Method for predicting oxygen load in iron and steel enterprises based on production plan

Patent metadata

Field	Value
Publication number	US-11755976-B2
Application number	US-202017297939-A
Country	US
Kind code	B2
Filing date	Jul 28, 2020
Priority date	Mar 23, 2020
Publication date	Sep 12, 2023
Grant date	Sep 12, 2023

Abstract

The present disclosure discloses a method for predicting oxygen load in iron and steel enterprises based on production plan, which relates to influencing factor extraction, neural network modeling and similar sequence matching technologies. The method uses the actual industrial operation data to first extract the relevant data such as the production plan and production performance of converter steel-making, analyze the influencing factors, and extract the main influencing variables of oxygen consumption. Then, the neural network prediction model of oxygen consumption of a single converter is established, the mean square error is taken as the evaluation index, and the predicting result of time granularity of a converter in the blowing stage is given. Finally, in combination with the information of smelting time and smelting duration of each device in the converter production plan, the prediction value of oxygen load in a planned time period is given.

First claim

What is claimed is:

1. A method for predicting oxygen load based on a production plan, comprising:
Step 1:
extracting the production plan and a production performance table of steel-making and iron-making from a real-time database of an industrial site, filtering production plan information and oxygen consumption data, and filling missing values;
Step 2: screening input variables and output variables of a converter oxygen prediction model based on a neural network;
wherein the converter oxygen prediction model is configured to predict a total amount of oxygen consumption of a single converter furnace, the output variables of the neural network are the total amount of the oxygen consumption of the single converter furnace, and the input variables of the neural network are determined by screening a mechanism model and analyzing maximum influencing factors; six input variables, including molten iron weight, molten iron temperature, molten iron carbon content, molten iron phosphorus content, end point carbon content and scrap steel quantity, are selected, which are all available variables in a production schedule;
Step 3: establishing a neural network model as the converter oxygen prediction model and training the neural network model;
wherein a neural network algorithm is used to construct the neural network model, and a mapping relationship between input and output is automatically learned through a large number of training samples, which approximates any nonlinear function in theory; a structure of the neural network model is divided into three layers: an input layer, a hidden layer and an output layer; assuming that the output layer of the neural network model is a Qth layer, assuming that a number of neural nodes in a q(q=1,2, . . . , Q) layer is n q , a connection coefficient from a jth neural node in a (q-1)th layer to an ith neural node in the qth layer is w, and a relationship between the input and the output of each layer of network layers is shown in Formula (1):

x
i

(
q
)

=

f
⁡
(

S
i

(
q
)

)

=

1

1
+

e

-
μ

⁢

s
i

(
q
)

(
1
)

s
i

(
q
)

=

∑

j
=
0

n
y

-
1

w
ij

(
q
)

⁢

x
j

(

q
-
1

)

(
2
)

f
⁡
(
x
)

=

1

1
+

e

-
x

(
3
)

wherein x i (q) is an ith output variable in the qth layer, S i (q) in Formula (1) is shown in Formula (2), a structure of f relationship in Formula (1) is shown in Formula (3), μ is a set parameter, S i (q) is a sum of products of connection coefficients from variables in the (q-1)th layer to an ith neural node in the qth layer, x i (q-1) is a jth input variable in the (q-1)th layer, W ij (q) represents impact of a jth neural node in the (q-1)th layer on the ith neural node in the qth layer, when j=0, W i0 (q) =−1, X 0 (q-1) =θ i (q) , where θ i (q) represents a threshold of the i-th neural node in the qth layer;
wherein historical data in the production performance table is extracted, and the input variables selected in Step 2 are taken as input parameters, so that an input sample is shown in Formula (4); the output variables are used as output parameters, and a relationship between an output sample and the input sample is shown in Formula (5); wherein a model structure then changes and is trained one by one, and a structure with a lowest Mean Absolute Percentage Error (MAPE) is used as a final model structure, and its formula is shown in Formula (6);

X

Assignees

• Univ Dalian Tech

Inventors

• Zhao Jun
• Jin Feng
• Yang Guanghui
• Wang Wei

Classifications

• G06Q10/04 Primary

  Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" (market predictions or forecasting for commercial activities G06Q30/0202) · CPC title

• G06Q10/06375 Primary

  Prediction of business process outcome or impact based on a proposed change · CPC title

• G06Q50/04

  Manufacturing · CPC title

• G06N3/08

  Learning methods · CPC title

• Y02P90/30

  Computing systems specially adapted for manufacturing · CPC title