process to obtain good economic and technical indicators and technological progress. In addition, this process is also the largest energy-consuming process in the iron and steel metallurgical process except for the blast furnace ironmaking process. Its energy consumption accounts for about 10%-15% of the total energy consumption of iron and steel metallurgy. Among its energy consumption, coke fuel consumption is about 80%, electricity consumption is about 13.5%, gas consumption is about 6%, and other is about 0.5%. It can be seen that in order to improve energy utilization, reduce pollution emissions, and achieve green manufacturing, it is necessary to achieve accurate dynamic prediction of sintering carbon consumption, which is also a key issue to improve energy utilization in the steel industry, reduce pollution emissions, and achieve green manufacturing and intelligent manufacturing. Summary of the invention

[0004] In order to solve the problems of serious pollution and low energy utilization in the sintering process, the present invention provides a modeling method of a carbon consumption adaptive weighted width echo state learning system in the sintering process, which mainly includes the following steps:

[0005] S1: Based on the sintering process mechanism analysis and data correlation analysis, the process parameters that directly affect the sintering carbon consumption are determined. The process parameters include: BRP, BRP temperature, BTP, BTP temperature, wind box negative pressure, vertical combustion speed, trolley speed, material layer thickness, return ore, coke powder ratio, FeT content, SiO2 content, CaO content and MgO content;

[0006] S2: According to the adaptive weighted width echo state learning system, an intelligent prediction model for sintering carbon consumption is established and trained;

[0007] S3: Based on the actual production data, the process parameters that directly affect the sintering carbon consumption are taken as input, and the sintering carbon consumption is taken as output. The trained sintering carbon consumption intelligent prediction model is used to dynamically predict the sintering carbon consumption.

[0008] A storage device stores instructions and data for implementing the sintering process carbon consumption adaptive weighted width echo state learning system modeling method.

[0009] A sintering process carbon consumption adaptive weighted width echo state learning system modeling device, comprising: a processor and a storage device; the processor loads and executes instructions and data in the storage device to implement the sintering process carbon consumption adaptive weighted width echo state learning system modeling method.

[0010] The technical solution provided by the present invention has the following beneficial effects: based on the analysis of the sintering process mechanism and data correlation analysis, the process parameters that directly affect the sintering carbon consumption can be determined: BRP, BRP temperature, BTP, BTP temperature, bellows negative pressure, vertical combustion speed, trolley speed, material layer thickness, return ore, coke powder ratio, FeT content, SiO2 content, CaO content and MgO content; according to the adaptive weighted width echo state learning system, an intelligent prediction model for sintering carbon consumption is established; according to actual production data, the process parameters that directly affect the sintering carbon consumption are used as input, and the sintering carbon consumption is used as output, and the trained sintering carbon consumption intelligent prediction model is used to dynamically predict the sintering carbon consumption. In this way, the dynamic characteristics of the sintering process can be captured in real time, and accurate dynamic prediction of sintering carbon consumption can be achieved, thereby laying the foundation for realizing green manufacturing and intelligent manufacturing in the steel industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0012] Figure 1 It is a flow chart of a method for modeling a sintering process carbon consumption adaptive weighted width echo state learning system in an embodiment of the present invention;

[0013] Figure 2 This is a principle block diagram of the intelligent prediction model for sintering carbon consumption in an embodiment of the present invention;

[0014] Figure 3 is a comparison chart of the dynamic prediction result of sintering carbon consumption and the actual value in the embodiment of the present invention;

[0015] Figure 4 is a prediction error result diagram of the dynamic prediction result of sintering carbon consumption in an embodiment of the present invention;

[0016] Figure 5 It is a schematic diagram of the operation of the hardware device in the embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0018] Example 1

[0019] Please refer to Figure 1 , Figure 1 1 is a flow chart of a method for modeling a sintering process carbon consumption adaptive weighted width echo state learning system in an embodiment of the present invention, and the specific steps are as follows:

[0020] S1: The sintering process is a kind of production process with complex chemical and physical changes. The main factor affecting the sintering carbon consumption is the sintering thermal state. Based on the analysis of the sintering process mechanism and data correlation analysis, the process parameters that directly affect the sintering carbon consumption are determined. The process parameters that affect the sintering carbon consumption include the sintering ore output and carbon consumption. The sintering ore output is determined by the raw material parameters in the sintering process. The raw material parameters include coke powder ratio, SiO2 content, TFe content, MgO content, CaO content and return ore. The raw material parameters determine the amount of sintered ore and the amount of coke in the sintering process. The carbon consumption includes the state parameters and operating parameters in the sintering process. The state parameters and operating parameters can reflect the thermal state of the sintering process. The operating parameters include the trolley speed and the material layer thickness. The state parameters include: BTP, BTP temperature, bellows negative pressure, vertical combustion speed, BRP and BRP temperature, where BTP represents the sintering end point and BRP represents the rising point.

[0021] Therefore, the process parameters include: BRP, BRP temperature, BTP, BTP temperature, wind box negative pressure, vertical combustion speed, trolley speed, material layer thickness, return ore, coke powder ratio, FeT content, SiO2 content, CaO content and MgO content; these process parameters are used as inputs of the sintering carbon consumption intelligent prediction model.

[0022] S2: Based on the adaptive weighted width echo state learning system, an intelligent prediction model for sintering carbon consumption is established; please refer to Figure 2 , Figure 2 This is a principle block diagram of the sintering carbon consumption intelligent prediction model in this embodiment. The sintering carbon consumption intelligent prediction model includes a characteristic node layer and a reservoir layer, and its structure can better reflect the dynamic characteristics of the sintering process. The proposed adaptive weighted width echo state learning system can obtain the dynamic characteristics of the production process in real time, laying the foundation for establishing an accurate and effective sintering carbon consumption intelligent prediction model.

[0023] The process of establishing an intelligent prediction model for sintering carbon consumption is as follows:

[0024] S2.1: The training sample data set consisting of process parameters that directly affect sintering carbon consumption and the corresponding sintering carbon consumption is {(x t ,Y t ),t=1,2,...,N},x t is the tth training sample, which is used as the input of the sintering carbon consumption intelligent prediction model. t is the t-th sintering carbon consumption, N represents the total number of samples in the training sample data set; input data set Randomly map X to n sets of features, and the feature F of the i-th mapping i Denoted as:

[0025]

[0026] Among them, φ() is the activation function; is the weight vector; is the deviation vector; n represents the number of feature mappings, and its value is a positive integer greater than or equal to 1; the input data set X is transformed through n feature mappings to obtain:

[0027] F=[F1,F2,...,F n ]

[0028] S2.2: In order to capture the dynamic characteristics of the system, F is transferred to the reservoir through nonlinear transformation, and the reservoir state update equation is expressed as:

[0029]

[0030] in, is the hyperbolic tangent function, W in and W back Learning stochastic generation via echo state networks.

[0031] S2.3: The output of the reservoir layer is combined with all the outputs of the feature node layer:

[0032] A=[F|E]

[0033] S2.4: Assuming that the weights in the output layer are recorded as W, we can consider the following optimization problem to find its solution:

[0034] arg min:||θ(AW-Y)|| 2 +λ||W|| 2

[0035] Among them, θ is a diagonal matrix associated with each input sample and λ represents the trade-off parameter. At the same time,

[0036] θ=diag(ω t ),t=1,2,...,N

[0037] Among them, ω t is a weighted finite element, diag() represents a diagonal matrix, ω t =g(r t ), g() represents the residual probability density function, r t represents the residual of the tth sample.

[0038] Based on kernel density estimation, the residual probability density function is expressed as:

[0039]

[0040] in, represents the width of the estimation window, Ker(·) represents the kernel function, and in this embodiment, the kernel function is selected as a Gaussian function.

[0041] S2.5: This embodiment uses a pseudo-inverse algorithm to solve the weights in the output layer, and obtains:

[0042] W=(λI+A T θ 2 A) -1 A T θ 2 Y

[0043] Where I is the identity matrix.

[0044] S2.6: When W is obtained, the model training process is completed and no fine-tuning is required. The sintering carbon consumption intelligent prediction model is formed, and the predicted value of the sintering carbon consumption intelligent prediction model is recorded as

[0045]

[0046] S2.7: When the sintering carbon consumption intelligent prediction model network is trained, the output is used as feedback for residual calculation to improve the accuracy of the sintering carbon consumption intelligent prediction model:

[0047] R=Y-AW

[0048] Where R = [r1, r2, ..., r N ] T represents the residual matrix, r t represents the residual of the t-th sample, t=1,2,...,N.

[0049] The process parameter samples in the test data set in step S2 are input into the sintering carbon consumption intelligent prediction model to obtain the predicted value of sintering carbon consumption, and the predicted value of sintering carbon consumption is compared with the sample value of sintering carbon consumption corresponding to the sintering parameter sample to obtain the following: Figure 3 The results shown by Figure 3It can be seen that the predicted value of sintering carbon consumption is very close to the sample value. The sample value is the actual value of sintering carbon consumption corresponding to the actual production process. Figure 3 The prediction error between the predicted value of sintering carbon consumption and the sample value is as follows: Figure 4 As shown by Figure 3 and Figure 4 It can be seen that the prediction error of the dynamic prediction of sintering carbon consumption is within [-2,3] kg / t, and most of the prediction results are within [-2,2] kg / t. Therefore, the sintering carbon consumption intelligent prediction model can accurately perform dynamic prediction of sintering carbon consumption, meet the production requirements of the actual sintering process, and is conducive to promoting green manufacturing and intelligent manufacturing of the sintering process.

[0050] S3: According to the actual production data of sintering carbon consumption, the process parameters that directly affect the sintering carbon consumption (BRP, BRP temperature, BTP, BTP temperature, bellows negative pressure, vertical combustion speed, trolley speed, material layer thickness, return ore, coke powder ratio, FeT content, SiO2 content, CaO content and MgO content) are taken as input, and the sintering carbon consumption is taken as output. The sintering carbon consumption is dynamically predicted using the sintering carbon consumption intelligent prediction model.

[0051] Example 2

[0052] A sintering process carbon consumption adaptive weighted width echo state learning system modeling device 501, such as Figure 5 As shown, it includes: a processor 502 and a storage device 503; the processor 502 loads and executes the instructions and data in the storage device 503 to implement the sintering process carbon consumption adaptive weighted width echo state learning system modeling method.

[0053] Example 3

[0054] A storage device stores instructions and data for implementing the sintering process carbon consumption adaptive weighted width echo state learning system modeling method.

[0055] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A modeling method for a carbon consumption adaptive weighted width echo state learning system in a sintering process, characterized by: Includes the following step: S1: Based on the sintering process mechanism analysis and data correlation analysis, the process parameters that directly affect the sintering carbon consumption are determined. The process parameters include: BRP, BRP temperature, BTP, BTP temperature, wind box negative pressure, vertical combustion speed, trolley speed, material layer thickness, return ore, coke powder ratio, FeT content, SiO2 content, CaO content and MgO content; S2: establishing and training a sintering carbon consumption intelligent prediction model according to an adaptive weighted width echo state learning system; the sintering carbon consumption intelligent prediction model includes a characteristic node layer and a reservoir; The process of establishing an intelligent prediction model for sintering carbon consumption is as follows: S2.1: The training sample data set consisting of process parameters that directly affect sintering carbon consumption and the corresponding sintering carbon consumption is {(x t ,Y t ),t=1,2,...,N},x t is the tth training sample, which is used as the input of the sintering carbon consumption intelligent prediction model. t is the t-th sintering carbon consumption, N represents the total number of samples in the training sample data set; input data set Randomly map X to n sets of features, and the feature F of the i-th mapping i Denoted as: Among them, φ( ) is the activation function; is the weight vector; is the bias vector; n represents the number of feature mappings; After the input data set X is transformed by n feature maps, we get: F=[F1,F2,...,F n ] Where F represents the mapping feature vector output by the feature node layer; S2.2: In order to capture the dynamic characteristics of the adaptive weighted width echo state learning system, F is transferred to the reservoir through nonlinear transformation, and the reservoir state update equation is expressed as: in, is the hyperbolic tangent function, W in and W back Random generation through echo state network learning; E(n) represents the reservoir state of the nth feature map, F(n) represents the feature of the nth map, and E(n-1) represents the reservoir state of the n-1th feature map; S2.3: The output of the reservoir layer is combined with all the outputs of the feature node layer: A=[F|E] Among them, E represents the output of the reservoir; F represents the output of the feature node layer, that is, the mapping feature vector; S2.4: Solve the optimization problem of the system: <h2 style=";text-align:left;direction:ltr">argmin:||θ(AW-Y)||<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> +λ··W||<h2 style=";text-align:left;direction:ltr"> 2 Among them, θ is the diagonal matrix associated with each input sample, λ represents the trade-off parameter, W is the weight in the output layer, and Y represents the sintering carbon consumption; S2.5: When W is obtained, the adaptive weighted width echo state learning system training is completed, forming a sintering carbon consumption intelligent prediction model: in, It indicates the predicted value of the sintering carbon consumption intelligent prediction model; S3: Based on the actual production data, the process parameters that directly affect the sintering carbon consumption are taken as input, and the sintering carbon consumption is taken as output. The trained sintering carbon consumption intelligent prediction model is used to dynamically predict the sintering carbon consumption.

2. A sintering process carbon consumption adaptive weighted width echo state learning system modeling method as claimed in claim 1, characterized in that: In step S1, according to the analysis of the sintering process mechanism, the process parameters affecting the sintering carbon consumption include two aspects: sintering ore output and carbon consumption; the sintering ore output is determined by the raw material parameters in the sintering process, and the raw material parameters include coke powder ratio, SiO2 content, TFe content, MgO content, CaO content and returned ore; the carbon consumption includes state parameters and operating parameters in the sintering process, and the operating parameters include trolley speed and material layer thickness, and the state parameters include: BTP, BTP temperature, wind box negative pressure, vertical combustion speed, BRP and BRP temperature, wherein BTP represents the sintering end point and BRP represents the rising point.

3. A sintering process carbon consumption adaptive weighted width echo state learning system modeling method as claimed in claim 1, characterized in that: In step S2.4, the formula for calculating the diagonal matrix θ associated with each input sample and is: θ=diag(ω t ),t=1,2,...,N Among them, ω t is a weighted finite element, diag() represents a diagonal matrix, and N represents the total number of samples in the training sample data set.

4. A sintering process carbon consumption adaptive weighted width echo state learning system modeling method as claimed in claim 3, characterized in that: Weighted Finite Element ω t The calculation formula is: oh t =g(r t ) Among them, g() represents the residual probability density function, r t represents the residual of the tth sample.

5. A sintering process carbon consumption adaptive weighted width echo state learning system modeling method as claimed in claim 4, characterized in that: The residual probability density function is: in, represents the width of the estimation window, x represents the input data, and Ker(·) represents the kernel function.

6. A sintering process carbon consumption adaptive weighted width echo state learning system modeling method as claimed in claim 1, characterized in that: Use the pseudo-inverse algorithm to solve the parameters in the optimization problem of step S2.4 and obtain: W=(λI+A T i 2 A) -1 A T i 2 Y Where I is the identity matrix.

7. A sintering process carbon consumption adaptive weighted width echo state learning system modeling method as claimed in claim 1, characterized in that: In step S2, when the sintering carbon consumption intelligent prediction model is trained, the model output is used as feedback for residual calculation to improve the accuracy of the sintering carbon consumption intelligent prediction model: R=Y-AW Where R = [r1, r2, ..., r N ] T represents the residual matrix, r t represents the residual of the t-th sample, t=1,2,...,N.

8. A storage device, characterized in that: The storage device stores instructions and data for implementing the sintering process carbon consumption adaptive weighted width echo state learning system modeling method as described in any one of claims 1 to 7.

9. A sintering process carbon consumption adaptive weighted width echo state learning system modeling device, characterized by: include: Processor and storage device; the processor loads and executes instructions and data in the storage device to implement the sintering process carbon consumption adaptive weighted width echo state learning system modeling method as described in any one of claims 1 to 7.

Images