CN108764517B - Method, equipment and storage medium for predicting change trend of silicon content in molten iron of blast furnace - Google Patents
Method, equipment and storage medium for predicting change trend of silicon content in molten iron of blast furnace Download PDFInfo
- Publication number
- CN108764517B CN108764517B CN201810305865.7A CN201810305865A CN108764517B CN 108764517 B CN108764517 B CN 108764517B CN 201810305865 A CN201810305865 A CN 201810305865A CN 108764517 B CN108764517 B CN 108764517B
- Authority
- CN
- China
- Prior art keywords
- silicon content
- blast furnace
- molten iron
- trend
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Operations Research (AREA)
- Human Resources & Organizations (AREA)
- Computational Mathematics (AREA)
- Pure & Applied Mathematics (AREA)
- Economics (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Evolutionary Biology (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Algebra (AREA)
- Development Economics (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Manufacture Of Iron (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a method, equipment and a storage medium for predicting the change trend of the silicon content of molten iron in a blast furnace, wherein the method comprises the following steps: acquiring characteristic parameters related to the change of the silicon content in the blast furnace molten iron in the iron-making process, and predicting the change trend of the silicon content based on a trained prediction model of the silicon content in the blast furnace molten iron; the blast furnace molten iron content prediction model comprises a first layer prediction model and a second layer prediction model, the first layer prediction model is used for primarily predicting the silicon content change trend according to an original sample, and the second layer prediction model carries out secondary prediction according to the prediction result of the first layer prediction model to obtain the silicon content change trend. A fusion model of extreme gradient promotion and a long-short term memory network is established to predict the trend of the silicon content in the molten iron, and a reference basis is provided for a blast furnace operator to judge the change trend of the furnace condition and the regulation and control amplitude in advance, so that the smooth operation of the iron-making process and the quality of the molten iron are kept in a normal range.
Description
Technical Field
The invention relates to the technical field of blast furnace smelting automation control, in particular to a method, equipment and a storage medium for predicting the change trend of the silicon content of molten iron in a blast furnace.
Background
Blast furnace smelting is a continuous production process, the operation condition in the furnace is complex, and various physical and chemical reactions are numerous. The temperature of the blast furnace, i.e. the temperature of molten iron in the furnace hearth, is one of the important parameters for measuring the furnace condition of the blast furnace, and over-high and over-low both cause the quality of the molten iron to be reduced and the furnace condition to be unsmooth. The furnace temperature is low, the physical heat of molten iron is insufficient, the heat storage of a furnace hearth is insufficient, and the freezing accident of the furnace hearth is easily caused; the furnace temperature is high, and the gas flow in the furnace hearth is too vigorous, so that furnace condition faults such as suspension, material collapse and the like and waste of raw materials are caused. Therefore, the realization of the online prediction of the furnace temperature has great significance for ensuring the safe, stable and smooth operation of the blast furnace.
Because the closed production process of the blast furnace cannot directly detect the temperature of molten iron in the hearth, the temperature measured by the thermocouple at the skimmer has certain heat loss and cannot accurately represent the temperature of the blast furnace. Research shows that in the process of flowing molten iron from the hearth to the skimmer, information of the silicon content of the molten iron is not lost, and the silicon content of the molten iron has a strong positive correlation with the temperature of the molten iron in the hearth, so that the problem of forecasting the furnace temperature can be converted into the forecasting of the silicon content of the molten iron in the blast furnace. In the actual production process, however, a blast furnace operator does not frequently adjust parameters due to the change of the silicon content values of the previous and subsequent times, and only when the change trend of the silicon content exceeds a certain allowable range, a corresponding regulation and control means is adopted, so that the prediction of the change trend of the silicon content of the molten iron of the blast furnace has great significance for the fine regulation and control of the furnace temperature of the blast furnace on site from the aspect of providing guidance on site. Particularly, under the conditions of large-scale customized production of steel and increasingly degraded raw materials required by iron making in China, the external environment for blast furnace iron making presents obvious uncertainty, the variation trend of the molten iron silicon content of the blast furnace presents obvious high dynamics, and the stability control difficulty of the blast furnace is obviously increased, so that the real-time and accurate prediction of the trend of the molten iron silicon content has important significance for stabilizing the furnace condition of the blast furnace and ensuring the smooth operation of the blast furnace.
The forecasting of the silicon content trend of the molten iron is mainly realized by establishing a mechanism and a data model, the accuracy of the mechanism model forecasting is high, the application range is wide, the relation between the silicon content trend in the blast furnace and detection parameters in the iron-making process can be intuitively reflected, the accurate mechanism model is time-consuming and labor-consuming to establish, the modeling condition requirements are strict, and the parameters required by the model cannot be accurately obtained. Therefore, the establishment of an intelligent forecasting model based on data driving becomes another good method, and the forecasting model between the detection parameters and the silicon content trend in the iron-making process is established by utilizing rich online and offline historical data in the iron-making process, so that the online, real-time and accurate forecasting of the silicon content trend is realized. The Chinese patent of the published patent No. CN105574297A provides a self-adaptive blast furnace molten iron silicon content trend forecasting method, which is based on an online Least square Support Vector machine (LS-SVMS) model, establishes a self-adaptive predictor based on the online Least square Support Vector machine (LS-SVMS) model, performs self-adaptive updating on the trend forecasting model by continuously acquiring new samples, tracks the dynamic change of the blast furnace smelting process, and has good real-time performance and reliability; however, the data acquired by the content of silicon in molten iron of the blast furnace is converted into the trend which is determined by the difference value of the content of silicon in the front and the back furnaces, wherein the difference value is positive and represents the rising trend, and the difference value is negative and represents the falling trend. The period of forecasting the trend is short, the trend division is not comprehensive enough, and only the direction of the silicon content variation trend can be provided, but the information of the variation amplitude cannot be provided. The invention discloses a Chinese patent publication No. CN104899463A which discloses a four-classification trend forecasting model of the silicon content of molten iron of a blast furnace, wherein rich online off-line data acquired on site of the blast furnace are utilized, samples of the silicon content of the molten iron are effectively clustered according to a fuzzy C-means clustering method, a division standard of four-classification change trend intervals of model output variables is obtained, and the four-classification trend forecasting model is established by utilizing an extreme learning machine. However, the extreme learning machine used by the method is a single hidden layer forward neural network, the input node and hidden node weights are given in a random mode so as to ensure the rapidity of the model, but the output weights are obtained on the basis of the input weights and hidden node weights, and the accuracy of the model cannot be ensured because the output weights are not optimal. In summary, the existing silicon content trend pre-method mainly has two defects, one is that the trend of the silicon content is defined by the difference between the front and back furnaces, and the division method is rough and cannot reflect comprehensive information; secondly, the accuracy of the model used for forecasting is not guaranteed.
Disclosure of Invention
The invention provides a method, equipment and a storage medium for predicting the silicon content variation trend of molten iron of a blast furnace, which overcome the problems or at least partially solve the problems, and solves the problems that in the silicon content trend prediction method in the prior art, the trend of the silicon content is only defined by the difference between the front and back furnaces, the division method is rough in division and cannot reflect comprehensive information, and the accuracy of a model for prediction cannot be guaranteed.
According to one aspect of the invention, the method for predicting the change trend of the silicon content in the molten iron of the blast furnace comprises the following steps:
acquiring characteristic parameters related to the change of the silicon content in the blast furnace molten iron in the iron-making process, and predicting the change trend of the silicon content based on a trained prediction model of the silicon content in the blast furnace molten iron;
the blast furnace molten iron content prediction model comprises a first layer prediction model and a second layer prediction model, the first layer prediction model is used for primarily predicting the silicon content change trend according to an original sample, and the second layer prediction model carries out secondary prediction according to the prediction result of the first layer prediction model to obtain the silicon content change trend.
A blast furnace molten iron silicon content variation trend prediction device comprises:
at least one processor, at least one memory, a communication interface, and a bus; wherein the content of the first and second substances,
the processor, the memory and the communication interface complete mutual communication through the bus;
the communication interface is used for information transmission between the test equipment and the communication equipment of the display device;
the memory stores program instructions which can be executed by the processor, and the processor calls the program instructions to execute the method for predicting the change trend of the silicon content of the blast furnace molten iron.
A blast furnace molten iron silicon content variation trend prediction device comprises:
at least one processor; and
at least one memory communicatively coupled to the processor, wherein:
the memory stores program instructions which can be executed by the processor, and the processor calls the program instructions to execute the method for predicting the change trend of the silicon content of the blast furnace molten iron.
A non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the method for predicting the trend of change in the silicon content of molten iron in a blast furnace as described above.
The invention provides a method, equipment and a storage medium for predicting the change trend of the silicon content of molten iron in a blast furnace, wherein the trend of the silicon content of the molten iron in the blast furnace is predicted by a data driving method, an input sample is constructed by data preprocessing, and the output trend category is determined by a method of regression linear fitting on the silicon content value in a sliding window; a fusion model based on extreme gradient improvement and a long-term and short-term memory network is established to predict the trend of the silicon content in the molten iron; the method has the advantages that data which can be detected by the blast furnace and influence the silicon content of the molten iron are fully utilized to predict the direction and the amplitude of the trend of the silicon content of the molten iron in a half iron-making shift, reference basis is provided for blast furnace operators to judge the change trend and the regulation and control amplitude of the furnace conditions in advance, and accordingly smooth operation of the iron-making process and maintenance of the quality of the molten iron in a normal range are guaranteed.
Drawings
FIG. 1 is a schematic diagram illustrating a method for predicting the variation trend of the silicon content in molten iron in a blast furnace according to an embodiment of the present invention;
FIG. 2 is a flow chart of silicon content trend division according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of multi-model fusion according to an embodiment of the present invention;
FIG. 4(a) is a hit chart of the trend prediction result of partial Si content in the practical case according to the embodiment of the present invention;
FIG. 4(b) is a hit chart of the trend prediction result of silicon content in another part of the practical case according to the embodiment of the present invention;
FIG. 5 is a schematic diagram of a device for predicting the variation trend of the silicon content in molten iron in a blast furnace according to an embodiment of the present invention.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
The embodiment shows a method for predicting the variation trend of the silicon content in molten iron of a blast furnace, as shown in fig. 1, the method includes:
acquiring characteristic parameters related to the change of the silicon content in the blast furnace molten iron in the iron-making process, and predicting the change trend of the silicon content based on a trained prediction model of the silicon content in the blast furnace molten iron;
the blast furnace molten iron content prediction model comprises a first layer prediction model and a second layer prediction model, the first layer prediction model is used for primarily predicting the silicon content change trend according to an original sample, and the second layer prediction model carries out secondary prediction according to the prediction result of the first layer prediction model to obtain the silicon content change trend.
In this embodiment, the variation trend of the silicon content in the molten iron is closely related to the variation of a plurality of parameters in the whole iron making process, and the parameters can be divided into two types: one type is partially measurable and can reflect the state parameters of the smelting process state of the blast furnace; the other is a control parameter for field workers to regulate and control when the furnace condition of the blast furnace is abnormal. The method mainly comprises the steps of oxygen enrichment rate, air permeability index, standard wind speed, oxygen enrichment flow, cold wind flow, coal injection quantity, molten iron component, coal gas component, top pressure, total pressure difference, hot wind pressure, actual wind speed, cold wind pressure, theoretical combustion temperature, hot wind temperature, blowing kinetic energy, oxygen enrichment flow, oxygen enrichment pressure, coal gas quantity of a furnace belly and the like.
In the embodiment, a first layer of prediction model is established through a limit gradient lifting model and a variant long-short term memory model of a recurrent neural network, and characteristic parameter data are input into the limit gradient lifting model to obtain the prediction result of the model and the importance ranking of the characteristic parameters influencing the silicon content trend;
selecting characteristic parameter data with high importance from the importance sequence, and inputting the characteristic parameter data into the long-term and short-term memory model to obtain a prediction result of the model;
and establishing a fusion model of the extreme gradient lifting model and the long-term and short-term memory model to obtain a prediction result of the seven-classification trend of the silicon content.
The historical database of the blast furnace stores a large amount of historical data, wherein some data are collected in real time on line, and some data are manually input by blast furnace operators after testing. In the process of collecting data, due to various reasons such as equipment failure or manual operation errors, the data can not truly and accurately reflect the furnace condition, so that the data needs to be preprocessed to eliminate the problems and improve the quality of the data and the precision of a subsequent model.
Therefore, in this embodiment, before obtaining the characteristic parameter related to the change of the silicon content in the blast furnace molten iron during the iron making process, the method further includes:
acquiring characteristic parameters and silicon content change trends related to the change of the silicon content in the blast furnace molten iron in the historical ironmaking process, and training a blast furnace molten iron content prediction model;
and training the first layer prediction model according to the characteristic parameters and the silicon content variation trend, and training the second layer prediction model according to the prediction result of the trained first layer prediction model.
Acquiring field data in the iron-making process, and performing abnormal value elimination, missing value filling and normalization processing on the characteristic parameters to obtain high-quality data required by modeling;
and (4) dividing the silicon content in the blast furnace molten iron on a time axis through a sliding window to obtain the variation trend of the silicon content in the blast furnace molten iron.
Specifically, the method comprises the following steps:
1) abnormal value processing
When the blast furnace is in failure, or due to abnormal furnace conditions, the collected characteristic parameter data can deviate from the normal range. So multiple outliers are required for culling. Firstly, the obvious abnormal data in the sequence are directly removed, and the unobvious data are directly removed through a box-line graph method. The box diagram is drawn only by means of actual data, the data does not need to be assumed to be subject to a specific distribution form in advance, no limiting requirement is made on the data, and the result of abnormal value identification is objective.
2) Missing value filling
The abnormal values deleted in the step 1), manual errors, damping down, equipment faults and other reasons can cause data loss. If the missing data is not deleted or filled reasonably, the accuracy of the model is influenced to a certain extent.
The direct deletion of all missing values inevitably loses a part of the information, and therefore selective deletion and padding of the missing values are required. Firstly, counting the number of attribute missing values of each sample according to rows, sequencing the number of the missing values from small to large, and taking the serial number as an abscissa and the number of the missing values as an ordinate. And respectively drawing a scatter diagram for counting the number of missing values of the training set and the test set. From the perspective of the model, the more consistent the distribution of the training set and the test set is, the more the occurrence of the overfitting phenomenon can be prevented, the places with inconsistent distribution in the scatter diagram are found out, and the points can be regarded as outliers, and the whole row of data is removed.
The remaining missing value should be filled up, and if the data at the first time is missing, the data x at the time immediately before the data is usedimAnd data x at the subsequent timeinAnd (4) estimating:
3) normalization processing
Because the dimension difference of the data collected on site is large, the influence of the large dimension on the model is large, and therefore the data needs to be normalized before modeling, and the data is normalized by the following formula:
xnormdenotes the result after normalization of the variables, xmax，xminRespectively representing the maximum value and the minimum value of the ith variable.
And (4) dividing the silicon content in the blast furnace molten iron on a time axis through a sliding window to obtain the variation trend of the silicon content in the blast furnace molten iron.
In the actual smelting operation process, the work length does not frequently adjust corresponding control parameters due to small fluctuation of the silicon content of the molten iron in each shift (8 hours). Only when the furnace condition deviates from the normal range, the operator can regulate and control the control parameters such as blast kinetic energy, hot air temperature, coal injection quantity and the like. Therefore, the prediction period is too short, the guidance significance to field workers is not great, the prediction period is too long, and the furnace condition is abnormal and cannot be regulated and controlled in time. Generally, each parameter has different lag time, for example, the air supply has 2-3 hours lag to the silicon content of the molten iron, and the coal injection amount has 3-4 hours lag to the silicon content of the molten iron, so that the regulation and control result can be correspondingly embodied after several hours. And according to the experience of field experts, the prediction of the trend of half a shift has great significance for field guidance. In summary, in this embodiment, a time period of half a shift (4 hours) is selected as a period for analyzing the prediction of the silicon content trend of the molten iron, so as to provide the change trend of the silicon content in the next 4 hours for the operating staff of the blast furnace, which is beneficial for the operating staff to comprehensively grasp the change direction of the furnace condition in the future, and to adopt a proper regulation and control means timely to ensure the stable and smooth operation of the blast furnace.
From the time sequence chart of the silicon content in one month, it can be seen that the value of the silicon content in the molten iron has no definite periodicity and clustering, and it is difficult to directly mine trend information on the original data. Therefore, the value of the silicon content in the molten iron can be divided on a time axis by using a sliding window, the equation of the value is obtained by adopting a polynomial regression fitting method for data in the sliding window, and the trend information of the silicon content is determined by fitting the positive and negative of the first and second derivatives of the equation. When the first derivative is 0, a steady and constant trend is represented; when the first derivative is a positive number, the linear ascending trend is shown; when the first derivative is negative, it shows a linear downward trend. The second derivative reflects the variable shape, and the second derivative is convex when being a positive number; the second derivative is negative and concave. The trends can be divided into 7 types according to the combination of the first and second derivatives: concave ascending, linear ascending, convex ascending, stable and unchanged, concave descending, linear descending and convex descending. The overall flow chart of the silicon content trend division is shown in fig. 2, and the specific steps are as follows:
step1, setting the width of a total window, determining the prediction period of the silicon content change in the blast furnace molten iron according to the number of ironmaking shifts, and setting the width of a sliding window according to the prediction period; in this example, the total window width M is determined, i.e. there are M data ((x) in the total time series of the problem under study1,y1),(x2,y2)...(xM,yM) Determining the width N of the sliding window to be 8(N < M) according to the prediction period;
step2, fitting the silicon content value in each sliding window through a zeroth order polynomial to obtain a fitting equation, judging the fitting effect through variance ratio test (F-test, F significance test), if the fitting equation is not significant, recording that the silicon content in the sliding window is in a stable and unchanged trend, otherwise, turning to Step 3;
step 3: fitting the silicon content value in the sliding window through a first-order polynomial to obtain a fitting equation, judging the fitting effect through F significance test, if the fitting equation is not significant, judging that the silicon content in the sliding window is in a linear ascending trend or a linear descending trend according to the positive and negative slope, otherwise, switching to Step 4;
step4, fitting the silicon content value in the sliding window through a second-order polynomial to obtain a fitting equation, solving a first derivative and a second derivative of the fitting equation, and if the first derivative is positive and the second derivative is positive, judging that the silicon content in the sliding window is in a convex rising trend; if the first derivative is positive and the second derivative is negative, judging that the silicon content in the sliding window is in a concave ascending trend; if the first derivative is negative and the second derivative is positive, judging that the silicon content in the sliding window is in a convex descending trend; if the first derivative is negative and the second derivative is negative, judging that the silicon content in the sliding window is in a concave descending trend;
and Step5, operating the data in each sliding window through the steps to obtain the silicon content variation trend in each sliding window, and finishing the condition that the data in the total window are fitted.
And training the first layer prediction model according to the characteristic parameters and the silicon content variation trend, and training the second layer prediction model according to the prediction result of the trained first layer prediction model.
In this embodiment, the first layer prediction model includes a extreme gradient boost model and a variant long-short term memory model of a recurrent neural network;
the first layer prediction model is trained by original samples, and comprises the following steps:
training a limit gradient lifting model through the characteristic parameters and the silicon content variation trend to obtain the trained limit gradient lifting model and the importance ranking of process variable parameters influencing the silicon content variation trend;
training a plurality of process variable characteristic parameters ranked at the top in the process variable parameter importance ranking based on the long-short term memory model, in the embodiment, selecting the characteristic parameters which affect the characteristic importance ranking of the silicon content trend, and inputting the characteristic parameters into the long-short term memory model (LSTM) to predict the trend of the molten iron silicon content.
A limit gradient boost model (XGBoost) is selected as feasibility analysis of a prediction model, from the data perspective, a plurality of blast furnace detection devices are provided, the obtained actual production data volume is large, and the data has diversity and hierarchy. XGboost adopts parallel and distributed computation, so that the learning speed of the model is accelerated. From the perspective of a model, XGboost is an effective high-performance implementation of a Machine learning algorithm (GBM) class algorithm, is not high in computational complexity, has a large number of applications in the industry, and has significant advantages in terms of computational performance, accuracy, and runtime in classification, regression, and ranking problems.
XGboost achieves accurate prediction results through iterative computation of a set of weak classifiers, the basic ones being classification and regression. In order to ensure the accuracy of the model, a common solution idea is to adopt an integrated model of trees and integrate the prediction results of multiple trees. Assume that a given data set has n samplesm number of features
And (3) predicting an output result by using K trees:
wherein the content of the first and second substances,
the function is a function of a function space and corresponds to a set of all trees, T is the number of leaf nodes in the tree, and omega is the weight corresponding to the leaf nodes. The objective function includes a loss function and a regularization term:
where l is the predicted value
And the actual value yiThe training error of the model is that omega represents a regular term punishment on the complexity of the model, which is beneficial to smoothing the final learning weight and avoiding overfitting, because the current parameters are in a function space rather than the traditional Euclidean space, the traditional random gradient descent method cannot be adopted to optimize the target function, the training of the model adopts an addition strategy, namely, the learned tree is fixed, the original model is kept unchanged, and a new tree is added into the model at one time. General assumptions
Each time a new tree is added, iteratively calculating the output forecast value of the model of the t round:
wherein the content of the first and second substances,
representing the prediction of the model of the t-th round,
representing the prediction of the model of round t-1, ft(xi) A function (tree) representing the new addition, which is given by the above equation:
according to the second order taylor expansion, the loss function in the target is simplified, and then the target function can be approximated as:
wherein the content of the first and second substances,
removing the constant term, the final target function is only related to the first and second derivatives of each sample point on the error function, i.e.:
the training error component of the objective function is determined, and the complexity of the tree is then defined, assuming fkThe decision tree is divided into a structural part q and a leaf node omega, the structural function q maps the input to the index number of the leaf, and omega gives the leaf score corresponding to each index number.
f(x)＝ωq(x),ω∈RT,q:Rd→{1,2...T}
After the decision tree is defined as above, the complexity of the tree can be defined as follows:
where I is defined as the set of samples I above each leafj＝{i|q(xi)J, the objective function can be rewritten as follows:
the target function comprises T independent univariate quadratic functions and is defined
The objective function can be further rewritten to the form:
assuming that the structure of the tree, q (x), is fixed, for ωjAnd (3) obtaining the optimal weight of the leaf node and the optimal value of the objective function by derivation:
in the above formula, obj(t)To represent the structural score of the tree, obj(t)The smaller the tree structure, the better the tree structure, the scoring function is used to find out a tree with an optimal structure, the tree is added into the model, and the operation is repeated. However, enumerating all tree structures is not feasible, so the common method is a greedy method, each time a partition is added to an existing leaf, the partition with the minimum objective function and the maximum gain is selected, and the gain formula is as follows:
wherein G isL，GRThe sum of the gains of the left tree and the right tree are respectively expressed, and the formula is used for selecting the split candidate variables.
The circulating neural network is trained, because the smelting of the blast furnace is a dynamic time sequence process and has the characteristics of large time lag, strong lag and the like, and the smelting process of the blast furnace is gradually changed, the current furnace condition has certain correlation with the historical furnace condition, so the trend of the molten iron silicon content is not only related to the current parameter, but also related to the parameter at the previous moment. The circulating neural network can learn new information and store historical information for selective persistence, and a variant Long-Short Term Memory model (LSTM) of the circulating neural network is selected to predict the trend of the silicon content of the molten iron.
In order to overcome the problem of disappearance of the gradient of the recurrent neural network, a long-short term memory model (LSTM) is provided, the LSTM constructs a special memory storage unit, the problem that the hidden layer continuously superposes an input sequence under a new time node state to cause fuzzy information in the front and cannot be continuously transmitted backwards can be solved, and the method is more suitable for prediction in the aspect of time sequences. The basic structure of the long-short term memory model (LSTM) is as follows:
the long-short term memory model (LSTM) is defined as: input time series X ═ X1,x2...xnAnd each input is provided with a corresponding input gate, a forgetting gate and an output gate to protect and control the state of the cell. The cell state controls the memory and forgetting of the long term memory unit.
In the long-short term memory model (LSTM), forget gate ftThe first step decides what information to discard from the cell state:
ft＝σ(ωf[ht-1,xt]+bf)
ht-1representing the output of the last moment, xtRepresenting the input at the current time. f. oftOutputting a value between 0 and 1 to the cell state ct-1If the output is 1, the data is completely retained,if the output is 0, it is discarded completely.
The next step is to determine what information is deposited in the cell state:
it＝σ(ωi[ht-1,xt]+bi)
input door itThe number between outputs 0-1 determines what value is updated,
creating a new candidate variable by tanh, followed by updating the state of the old cell, and updating the state of the old cell c by the above equationt-1Renewal to New cell State Ct. Old cell State ct-1And the output f of the forgetting gatetMultiplication, discarding information not to be retained, and addition of input gate itOutput of (2) and new candidate vector
Product of (a) according to CtTo determine the degree of renewal of each cellular state.
The final output is then determined, which depends on our previous cell state:
Ot＝σ(ωo[ht-1,xt]+bo)
ht＝Ot*tanh(Ct)
firstly, which part of the cell state is output is determined through a Sigmoid function, and then, the cell state is multiplied by the output of the output gate through tanh processing, and finally, the determined output part is obtained. The above-mentioned control gates and memory cells allow the LSTM unit to adaptively forget, memorize and present memory content. The opening and closing of the forgetting door can simultaneously occur in different LSTM units. The long-term and short-term memory model based on the recurrent neural network can discover and establish the long-term dependency relationship between the input value and the output value.
In this embodiment, the two models can achieve a better classification effect after simple parameter adjustment, but a complex parameter adjustment process is required to further improve the accuracy of the models, and in order to avoid the parameter adjustment process which is time-consuming and labor-consuming and has not necessarily significantly improved effect, model fusion becomes a good method for improving the prediction accuracy of the models. The model fusion fully utilizes the characteristics of a plurality of existing models, has better fitting capability and can form the complementary relation of the two models.
In the present embodiment, model fusion is performed by stacking (also referred to as "stacking"), which is a model combination technique for combining information from a plurality of prediction models to generate a new model. The stacked model is also referred to as a secondary model because of its smoothness and ability to highlight each base model where it performs best, and blacking out each base model where it performs poorly, it will be better than each individual model. The specific fusion process is shown in fig. 3, the first layer prediction model is a maximum gradient lifting model and a long-short term memory model, the second layer prediction model is a logistic regression model, and the stacking (stacking) specific steps are as follows:
step1: dividing an original sample set into a training set and a testing set according to a certain proportion, and averagely dividing the training set into five parts;
step2: training a limit gradient lifting model of the first layer prediction model by four parts in the training set, predicting the fifth part of data on the trained limit gradient lifting model and reserving a prediction result, and iterating the process for five times;
step 3: changing the extreme gradient lifting model in Step2 into a long-short term memory model, and executing the operation of Step 2;
step4: synthesizing the prediction results of the extreme gradient lifting model and the growth short-term memory model into a row, inputting the row into a logistic regression model of a second-layer prediction model, and finishing the training of the model;
step5: predicting the test set by using the trained first-layer prediction model and reserving the results, obtaining five groups of different prediction results after five iterations of each model, and obtaining a group of prediction results by taking the mode of the five groups of results on the values on the positions;
step 6: and synthesizing the prediction results of the two models in the first layer of prediction model into a column, and inputting the column into a logistic regression model in the second layer of prediction model to obtain a final prediction result.
The method in this example is applied to a certain steel mill 2650m3The blast furnace is used for carrying out test tests, and the method specifically comprises the following steps:
1) and (4) preprocessing data. The data collected from the blast furnace detection device is processed in a relevant mode to improve the data quality, the data from 1 month in 2017 to 6 months in 2017 are obtained in the scheme, and the specific steps are as follows:
the abnormal data is removed by adopting a box line graph method, and 248 abnormal data are removed in total by processing the abnormal values.
And counting the number of attribute missing values of each sample according to rows, sequencing the number of the missing values from small to large, and taking the serial number as an abscissa and the number of the missing values as an ordinate. And respectively drawing a scatter diagram of the statistics of the number of missing values of the training set and the test set, deleting inconsistent places in the scatter diagram as outliers, and filling other missing values according to values of two moments before and after, wherein 582 missing data are filled in the embodiment.
And (6) normalization processing.
2) And determining the variation trend of the silicon content of the molten iron. Determining the prediction period to be half a shift (four hours) according to experience and time lag analysis of field experts, fitting the numerical value of the silicon content in every four hours by using an n-order polynomial equation (n <3) to obtain a first-order and a second-order derivative of the equation, and when the first-order derivative is 0, indicating that the equation is in a stable and constant trend; when the first derivative is positive, the linear ascending trend is shown; when the first derivative is negative, the linear descending trend is shown; the positive second derivative of the first derivative is regular to be a convex rising trend; the positive second derivative of the first derivative is a concave ascending trend; the negative second derivative of the first derivative is regular to be a convex descending trend; the negative first derivative and the negative second derivative are concave descending trends.
3) And establishing a limit gradient forecasting model. And forming a training set and a testing set by the samples processed by the process according to a certain proportion for training and predicting the model. Inputting the data into a limit gradient lifting model, finding out the optimal parameters of the model by a grid search method to obtain the prediction result of the content trend of the silicon in the molten iron, wherein the total number of test sets is 100, 67 test sets are hit totally, and the hit rate is 67%. And obtaining the importance degree pre-ordering of the input characteristics, and selecting 15 characteristics before pre-ordering according to the importance degree ordering of the extreme gradient prediction model, namely the oxygen-enriched pressure, the hot air temperature, the blast temperature and the CO2Theoretical combustion temperature, H2The arch temperature northeast, CO, the permeability index, the arch temperature southeast, the cold air flow, the coal injection quantity in the last hour, the arch temperature southwest, the oxygen-enriched flow and the oxygen-enriched rate are input into a circulating neural network.
4) And establishing a recurrent neural network forecasting model. The samples processed by the above process are formed into a training set and a testing set according to a certain proportion, and data are input into a Long-Short Term Memory model (LSTM) to obtain a prediction result of the content trend of the molten iron silicon, wherein the number of the testing sets is 100, the total number of the testing sets is 62, and the hit rate is 62%.
5) Establishing a fusion model of the extreme gradient elevation and long-short term memory model, wherein the predicted result of the fusion model is shown in fig. 4(a) and 4(b), each of fig. 4(a) and 4(b) comprises 50 results, 0 represents a steady trend, 1 represents a convex upward trend, 2 represents a linear upward trend, 3 represents a concave upward trend, 1 represents a convex downward trend, 2 represents a linear downward trend, and 3 represents a concave downward trend. In the graph, circles are actual categories, pentagons are prediction categories, the coincidence of the pentagons and the circles indicates prediction hits, 100 test sets hit 75 in total, the hit rate is 75%, the prediction accuracy is improved to a certain extent, the advantages of an extreme gradient boost (XGBoost) prediction model and a long-short term memory model (LSTM) are integrated, the intelligent prediction of the silicon content trend of molten iron is realized, and important information is provided for a blast furnace operator to regulate the furnace condition and guide production.
Fig. 5 is a block diagram showing a configuration of a blast furnace molten iron silicon content trend prediction apparatus according to an embodiment of the present application.
Referring to fig. 5, the blast furnace molten iron silicon content variation trend prediction apparatus includes: a processor (processor)810, a memory (memory)830, a communication Interface (Communications Interface)820, and a bus 840;
wherein the content of the first and second substances,
the processor 810, the memory 830 and the communication interface 820 complete communication with each other through the bus 840;
the communication interface 820 is used for information transmission between the test equipment and the communication equipment of the display device;
the processor 810 is configured to call the program instructions in the memory 830 to execute the method for predicting the trend of the content of silicon in molten iron in a blast furnace provided by the above embodiments of the method, including, for example:
acquiring characteristic parameters related to the change of the silicon content in the blast furnace molten iron in the iron-making process, and predicting the change trend of the silicon content based on a trained prediction model of the silicon content in the blast furnace molten iron;
the blast furnace molten iron content prediction model comprises a first layer prediction model and a second layer prediction model, the first layer prediction model is used for primarily predicting the silicon content change trend according to an original sample, and the second layer prediction model carries out secondary prediction according to the prediction result of the first layer prediction model to obtain the silicon content change trend.
The embodiment discloses a blast furnace molten iron silicon content variation trend prediction device, including:
at least one processor; and
at least one memory communicatively coupled to the processor, wherein:
the memory stores program instructions which can be executed by the processor, and the processor calls the program instructions to execute the method for predicting the change trend of the silicon content of the blast furnace molten iron.
The embodiment discloses a computer program product, which comprises a computer program stored on a non-transitory computer readable storage medium, wherein the computer program comprises program instructions, and when the program instructions are executed by a computer, the computer can execute the method for predicting the variation trend of the silicon content in the molten iron of the blast furnace provided by the above-mentioned method embodiments, for example, the method comprises the following steps:
acquiring characteristic parameters related to the change of the silicon content in the blast furnace molten iron in the iron-making process, and predicting the change trend of the silicon content based on a trained prediction model of the silicon content in the blast furnace molten iron;
the blast furnace molten iron content prediction model comprises a first layer prediction model and a second layer prediction model, the first layer prediction model is used for primarily predicting the silicon content change trend according to an original sample, and the second layer prediction model carries out secondary prediction according to the prediction result of the first layer prediction model to obtain the silicon content change trend.
The present embodiment provides a non-transitory computer-readable storage medium, which stores computer instructions, which cause the computer to execute the method for predicting the variation trend of the silicon content in molten iron of a blast furnace provided by the above method embodiments, for example, the method includes:
acquiring characteristic parameters related to the change of the silicon content in the blast furnace molten iron in the iron-making process, and predicting the change trend of the silicon content based on a trained prediction model of the silicon content in the blast furnace molten iron;
the blast furnace molten iron content prediction model comprises a first layer prediction model and a second layer prediction model, the first layer prediction model is used for primarily predicting the silicon content change trend according to an original sample, and the second layer prediction model carries out secondary prediction according to the prediction result of the first layer prediction model to obtain the silicon content change trend.
In summary, the invention provides a method, a device and a storage medium for predicting the variation trend of the silicon content of the molten iron of the blast furnace, the trend of the silicon content of the molten iron of the blast furnace is predicted by a data-driven method, the quality of an input sample is improved by preprocessing characteristic parameters influencing the trend of the silicon content, and the output trend category is determined by fitting the value regression of the silicon content in a sliding window; a fusion model of extreme gradient promotion and a long-term and short-term memory network is established to predict the trend of the silicon content in the molten iron; the method has the advantages that the direction and the amplitude of the trend of the silicon content of the molten iron in half an iron-making shift (four hours) are predicted by fully utilizing the data which can be detected by the blast furnace and influence the silicon content of the molten iron, reference basis is provided for a blast furnace operator to judge the change trend of the furnace condition and the regulation and control amplitude in advance, and the smooth operation of the iron-making process and the maintenance of the quality of the molten iron in a normal range are further ensured.
Those of ordinary skill in the art will understand that: all or part of the steps for implementing the method embodiments may be implemented by hardware related to program instructions, and the program may be stored in a computer readable storage medium, and when executed, the program performs the steps including the method embodiments; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.
The above-described embodiments of the test equipment and the like of the display device are merely illustrative, wherein the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only used for illustrating the technical solutions of the embodiments of the present invention, and are not limited thereto; although embodiments of the present invention have been described in detail with reference to the foregoing embodiments, those skilled in the art will understand that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Finally, the method of the present invention is only a preferred embodiment and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (6)
1. A method for predicting the variation trend of the silicon content of molten iron in a blast furnace is characterized by comprising the following steps:
acquiring characteristic parameters related to the change of the silicon content in the blast furnace molten iron in the iron-making process, and predicting the change trend of the silicon content based on a trained prediction model of the silicon content in the blast furnace molten iron;
the blast furnace molten iron content prediction model comprises a first layer prediction model and a second layer prediction model, wherein the first layer prediction model is used for primarily predicting the silicon content change trend according to an original sample, and the second layer prediction model is used for secondarily predicting according to the prediction result of the first layer prediction model to obtain the silicon content change trend;
before obtaining characteristic parameters related to the change of the silicon content in the blast furnace molten iron in the iron-making process, the method further comprises the following steps:
acquiring characteristic parameters and silicon content change trends related to the change of the silicon content in the blast furnace molten iron in the historical ironmaking process, and training a blast furnace molten iron content prediction model;
and training the first layer prediction model according to the characteristic parameters and the silicon content variation trend, and training the second layer prediction model according to the prediction result of the trained first layer prediction model.
The method for acquiring characteristic parameters and silicon content change trends related to the silicon content change in the blast furnace molten iron in the historical ironmaking process comprises the following steps:
acquiring field data in an iron-making process, and performing outlier elimination, missing value filling and normalization processing on the field data to obtain characteristic parameters related to the silicon content change in the blast furnace molten iron;
and (4) dividing the silicon content in the blast furnace molten iron on a time axis through a sliding window to obtain the variation trend of the silicon content in the blast furnace molten iron.
The method comprises the following steps of segmenting the silicon content in the blast furnace molten iron on a time axis through a sliding window, and specifically comprises the following steps:
s1, setting the width of a total window, determining the prediction period of the silicon content change in the blast furnace molten iron according to the number of ironmaking shifts, and setting the width of a sliding window according to the prediction period;
s2, for each sliding window, fitting the silicon content in the sliding window through a zeroth order polynomial to obtain a fitting equation, judging the fitting effect through variance ratio test, if the fitting equation is not obvious, recording the silicon content in the sliding window as a stable and invariable trend, otherwise, entering the step S3;
s3, fitting the silicon content in the sliding window through a first-order polynomial to obtain a fitting equation, judging the fitting effect through variance ratio test, if the fitting equation is not significant, judging the silicon content in the sliding window to be in a linear ascending trend or a linear descending trend according to the positive and negative slope, otherwise, entering the step S4;
s4, fitting the silicon content in the sliding window through a second-order polynomial to obtain a fitting equation, solving a first derivative and a second derivative of the fitting equation, and if the first derivative is positive and the second derivative is positive, judging that the silicon content in the sliding window is in a convex rising trend; if the first derivative is positive and the second derivative is negative, judging that the silicon content in the sliding window is in a concave ascending trend; if the first derivative is negative and the second derivative is positive, judging that the silicon content in the sliding window is in a convex descending trend; and if the first derivative is negative and the second derivative is negative, judging that the silicon content in the sliding window is in a concave descending trend.
2. The method of claim 1, wherein the prediction period is one-half of an iron shift.
3. The method of claim 1, wherein the first layer prediction model comprises a extreme gradient boosting model and a long-short term memory model;
the first layer prediction model is trained by original samples, and comprises the following steps:
training a limit gradient lifting model through the characteristic parameters and the silicon content variation trend to obtain the trained limit gradient lifting model and the parameter importance ranking influencing the silicon content variation trend;
and training a plurality of characteristic parameters ranked at the top in the parameter importance ranking based on the long-short term memory model.
4. The method of claim 3, wherein the training of the second-layer prediction model by the prediction result of the first-layer prediction model comprises:
and predicting the original sample based on the trained extreme gradient lifting model and the long-short term memory model, combining the prediction results of the extreme gradient lifting model and the long-short term memory model into a row, and inputting the combined prediction results into the second layer of prediction model for training.
5. A blast furnace molten iron silicon content variation trend prediction device is characterized by comprising:
at least one processor, at least one memory, a communication interface, and a bus; wherein the content of the first and second substances,
the processor, the memory and the communication interface complete mutual communication through the bus;
the communication interface is used for information transmission between the test equipment and the communication equipment of the display device;
the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method of any of claims 1 to 4.
6. A non-transitory computer-readable storage medium storing computer instructions that cause a computer to perform the method of any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810305865.7A CN108764517B (en) | 2018-04-08 | 2018-04-08 | Method, equipment and storage medium for predicting change trend of silicon content in molten iron of blast furnace |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810305865.7A CN108764517B (en) | 2018-04-08 | 2018-04-08 | Method, equipment and storage medium for predicting change trend of silicon content in molten iron of blast furnace |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108764517A CN108764517A (en) | 2018-11-06 |
| CN108764517B true CN108764517B (en) | 2020-12-04 |
Family
ID=63981135
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810305865.7A Active CN108764517B (en) | 2018-04-08 | 2018-04-08 | Method, equipment and storage medium for predicting change trend of silicon content in molten iron of blast furnace |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108764517B (en) |
Families Citing this family (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109657285A (en) * | 2018-11-27 | 2019-04-19 | 中国科学院空间应用工程与技术中心 | The detection method of turbine rotor transient stress |
| CN109325640B (en) * | 2018-12-07 | 2022-04-26 | 中山大学 | User value prediction method, device, storage medium and equipment |
| CN109656918A (en) * | 2019-01-04 | 2019-04-19 | 平安科技（深圳）有限公司 | Prediction technique, device, equipment and the readable storage medium storing program for executing of epidemic disease disease index |
| CN110147525B (en) * | 2019-05-21 | 2023-02-14 | 内蒙古蒙树生态环境有限公司 | Method, device and equipment for predicting growth state of nursery stock and storage medium |
| CN110288096B (en) * | 2019-06-28 | 2021-06-08 | 满帮信息咨询有限公司 | Prediction model training method, prediction model training device, prediction model prediction method, prediction model prediction device, electronic equipment and storage medium |
| CN111103420B (en) * | 2019-11-20 | 2022-05-10 | 华东理工大学 | Phenolic resin product quality prediction method under uncertain raw materials |
| CN111679584B (en) * | 2020-06-23 | 2022-05-03 | 武汉钢铁有限公司 | Regulating and controlling method and device for blast furnace smelting |
| CN111832703B (en) * | 2020-06-29 | 2022-05-13 | 中南大学 | Irregular sampling dynamic sequence modeling method for process manufacturing industry |
| CN111984907B (en) * | 2020-07-31 | 2022-03-22 | 新兴铸管股份有限公司 | Method for judging development trend of blast furnace temperature |
| CN111949135B (en) * | 2020-08-31 | 2022-06-17 | 福州大学 | Touch communication fault tolerance method and system based on hybrid prediction |
| CN112434961B (en) * | 2020-12-01 | 2022-04-15 | 内蒙古科技大学 | Method and device for predicting temperature drop of molten iron on iron-steel interface and terminal equipment |
| CN113223634B (en) * | 2021-03-22 | 2022-09-16 | 浙江大学 | Blast furnace molten iron silicon content prediction method based on two-dimensional self-attention enhanced GRU model |
| CN113325811B (en) * | 2021-05-20 | 2022-03-22 | 杭州电子科技大学 | Online industrial process anomaly detection method based on memory and forgetting strategy |
| CN113205368B (en) * | 2021-05-25 | 2022-11-29 | 合肥供水集团有限公司 | Industrial and commercial customer clustering method based on time sequence water consumption data |
| CN113656930B (en) * | 2021-06-24 | 2023-11-17 | 华北理工大学 | Prediction method for smelting endpoint phosphorus content by adopting machine learning algorithm |
| CN114626303A (en) * | 2022-03-18 | 2022-06-14 | 山东莱钢永锋钢铁有限公司 | Blast furnace temperature prediction and operation guidance method based on neural network |
| CN114398140B (en) * | 2022-03-25 | 2022-05-31 | 深圳市鼎阳科技股份有限公司 | Dynamic generation method of trend graph, electronic measurement device and storage medium |
| CN114943186A (en) * | 2022-07-19 | 2022-08-26 | 数皮科技(湖北)有限公司 | Granite thermal conductivity limit lifting gradient prediction method based on whole-rock chemical data |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104778361A (en) * | 2015-04-14 | 2015-07-15 | 浙江大学 | Improved method for predicting hot-metal silicon content by EMD-Elman (empirical mode decomposition-Elman) neural network |
| CN105574297A (en) * | 2016-02-16 | 2016-05-11 | 中国石油大学(华东) | Self-adaption blast-furnace melt silicon content tendency forecasting method |
| WO2016101182A1 (en) * | 2014-12-23 | 2016-06-30 | 清华大学 | Interval type indicator forecasting method based on bayesian network and extreme learning machine |
| CN106096637A (en) * | 2016-06-06 | 2016-11-09 | 浙江大学 | Molten iron silicon content Forecasting Methodology based on the strong predictor of Elman Adaboost |
| CN106250403A (en) * | 2016-07-19 | 2016-12-21 | 北京奇艺世纪科技有限公司 | Customer loss Forecasting Methodology and device |
| CN106709197A (en) * | 2016-12-31 | 2017-05-24 | 浙江大学 | Molten iron silicon content predicting method based on slide window T-S fuzzy neural network model |
| CN106909705A (en) * | 2016-12-22 | 2017-06-30 | 上海交通大学 | A kind of blast-melted quality prediction method and its system |
| CN107704966A (en) * | 2017-10-17 | 2018-02-16 | 华南理工大学 | A kind of Energy Load forecasting system and method based on weather big data |
-
2018
- 2018-04-08 CN CN201810305865.7A patent/CN108764517B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016101182A1 (en) * | 2014-12-23 | 2016-06-30 | 清华大学 | Interval type indicator forecasting method based on bayesian network and extreme learning machine |
| CN104778361A (en) * | 2015-04-14 | 2015-07-15 | 浙江大学 | Improved method for predicting hot-metal silicon content by EMD-Elman (empirical mode decomposition-Elman) neural network |
| CN105574297A (en) * | 2016-02-16 | 2016-05-11 | 中国石油大学(华东) | Self-adaption blast-furnace melt silicon content tendency forecasting method |
| CN106096637A (en) * | 2016-06-06 | 2016-11-09 | 浙江大学 | Molten iron silicon content Forecasting Methodology based on the strong predictor of Elman Adaboost |
| CN106250403A (en) * | 2016-07-19 | 2016-12-21 | 北京奇艺世纪科技有限公司 | Customer loss Forecasting Methodology and device |
| CN106909705A (en) * | 2016-12-22 | 2017-06-30 | 上海交通大学 | A kind of blast-melted quality prediction method and its system |
| CN106709197A (en) * | 2016-12-31 | 2017-05-24 | 浙江大学 | Molten iron silicon content predicting method based on slide window T-S fuzzy neural network model |
| CN107704966A (en) * | 2017-10-17 | 2018-02-16 | 华南理工大学 | A kind of Energy Load forecasting system and method based on weather big data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108764517A (en) | 2018-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108764517B (en) | Method, equipment and storage medium for predicting change trend of silicon content in molten iron of blast furnace | |
| CN102831269B (en) | Method for determining technological parameters in flow industrial process | |
| CN114678080B (en) | Converter end point phosphorus content prediction model, construction method and phosphorus content prediction method | |
| CN113095550B (en) | Air quality prediction method based on variational recursive network and self-attention mechanism | |
| CN108875118B (en) | Method and device for evaluating accuracy of prediction model of silicon content of blast furnace molten iron | |
| CN110097929A (en) | A kind of blast furnace molten iron silicon content on-line prediction method | |
| CN114169434A (en) | Load prediction method | |
| CN114841073A (en) | Instant learning semi-supervised soft measurement modeling method based on local label propagation | |
| CN117139380A (en) | Camber control method based on self-learning of regulation experience | |
| CN114117919B (en) | Instant learning soft measurement modeling method based on sample collaborative representation | |
| CN112257348B (en) | Method for predicting long-term degradation trend of lithium battery | |
| CN114548494A (en) | Visual cost data prediction intelligent analysis system | |
| CN114139446A (en) | On-line detection soft measurement method for components in special rectification process | |
| CN112766548A (en) | Order completion time prediction method based on GASA-BP neural network | |
| CN110570339B (en) | Online education and teaching intervention method | |
| CN113377075B (en) | Method and device for real-time optimization of rare earth extraction process and computer readable storage medium | |
| CN115204040A (en) | Digital heating furnace simulation method based on neural network model | |
| Jiang et al. | A trend prediction method based on fusion model and its application | |
| He et al. | GA-based optimization of generative adversarial networks on stock price prediction | |
| Kovačević et al. | Effect of labeling algorithms on financial performance metrics | |
| CN114798763A (en) | Method and system for predicting rough rolling outlet temperature of tapping plate blank of heating furnace | |
| CN111650894A (en) | Bayesian network complex industrial process soft measurement method based on hidden variables | |
| JP6933196B2 (en) | Blast furnace unloading speed prediction model learning method, blast furnace unloading speed prediction method, blast furnace operation guidance method, blast furnace unloading speed control method, hot metal manufacturing method, blast furnace operation method, and blast furnace unloading speed prediction Model learning device | |
| CN114549026B (en) | Method and system for identifying unknown fraud based on algorithm component library analysis | |
| CN114841465A (en) | Method for predicting gas production and consumption of blast furnace |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |