CN107038307B - The Roller Conveying Kiln for Temperature that mechanism is combined with data predicts integrated modelling approach - Google Patents

Abstract

The invention discloses a kind of Roller Conveying Kiln for Temperature for combining mechanism with data to predict integrated modelling approach, by establishing mechanism model from the angle of temperature change and energy to the factor analysis for influencing temperature change；Then in view of roller kilns sintering is a sufficiently complex process, entire sintering process can not be described by a single mechanism model, and mechanism model is that there are model errors by simplifying, data model is established to predict model error, it is exported with this to make up mechanism, i.e., establishes the error prediction model of the nonlinear time-varying process returned based on local weighted core principle component using error as training sample；Last mechanism model is combined with data model establishes Roller Conveying Kiln for Temperature prediction integrated model.The state change of process can be preferably tracked using the model that the present invention obtains, and good directive function is provided for Roller Conveying Kiln for Temperature control, to improve production quality and qualification rate.

Description

Integrated Modeling Method for Roller Kiln Temperature Prediction Combined with Mechanism and Data

Field of study

The invention belongs to the field of roller kiln smelting, and particularly relates to an integrated modeling method for roller kiln temperature prediction

Background technique

Lithium batteries have the characteristics of high operating voltage, large specific energy, long cycle life, light weight, and less environmental pollution. They are widely used in many fields around the world, such as mobile phones, electric vehicle technology, and medical equipment power supplies. Among them, the representative battery is a lithium battery with roller kiln as the production platform and lithium cobalt oxide as the positive electrode material. The roller kiln of the sintering device for the production of cathode materials for lithium batteries is a lightweight continuous industrial kiln, which has the characteristics of low energy consumption, short firing cycle and good furnace temperature uniformity. In addition, it is also a distributed heat flow field. The system is divided into three major areas: heating section, constant temperature section and cooling section, each of which is divided into several small areas. Among them, the roller kiln temperature is one of the most critical parameters in the operation of the system. Too high or too low temperature will adversely affect the product quality. Maintaining temperature stability is a necessary condition to ensure product quality.

At present, due to the limitation of working conditions, it is difficult to obtain operating information and process parameters such as oxygen flow distribution and charge distribution inside the roller kiln, and the partial differential equations have problems such as complicated solutions and strict known conditions. It is difficult to establish an accurate mathematical model of partial differential equations from the perspective of the field. In order to obtain the temperature variation trend of the roller kiln, researchers in recent years usually use the first-order time-delay system as the model of the system to approximate the temperature according to the characteristics of the roller kiln. There is a large error in the temperature value, so it is an unsolved problem to establish a mechanism model that can be solved easily and can better estimate the temperature of the roller kiln. However, due to the complexity of the sintering process, the mechanism model cannot reflect all the factors that affect the temperature, and there will inevitably be errors between the obtained results and the actual temperature. Therefore, in order to obtain a more accurate temperature estimate, it is also necessary to predict the errors generated by the model in advance. A big problem to solve.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide an integrated modeling method for roller kiln temperature prediction based on the combination of mechanism and data, namely establishing a mechanism model from the perspective of temperature change and energy change, and then using the mechanism model to output the difference between the actual temperature and the output As a training sample, a data model is established, errors are predicted, and finally the output of the mechanism model and the data model is summed as the final temperature output, so as to solve the problems existing in the prior art.

An integrated modeling method for roller kiln temperature prediction based on the combination of mechanism and data, comprising the following steps:

1) Data processing steps: organize the operation data of the roller kiln process and store it in the created data

library;

2) Mechanism modeling based on the relationship between temperature and energy change: the material temperature cannot be measured directly. The temperature measured by the thermocouple is regarded as the temperature of the entire upper temperature zone or the lower temperature zone. Roller kiln sintering is a slow time-varying process. The point power is regarded as the power within Δt=5min, assuming that the pressure in each temperature zone is constant;

Secondly, taking the upper temperature zone of the i-th temperature zone as the research object, the factors that affect the temperature change include the following aspects: heating of silicon carbide rods, heat transfer from high temperature zone to/from low temperature zone, atmosphere bringing in/carrying out, The bowl and material are brought in/out, water evaporation absorbs heat, material chemical reaction consumes, and the kiln wall dissipates heat; according to the relationship between temperature change and energy change, formula (1) is established:

Among them, Q _i1 , Q _i2 , and Q _i3 represent the internal energy change, sensible heat change and heat transfer change of the upper temperature zone of the i-th temperature zone, respectively, and a represents the heat transfer coefficient;

According to the energy change, Q _i1 , Q _i2 , and Q _i3 are obtained, which are substituted into the formula (1) and simplified to obtain the calculation model (2):

Among them, P _i1 (t) is the heating power of the upper temperature zone of the i-th temperature zone, x _i1 , x _i2 are the temperature of the upper/lower temperature zone of the i-th temperature zone, respectively, x _(i-1)1 , x _{(i- 1) 2} represents the temperature in the upper/lower temperature zone of the i-1th temperature zone, φ(x _i1 (t)) represents the heat consumed by the chemical reaction in the upper temperature zone, and vi is the flow rate of the atmosphere introduced into the _i -th temperature zone;

In the same way, the calculation model of the lower temperature zone of the i-th temperature zone can be obtained. Finally, the two calculation models are simplified to obtain the final mechanism model (3):

Among them, x _i1 and x _i2 respectively represent the temperature of the upper and lower temperature zones of the i-th temperature zone; u _i1 =P _i1 Δt, u _i2 =P _i2 Δt respectively represent the heat generated in the heating rod Δt of the i-th temperature zone at the upper and lower temperature zones; v _i represents the flow rate of the atmosphere entering the ith temperature zone; x _(i-1)1 , x _(i-1)2 represent the temperature of the upper and lower temperature zones of the previous temperature zone of the ith temperature zone, x _{(i+ 1) The temperature of the upper and lower temperature zones in the next temperature zone after 1} , x _(i+1)2 ; φ(x _i1 ), φ(x _i2 ) represent the heat consumed by the chemical reaction in the upper/lower temperature zone of the i-th temperature zone, other are the parameters to be identified by the system;

Again, considering that the temperature of the roller kiln is constantly changing, the heat consumed by the reaction cannot be represented by a specific relational formula, but under the conditions of constant pressure, constant volume and the flow of the incoming atmosphere, the energy consumed by the chemical reaction is only the same as the current temperature. Therefore, it can be expressed as a function related to temperature. The Gaussian kernel function is used to fit the heat consumed by the chemical reaction in pieces, such as formula (4):

Among them, x _i1 , x _i2 represent the temperature of the upper/lower temperature zone of the i-th temperature zone; x′ _i1 , x′ _i2 represent the segment point temperature of the upper/lower temperature zone, and x _max , x _min represent the upper temperature of the i-th temperature zone The maximum and minimum temperature of the zone, x′ _max , x′ _min represent the maximum and minimum temperature of the temperature zone under the ith temperature zone, α _i , β _i identification coefficients.

The two calculation models are simplified to a certain extent, and the final mechanism model (5) is obtained:

Among them, x _i1 , x _i2 represent the temperature of the upper/lower temperature zone of the i-th temperature zone respectively; u _i1 =P _i1 Δt, u _i2 =P _i2 Δt respectively represent the temperature generated in the heating rod Δt of the upper and lower temperature zones of the i-th temperature zone Heat; v _i represents the flow rate of the atmosphere entering the i-th temperature zone; x _(i-1)1 , x _(i-1)2 , x _(i+1)1 , x _(i+1)2 represent the i-th The temperature of the temperature zone before and after each temperature zone; x′ _i1 , x′ _i2 represent the segment point temperature of the upper/lower temperature zone, x _max , x _min represent the maximum and minimum temperature of the upper temperature zone of the i-th temperature zone, x′ _max , _x'min represents the maximum and minimum temperature of the temperature zone under the i-th temperature zone; the others are the parameters to be identified by the system;

Finally, using the sample data in the database, the parameter identification method of least squares is used to identify the parameters of the mechanism model, and the simulation verification is carried out;

3) Temperature error prediction modeling of nonlinear time-varying process based on locally weighted kernel principal component regression: According to the high dimensionality, strong nonlinearity and process time-varying characteristics of the sintering process data of roller kiln, principal component analysis method, kernel Skills and real-time learning methods to establish a temperature error prediction model for nonlinear time-varying processes based on locally weighted kernel principal component regression;

First, classify the sample data in the database: training samples, verification samples, test samples, and obtain the weight coefficient according to the distance between each sample and the test sample, that is, the similarity, among which, the distance between the historical sample and the test sample and the specified Weight value, using formula (6):

Among them, x _i is the historical data, x _q is the test sample, d _i represents the distance between the historical sample and the test sample, σ represents the parameter that adjusts the speed of the weight changing with the distance, and _wi represents the specified weight value;

Construct the weighted training samples φ ^w (x _i ) after nonlinear high-dimensional space projection, and calculate the weighted covariance matrix (7):

Secondly, in order to extract the nonlinear part of the data and make the obtained results more reflect the reality, a Gaussian kernel function is introduced to extract the nonlinear characteristics, that is, the score vector of each sample in the projection direction is calculated, including the score vector of training samples and test samples. , as shown in formula (8):

Among them, T ^{W, K} , t _q ^{W, K} represent training sample and test sample score vector respectively, K ^w , K _q ^w represent training sample and test sample projection kernel matrix respectively, α _d ^{W, K} represent projection to d-dimensional space eigenvector of ;

Then, establish the least squares regression model between the output variable and the nonlinear feature, and calculate the formula (9):

Finally, the data in the database is used as the input, and the difference between the actual temperature and the output of the mechanism model is used as the training sample, and the model parameters are optimized and identified, and the simulation verification is carried out;

4) The integrated modeling method of roller kiln temperature prediction combining mechanism and data: First, using the sample data of the database, the least square parameter identification method is used to carry out the parameters of the mechanism model, and the model output is obtained by simulation; The input is used as the input of the data model, the difference between the actual temperature and the output of the mechanism model is used as the output of the data model, and new sample data are established to optimize the parameters of the data model, and the temperature error prediction output is obtained by simulation; The output is summed with the error prediction output obtained by the data model, and finally the temperature prediction output of the integrated model is obtained.

The present invention has the following beneficial effects:

1) Compared with the first-order time-delay system, the mechanism model established by the present invention includes more influencing factors in the sintering process, and the final result is closer to reality;

2) Considering that roller kiln sintering is a very complex process, it is impossible to describe the entire sintering process through a single mechanism model, and the mechanism model is obtained through certain simplification, so there will inevitably be model errors, but the impact that cannot be reflected in the model The relationship between factors and errors is complex and cannot be described by establishing a mathematical model through a definite relationship. In order to obtain better prediction results, the input of the mechanism model is used as the input, and the difference between the actual temperature and the output of the mechanism model is used as the training sample. , establish a data-driven model, which supplements the important influencing factors that cannot be reflected by the mechanism model, so that the actual content of the entire system is more abundant, and the final output results are more accurate;

3) Summing the output obtained by the mechanism model and the temperature error prediction output of the data model, a more accurate prediction output can be obtained. Using this model can better track the state change of the process, and provide a good guide for the temperature control of the roller kiln, thereby improving the production quality and qualification rate of the product.

Description of drawings

Fig. 1 is the principle diagram of the roller kiln temperature prediction integrated model combining the mechanism and data of the present invention;

Fig. 2 is the effect schematic diagram of the 2nd temperature zone mechanism model output of the present invention;

Fig. 3 is the effect schematic diagram of the 3rd temperature zone mechanism model output of the present invention;

4 is a schematic diagram of the effect of the output of the temperature zone error prediction data model in the second temperature zone of the present invention;

5 is a schematic diagram of the effect of the output of the temperature zone error prediction data model in the third temperature zone of the present invention;

6 is a schematic diagram of the effect of the output of the second temperature zone roller kiln temperature prediction integrated model of the present invention;

FIG. 7 is a schematic diagram of the output of the integrated model for predicting the temperature of the roller kiln in the third temperature zone of the present invention.

Detailed ways

In order to better illustrate the present invention, a preferred embodiment is hereby described in detail with the accompanying drawings, as follows:

Example 1

Step 1: Data preprocessing: pre-organize the operation data of the roller kiln process, including data showing errors, missing data, etc., and store it in the created database after finishing, which mainly includes the following data: The temperature of the upper/lower temperature zone of the i-th temperature zone of the roller kiln is x _i1 , x _i2 , the heating power of the upper/lower temperature zone is P _i1 , P _i2 , the temperature of the temperature zone before and after the i-th temperature zone is x _{(i-1) 1} , x _(i-1)2 , x _(i+1)1 , x _(i+1)2 , the flow rate of the atmosphere entering each temperature zone vi , the speed _V of the material and the bowl moving; the obtained data call A certain amount of training sample data is used to identify model parameters and establish a data error prediction model;

Step 2: First, through the in-depth analysis of the factors affecting the temperature change in the i1th temperature zone, the temperature change is mainly affected by the following three aspects

①Internal heat change:

② Sensible heat change: Q _i2 = α ₀ C _i0 v _i1 x _q0 +C _i1 m _i1 x _(i-1)1 -C _i2 v _i2 x _q1 -C _i3 m _i2 x _i1 (t)

③ Changes in heat transfer:

Q _i3 =ω _i1 (x _(i+1)1 (t)-x _i1 (t))-ω _i2 (x _i1 (t)-x _(i-1)1 (t))-ω _i3 (x _i1 (t)-x _i2 (t))-ω _i4 (x _i1 (t)-x _q2 )

Among them, the internal heat change is composed of silicon carbon rod heating, water vapor heat consumption and chemical reaction heat consumption; the sensible heat change is composed of the heat brought in by the atmosphere, materials, and the pot body and the heat taken away by the atmosphere, materials, and the pot body; Thermal change refers to the change of heat transferred from the high temperature area to the low temperature area. P _i1 (t) is the power at time t in the i1th temperature zone, ω _i1 , ω _i2 , ω _i3 , ω _i4 represent the temperature and energy conversion coefficient; x _(i-1)1 , x _(i-1)2 , x _(i+1)1 , x _(i+1)2 represent the temperature at time t in the i-1, i+1 th temperature zone upper/lower temperature zone; x _i1 , x _i2 represent the i-th temperature zone upper/lower The temperature corresponding to time t in the temperature zone, α ₀ , C ₀ , C _i0 , C _i1 , C _i2 , C _i3 , x _q0 , x _q1 , x _q2 , m _i0 , m _i1 , and m _i2 are regarded as all under certain conditions Constant; v _i1 represents the flow rate of the atmosphere entering the i-th temperature zone in the standard state for one hour. v _i2 represents the atmospheric flow rate discharged from the i-th temperature zone

Secondly, formula (1) is established according to the relationship between temperature change and energy change:

Substitute the mathematical relationship of the three into the above formula and perform certain simplification to obtain the following calculation model for the temperature of the i1th temperature zone

Among them, P _i1 (t) is the heating power of the upper temperature zone of the ith temperature zone, x _i1 , x _i2 are the temperature of the upper/lower temperature zone of the ith temperature zone, respectively, x _(i-1)1 (t), x _(i+1)1 (t) represents the temperature in the upper temperature zone of the i-1, i+1 temperature zone, v _i The flow rate of the atmosphere entering the i-th temperature zone, φ(x _i1 (t)) represents the upper temperature Heat consumed by chemical reactions in the zone.

In the same way, the following calculation model (3) can be obtained for the temperature of the i2-th temperature zone:

Among them, P _i2 (t) is the heating power of the lower temperature zone of the i-th temperature zone, x _i1 , x _i2 are the temperatures of the upper/lower temperature zone of the i-th temperature zone, respectively, x _(i-1)2 (t), x _(i+1)2 (t) represents the temperature in the upper temperature zone of the i-1, i+1 temperature zone, v _i The atmospheric flow rate of the i-th temperature zone, φ(x _i2 (t)) represents the lower temperature Heat consumed by chemical reactions in the zone.

Again, considering that the temperature of the roller kiln is constantly changing, the heat consumed by the reaction cannot be represented by a specific relational formula, but under the conditions of constant pressure, constant volume and the flow of the incoming atmosphere, the energy consumed by the chemical reaction is only the same as the current temperature. Therefore, it can be expressed as a function related to temperature. The Gaussian kernel function is used to fit the heat consumed by the chemical reaction in pieces, such as formula (4):

Among them, x _i1 , x _i2 represent the temperature of the upper/lower temperature zone of the i-th temperature zone; x′ _i1 , x′ _i2 represent the segment point temperature of the upper/lower temperature zone, and x _max , x _min represent the upper temperature of the i-th temperature zone The maximum and minimum temperature of the zone, x′ _max , x′ _min represent the maximum and minimum temperature of the temperature zone under the ith temperature zone, α _i , β _i identification coefficients.

The two calculation models are simplified to a certain extent, and the final mechanism model (5) is obtained:

Among them, x _i1 , x _i2 represent the temperature of the upper/lower temperature zone of the i-th temperature zone respectively; u _i1 =P _i1 Δt, u _i2 =P _i2 Δt respectively represent the temperature generated in the heating rod Δt of the upper and lower temperature zones of the i-th temperature zone Heat; v _i represents the flow rate of the atmosphere entering the i-th temperature zone; x _(i-1)1 , x _(i-1)2 , x _(i+1)1 , x _(i+1)2 represent the i-th The temperature of the temperature zone before and after each temperature zone; x′ _i1 , x′ _i2 represent the segment point temperature of the upper/lower temperature zone, x _max , x _min represent the maximum and minimum temperature of the upper temperature zone of the i-th temperature zone, x′ _max , _x'min represents the maximum and minimum temperature of the temperature zone under the i-th temperature zone; the others are the parameters to be identified by the system;

Finally, the sample data in the database is used to identify the parameters of the mechanism model by the least squares parameter identification method, and the simulation verification is carried out. Figures 2 and 3 are the output of the mechanism model and the actual temperature effect in the second and third temperature zones. The black curves represent the actual temperature in the upper and lower temperature zones of the 2nd and 3rd temperature zones respectively, and the red and blue curves represent the output results of the mechanism model of the upper and lower temperature zones of the 2nd and 3rd temperature zones respectively; stored in the database.

Step 3: In order to effectively compensate the temperature error generated by the mechanism model, it is necessary to establish a data model for the temperature error generated by the mechanism model. First, classify the sample data in the database: training samples, verification samples, and test samples, where the data model input includes the input of the mechanism model, the temperature of the adjacent temperature zone, and the speed of the material moving. The weight coefficient is obtained according to the distance between each sample and the test sample, that is, the similarity, wherein, the distance between the historical sample and the test sample and the specified weight value, using formula (6):

Among them, x _i is the historical data, x _q is the test sample, d _i represents the distance between the historical sample and the test sample, σ represents the parameter that adjusts the speed of the weight changing with the distance, and _wi represents the specified weight value;

Construct the weighted training samples after nonlinear high-dimensional space projection, and calculate the weighted covariance matrix (7):

Secondly, in order to extract the nonlinear part of the data and make the obtained results more reflective of the real reality, a Gaussian kernel function is introduced to extract the nonlinear characteristics, that is, the score vector of each sample in the projection direction is calculated, including the score vector of training samples and query samples. , the calculation formula (8):

Again, establish the least squares regression model between the output variable and the nonlinear feature, and calculate the formula (9):

Finally, the database sample data is used to optimize the data model parameters, and the temperature error prediction output is obtained. Figure 4 and Figure 5 are schematic diagrams of the temperature error prediction output effect in the upper temperature zone of the second and third temperature zones. The red line represents the prediction result, and the blue Represents a test sample.

Step 4: In order to obtain a more accurate temperature prediction output, the mechanism model and the data model are combined to establish an integrated model of roller kiln temperature prediction. Figure 1 shows the principle diagram of the integrated model of roller kiln temperature prediction combined with mechanism and data; The output obtained by the mechanism model and the temperature error prediction output of the data model can be summed to obtain a more accurate prediction output. Figures 6 and 7 are schematic diagrams of the output of the integrated model output of the roller kiln temperature prediction in the upper temperature zone of the second and third temperature zones. . It can be seen from the analysis of the results that the model can better track the state change of the process, and provide a good guidance for the temperature control of the roller kiln, thereby improving the production quality and qualification rate of the product.

The above disclosure is only a specific embodiment of the present application, but the present application is not limited to any changes that can be conceived by those skilled in the art, and should fall within the protection scope of the present application.

Claims (1)

1. a kind of modeling method for the Roller Conveying Kiln for Temperature prediction for combining mechanism with data, it is characterised in that including walking as follows It is rapid:

1) data processing: arranging roller kilns process operation data, be stored in created database, the data It include following data: warm area temperature x on i-th of warm area of roller kilns in library_i1, lower warm area temperature x_i2, warm area adds on i-th of warm area Thermal power P_i1, lower warm area heating power P_i2, warm area temperature is respectively x to the previous warm area of i-th of warm area up and down_(i-1)1, x_(i-1)2, warm area temperature is respectively x to the latter warm area up and down_(i+1)1,x_(i+1)2, it is passed through the atmosphere flow v of each warm area_i, material and The mobile speed V of bowl；The data obtained is as training sample data, for recognizing model parameter and data error prediction model Foundation；

2) modelling by mechanism based on temperature Yu energy variation relationship: temperature of charge can not be measured directly, and thermocouple is measured temperature It is regarded as entirely going up warm area or lower warm area temperature, roller kilns sintering is the process of slow time-varying, and sampling point power is considered as Δ t= Power in 5min, it is assumed that each warm area pressure is constant；

Secondly, the factor for influencing temperature change includes the following aspects using the upper warm area of i-th of warm area as research object: Elema heating, high-temperature region are incoming/bringing into low-temperature space heat transfer, atmosphere/takes out of, bowl and material are brought into/taken out of, moisture evaporation Heat absorption, the heat dissipation of material consumption of chemical reaction, kiln wall；According to the relationship opening relationships formula (1) of temperature change and energy variation:

Wherein Q_i1,Q_i2,Q_i3The internal energy variation, sensible heat variation and heat transfer for respectively indicating the upper warm area of i-th of warm area become Change, a indicates the coefficient of heat transfer；

Q is obtained according to energy variation_i1,Q_i2,Q_i3, it is substituted into (1) formula and is simplified, obtain computation model (2):

Wherein, P_i1It (t) is warm area heating power on i-th of warm area, x_i1,x_i2Respectively i-th of warm area up/down warm area temperature, x_(i-1)1,x_(i-1)2Indicate (i-1)-th warm area up/down warm area temperature, φ (x_i1(t)) in expression warm area consumption of chemical reaction heat Amount, v_iThe atmosphere flow that i-th of warm area is passed through；

The computation model of the lower warm area of i-th of warm area can be similarly obtained, finally carries out simplifying processing for two computation models, obtain most Whole mechanism model (3):

Wherein, x_i1,x_i2Respectively indicate i-th of warm area warm area temperature up and down；u_i1=P_i1Δt,u_i2=P_i2Δ t respectively indicates i-th The heat generated in warm area heating rod Δ t above and below a warm area；v_iIndicate the atmosphere flow that i-th of warm area is passed through；x_(i-1)1,x_(i-1)2 Indicate the temperature of the previous warm area of i-th of warm area or more warm area, x_(i+1)1,x_(i+1)2The temperature of warm area above and below the latter warm area； φ(x_i1), φ (x_i2) indicate i-th of warm area up/down warm area consumption of chemical reaction heat, it is other be system parameter to be identified；

Again, it is contemplated that Roller Conveying Kiln for Temperature constantly changes, and the heat for reacting consumption can not be by a specific relational expression come table Show, but in constant pressure, constant volume and under being passed through atmosphere flow certain condition, the energy of consumption of chemical reaction only with current warm area temperature It spends related, it is possible to be expressed as function related with temperature, be carried out using heat of the gaussian kernel function to consumption of chemical reaction Piecewise fitting, such as (4) formula:

Wherein, x_i1,x_i2Indicate i-th of warm area up/down warm area temperature；x′_i1,x′_i2Indicate up/down warm area waypoint temperature, x_max, x_minIndicate warm area maximum temperature, minimum value, x ' on i-th of warm area_max,x′_minIndicate warm area maximum temperature under i-th of warm area, most Small value, α_i, β_iRecognize coefficient；

Two computation models are carried out certain simplifying to handle, obtain final mechanism model (5):

Wherein, x_i1,x_i2Respectively indicate i-th of warm area up/down warm area temperature；u_i1=P_i1Δt,u_i2=P_i2Δ t respectively indicates i-th The heat generated in warm area heating rod Δ t above and below a warm area；v_iIndicate the atmosphere flow that i-th of warm area is passed through；x_(i-1)1, x_(i-1)2,x_(i+1)1,x_(i+1)2Indicate the temperature of i-th of warm area front and back warm area；x′_i1, x '_i2Indicate up/down warm area waypoint temperature, x_max,x_minIndicate warm area maximum temperature, minimum value, x ' on i-th of warm area_max,x′_minIndicate that warm area temperature is most under i-th of warm area Greatly, minimum value；Other is system parameter to be identified；

Finally mechanism model parameter is recognized using least-squares parameter discrimination method using the sample data of lane database, And simulating, verifying；

3) temperature error of the nonlinear time-varying process returned based on local weighted core principle component predicts modeling: according to roller kiln burning Tie process data have high-dimensional, strong nonlinearity and process time-varying characteristics, be respectively adopted Principal Component Analysis, geo-nuclear tracin4 and Instant learning method establishes the temperature error prediction model of the nonlinear time-varying process returned based on local weighted core principle component；

The sample data of lane database is classified first: training sample verifies sample, and test sample according to each sample and is surveyed The distance between sample sheet size, i.e. similarity obtain weight coefficient, wherein training sample and test sample in historical sample The distance between and specified weight value, using (6) formula:

Wherein, x_iFor historical data, x_qFor test sample, d_iIndicate the training sample in historical sample and between test sample Distance, σ indicate to adjust parameter of the weight with distance change speed, w_iIndicate specified weight value；

Construct the weighting training sample φ after the projection of non-linear higher dimensional space^w(x_i), calculate weighting covariance matrix (7):

Secondly, making the result obtained that can more react true reality to extract data non-linear partial, gaussian kernel function pair is introduced Nonlinear characteristic extracts, that is, calculates each sample in the score vector of projecting direction, including training sample and test specimens This score vector, shown in calculation formula (8):

Wherein, T^W,K,t_q ^W,KRespectively indicate training sample, test sample score vector, K^w,K_q ^wRespectively indicate training sample, test Sample projects nuclear matrix, α_d ^W,KExpression projects to the feature vector of d dimension space；

Then, the least square regression model between output variable and nonlinear characteristic is established, calculation formula (9):

Finally using the data of lane database as input, the difference of actual temperature and mechanism model output as training sample, Identification, and simulating, verifying are optimized to model parameter；

4) Roller Conveying Kiln for Temperature that mechanism is combined with data predicts integrated modelling approach: firstly, using the sample data of database, Mechanism model parameter is carried out using least-squares parameter discrimination method, emulation obtains model output；Then, with mechanism model Input is inputted as data model, using actual temperature and the difference of mechanism model output as the output of data model, is built with this It founds new sample data to optimize data model parameters, emulation obtains temperature error prediction output；Finally, by mechanism model The error prediction output that obtained output and data model obtains is summed, and the temperature prediction output of integrated model is finally obtained.