Iron and steel industry carbon monitoring method and system based on Internet of Things sensing

Abstract

The invention relates to the technical field of carbon monitoring, in particular to a steel industry carbon monitoring method and system based on Internet-of-Things sensing, and the method comprises the steps: carrying out the data monitoring of a target steel production line according to multiple sensors, so as to obtain the observation value, timestamp and spatial position code of each sensor, and according to the spatial distance and time difference between the sensors, carrying out the data monitoring of the target steel production line; constructing a space-time difference matrix; constructing a space-time diagram among multiple sensors according to the space-time difference matrix, and performing information aggregation on the space-time diagram to obtain fusion data; according to the technological process, a technological adjacency matrix is constructed, and the technological adjacency matrix represents the material and energy flow relation between the steel production equipment. The data is trained in real time through the graph neural network, the carbon emission intensity is corrected, the real-time correction function can dynamically adjust the carbon emission monitoring result according to fluctuation and changes in the production process, and it is ensured that the carbon emission monitoring result is always accurate and reliable.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon monitoring, and in particular to a carbon monitoring method and system for the steel industry based on IoT sensing. Background Art

[0002] Traditional methods usually rely on a single sensor or limited measurement points and are unable to fully capture carbon emission data from all links in the steel production process. This method cannot accurately estimate the carbon emission intensity of different production stages and may result in large errors, especially in complex production environments. The accuracy of the monitoring results is difficult to guarantee. Traditional methods often rely on static models or fixed calculation formulas for carbon emission assessments and are unable to adjust and correct data in real time. This means that when fluctuations or changes occur in the production process, traditional methods are unable to promptly reflect changes in carbon emissions, which may result in delayed or inaccurate carbon emission data. Traditional monitoring methods generally focus on a single factor (such as a certain equipment or a certain process link) and lack a comprehensive analysis of multiple factors. For example, traditional methods may only consider the data from individual sensors and ignore multiple equipment, process links, and the interactions between links. Summary of the invention

[0003] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a carbon monitoring method and system for the steel industry based on IoT sensing.

[0004] The technical solution adopted to solve the above technical problems is: a carbon monitoring method for the steel industry based on IoT sensing, including: Data monitoring of the target steel production line is performed using multiple sensors to obtain the observation value, timestamp, and spatial position code of each sensor, and a spatiotemporal difference matrix is ​​constructed based on the spatial distance and time difference between the sensors; Constructing a spatiotemporal graph among multiple sensors according to the spatiotemporal difference matrix, and performing information aggregation on the spatiotemporal graph to obtain fused data; Constructing a process adjacency matrix according to the process flow, wherein the process adjacency matrix represents the material and energy flow relationship between steel production equipment; Monitoring the carbon emission intensity of the target steel production line according to the fusion data to obtain the carbon emission intensity of the target steel production line; Training the process adjacency matrix, the fusion data, and the carbon emission intensity according to a pre-trained graph neural network to obtain a carbon emission correction value; The carbon emission intensity is corrected according to the carbon emission correction value to obtain the actual carbon emission intensity of the target steel production line.

[0005] Preferably, the spatiotemporal graph includes a node set and a connection edge set, wherein the nodes in the node set correspond to the sensors, and when the spatiotemporal difference between each sensor in the spatiotemporal difference matrix is ​​less than a preset difference threshold, there is a connection edge between two nodes in the node set.

[0006] Preferably, information aggregation is performed on the spatiotemporal graph to obtain fused data, including: The attention weights between the nodes in the spatiotemporal graph are calculated according to the scaled dot product attention, wherein the calculation formula of the attention weights between the nodes is as follows: ; in, Indicates the duty cycle time-space diagram and The attention weights between nodes, represents the space-time difference calculation function, where the space-time difference is represented by the Euclidean distance, , and Indicates The observations, timestamps and spatial location encodings of each sensor, , and Indicates Observations, timestamps and spatial location encoding of each sensor; According to the attention weights between the nodes in the space-time graph, information aggregation is performed on each node in the space-time graph through a graph convolution operation to obtain fused data, wherein the calculation formula of the fused data is as follows: ; in, represents fused data, represents a multilayer perceptron, Represents a concatenation operation.

[0007] Preferably, the calculation formula for the carbon emission intensity of the target steel production line is as follows: ; in, represents the carbon emission intensity of the target steel production line, and Indicates the density of input and output materials, and Indicates the carbon content of the input and output materials, Indicates temperature and pressure Dependent reaction rate term.

[0008] Preferably, the graph neural network is used to correct the carbon emission intensity based on the material flow relationship between the steel production equipment and the fusion data.

[0009] Preferably, the calculation formula for the actual carbon emission intensity is as follows: ; in, Indicates the actual carbon emission intensity, Indicates the carbon emission correction value.

[0010] The technical solution adopted to solve the above technical problems is: a carbon monitoring system for the steel industry based on IoT sensing, which is applicable to the carbon monitoring method for the steel industry based on IoT sensing, including: A data monitoring unit, the data monitoring unit is used to perform data monitoring on the target steel production line according to multiple sensors to obtain the observation value, time stamp and spatial position code of each sensor, and construct a spatiotemporal difference matrix according to the spatial distance and time difference between the sensors; A data fusion unit, the data fusion unit is used to construct a spatiotemporal graph between multiple sensors according to the spatiotemporal difference matrix, and perform information aggregation on the spatiotemporal graph to obtain fused data; A process modeling unit, wherein the process modeling unit is used to construct a process adjacency matrix according to the process flow, wherein the process adjacency matrix represents the material and energy flow relationship between steel production equipment; A carbon monitoring unit, the carbon monitoring unit is used to monitor the carbon emission intensity of the target steel production line according to the fusion data to obtain the carbon emission intensity of the target steel production line; A correction calculation unit, the correction calculation unit is used to train the process adjacency matrix, the fusion data and the carbon emission intensity according to a pre-trained graph neural network to obtain a carbon emission correction value; An emission correction unit, wherein the emission correction unit is used to correct the carbon emission intensity according to the carbon emission correction value to obtain the actual carbon emission intensity of the target steel production line.

[0011] The beneficial effects of the present invention are as follows: (1) The present invention can obtain accurate observation values ​​of each sensor through real-time data monitoring of the target steel production line through multiple sensors, which helps to more accurately estimate the carbon emission intensity in the steel production process. Combined with the spatiotemporal difference matrix, the spatial position and time differences between different sensors can be considered, so as to more carefully capture the carbon emission sources in the process flow and improve the monitoring accuracy; and through the information aggregation of the spatiotemporal graph, the sensor information can be dynamically integrated, and the data can be trained in real time through the graph neural network to correct the carbon emission intensity. This real-time correction function can dynamically adjust the carbon emission monitoring results according to the fluctuations and changes in the production process to ensure that it is always accurate and reliable; (2) The process adjacency matrix constructed by the present invention can reflect the material and energy flow relationship between equipment in steel production. This model can better consider the interaction between various links and equipment in the production process, optimize the monitoring and evaluation of carbon emissions, and not only consider local sensor data, but also integrate the synergy of the entire production system; (3) The present invention can effectively compensate for the errors in traditional carbon emission monitoring methods by introducing carbon emission correction values. In particular, when the process flow changes or equipment failures occur, the correction mechanism can provide more stable and true emission data. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A schematic flow chart of the steps of an overall method in an embodiment of the present invention; Figure 2 A schematic diagram of the system architecture of an overall system in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] Embodiment 1, as Figure 1 As shown, the present invention proposes a carbon monitoring method for the steel industry based on IoT sensing, comprising: S1. Data monitoring of the target steel production line is performed based on multiple sensors to obtain the observation value, timestamp and spatial position code of each sensor, and a spatiotemporal difference matrix is ​​constructed based on the spatial distance and time difference between sensors; S2, constructing a spatiotemporal graph between multiple sensors according to the spatiotemporal difference matrix, and performing information aggregation on the spatiotemporal graph to obtain fused data; S3. construct a process adjacency matrix according to the process flow, wherein the process adjacency matrix represents the material and energy flow relationship between steel production equipment; S4. Monitor the carbon emission intensity of the target steel production line according to the fusion data to obtain the carbon emission intensity of the target steel production line; S5. Training the process adjacency matrix, fusion data, and carbon emission intensity according to a pre-trained graph neural network to obtain a carbon emission correction value; S6. Correct the carbon emission intensity according to the carbon emission correction value to obtain the actual carbon emission intensity of the target steel production line.

[0014] In the present invention, IoT sensing refers to connecting various sensor devices to the Internet through IoT technology, so that these sensors can collect data in real time and transmit the data to the central system for analysis and processing. In the steel industry, IoT sensors can be used to monitor parameters such as temperature, pressure, and gas composition on the production line; the space-time difference matrix is ​​a mathematical tool used to represent the spatial distance and time difference between sensors. Specifically, the elements of this matrix can represent the position difference and time difference between each sensor, and are used to capture the space-time relationship between observation data of different sensors, which is usually used for subsequent data processing and analysis; the space-time graph is a graph structure in which the nodes represent sensors and the edges represent the space-time relationship between sensors. The space-time graph can help aggregate and analyze data by visualizing the space-time relationship between sensors. In monitoring and prediction, the space-time graph can help better understand the mutual influence between different sensors; the process adjacency matrix is ​​a matrix used to describe the material and energy flow relationship between steel production equipment. In steel production, there will be certain material and energy flows between different equipment, such as the transmission of coal, molten steel, gas, etc. The adjacency matrix is ​​used to represent the relationship between these flows. Usually each row and each column represents A production equipment, the elements in the matrix represent the material flow or energy flow between the equipment; carbon emission intensity refers to the amount of carbon dioxide emitted per unit output value or unit production volume. In the steel production process, carbon emission is an important environmental indicator. Carbon emission intensity is usually expressed as the amount of carbon dioxide emitted per ton of steel products, which is a key indicator for measuring carbon emission efficiency; graph neural network is a deep learning model used to process tasks with graph data structures. In the graph neural network, the relationship between data nodes and nodes (i.e., the edges of the graph) is used to help the model learn and reason. In this method, the graph neural network is used to combine process adjacency matrix, fusion data, carbon emission intensity and other information to optimize the carbon emission prediction model through training; the carbon emission correction value is the value adjusted for the initial carbon emission intensity. Through the training of the graph neural network, the correction value can be obtained. This value can help correct the original carbon emission intensity data, making the final carbon emission intensity more accurate and true. This is because the actual situation of carbon emissions may be affected by multiple factors. The graph neural network corrects carbon emissions by learning these influencing factors; the actual carbon emission intensity is the corrected carbon emission intensity data, which represents the actual carbon emission level of the steel production line under specific conditions.

[0015] Embodiment 2: A carbon monitoring method for the steel industry based on Internet of Things sensing is proposed in the present invention. Compared with Embodiment 1, this embodiment also includes: a spatiotemporal graph includes a node set and a connection edge set, wherein the nodes in the node set correspond to sensors, and when the spatiotemporal difference between each sensor in the spatiotemporal difference matrix is ​​less than a preset difference threshold, there is a connection edge between the two nodes in the node set.

[0016] In an optional embodiment, information aggregation is performed on the spatiotemporal graph to obtain fused data, including: A1. Calculate the attention weights between nodes in the spatiotemporal graph based on the scaled dot product attention. The calculation formula for the attention weights between nodes is as follows: ; in, Indicates the duty cycle time-space diagram and The attention weights between nodes, represents the space-time difference calculation function, where the space-time difference is represented by the Euclidean distance, , and Indicates The observations, timestamps and spatial location encodings of each sensor, , and Indicates Observations, timestamps and spatial location encoding of each sensor; A2. According to the attention weights between the nodes in the space-time graph, the information of each node in the space-time graph is aggregated through the graph convolution operation to obtain fused data. The calculation formula of the fused data is as follows: ; in, represents fused data, represents a multilayer perceptron, Represents a concatenation operation.

[0017] It should be noted that scaled dot product attention is a form of attention mechanism, mainly used for self-attention mechanism (Self-Attention), which measures the correlation between "query (Query)" and "key (Key)" by calculating the dot product between them, and gives different weights according to this correlation. In order to avoid numerical instability caused by excessive dot product values, the dot product value is usually divided by a constant (usually the square root of the dimension of the vector), which is the "scaling" operation; the spatiotemporal difference calculation function is used to measure the difference between two nodes in the spatiotemporal graph, usually using Euclidean distance to measure the spatial distance, while also considering the difference in time. It helps define the relationship between nodes so that spatiotemporal information can be better integrated in the graph convolution operation; graph convolution is a convolution operation for graph structured data, which updates the feature representation of each node by aggregating the information of neighboring nodes. Graph convolution is designed to process the relationship between nodes, and can process data of arbitrary topological structure without considering the traditional grid structure (such as pictures); multilayer perceptron (MLP) is a feedforward neural network, usually composed of multiple fully connected layers (Dense Layers), which is a common neural network structure used for feature conversion and mapping, and can capture nonlinear relationships in input data.

[0018] In an optional embodiment, the calculation formula for the carbon emission intensity of the target steel production line is as follows: ; in, represents the carbon emission intensity of the target steel production line, and Indicates the density of input and output materials, and Indicates the carbon content of the input and output materials, Indicates temperature and pressure Dependent reaction rate term.

[0019] In an optional embodiment, a graph neural network is used to correct the carbon emission intensity based on the material flow relationship and fusion data between steel production equipment.

[0020] In an optional embodiment, the calculation formula of the actual carbon emission intensity is as follows: ; in, Indicates the actual carbon emission intensity, Indicates the carbon emission correction value.

[0021] Embodiment three, as Figure 2As shown, the present invention proposes a carbon monitoring system for the steel industry based on IoT sensing, which is applicable to the carbon monitoring method for the steel industry based on IoT sensing, including: A data monitoring unit is used to monitor the target steel production line based on multiple sensors to obtain the observation value, timestamp and spatial position code of each sensor, and to construct a spatiotemporal difference matrix based on the spatial distance and time difference between sensors; The data fusion unit is used to construct a spatiotemporal graph among multiple sensors according to the spatiotemporal difference matrix and aggregate information on the spatiotemporal graph to obtain fused data; A process modeling unit, which is used to construct a process adjacency matrix according to the process flow, wherein the process adjacency matrix represents the material and energy flow relationship between steel production equipment; A carbon monitoring unit, which is used to monitor the carbon emission intensity of the target steel production line based on the fusion data to obtain the carbon emission intensity of the target steel production line; A correction calculation unit, which is used to train the process adjacency matrix, fusion data and carbon emission intensity according to a pre-trained graph neural network to obtain a carbon emission correction value; The emission correction unit is used to correct the carbon emission intensity according to the carbon emission correction value to obtain the actual carbon emission intensity of the target steel production line.

[0022] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A carbon monitoring method for the steel industry based on IoT sensing, characterized in that: include: Data monitoring of the target steel production line is performed using multiple sensors to obtain the observation value, timestamp, and spatial position code of each sensor, and a spatiotemporal difference matrix is ​​constructed based on the spatial distance and time difference between the sensors; Constructing a spatiotemporal graph among multiple sensors according to the spatiotemporal difference matrix, and performing information aggregation on the spatiotemporal graph to obtain fused data; Constructing a process adjacency matrix according to the process flow, wherein the process adjacency matrix represents the material and energy flow relationship between steel production equipment; Monitoring the carbon emission intensity of the target steel production line according to the fusion data to obtain the carbon emission intensity of the target steel production line; Training the process adjacency matrix, the fusion data, and the carbon emission intensity according to a pre-trained graph neural network to obtain a carbon emission correction value; The carbon emission intensity is corrected according to the carbon emission correction value to obtain the actual carbon emission intensity of the target steel production line.

2. The carbon monitoring method for the steel industry based on IoT sensing according to claim 1 is characterized in that: The spatiotemporal graph includes a node set and a connection edge set, wherein the nodes in the node set correspond to the sensors, and when the spatiotemporal difference between each sensor in the spatiotemporal difference matrix is ​​less than a preset difference threshold, a connection edge exists between two nodes in the node set.

3. The carbon monitoring method for the steel industry based on IoT sensing according to claim 2 is characterized in that: Aggregating information on the spatiotemporal graph to obtain fused data includes: The attention weights between the nodes in the spatiotemporal graph are calculated according to the scaled dot product attention, wherein the calculation formula of the attention weights between the nodes is as follows: ； in, Indicates the duty cycle time-space diagram and The attention weights between nodes, represents the space-time difference calculation function, where the space-time difference is represented by the Euclidean distance, , and Indicates The observations, timestamps and spatial location encodings of each sensor, , and Indicates Observations, timestamps and spatial location encoding of each sensor; According to the attention weights between the nodes in the space-time graph, information aggregation is performed on each node in the space-time graph through a graph convolution operation to obtain fused data, wherein the calculation formula of the fused data is as follows: ； in, represents fused data, represents a multilayer perceptron, Represents a concatenation operation.

4. The carbon monitoring method for the steel industry based on IoT sensing according to claim 3 is characterized in that: The calculation formula for the carbon emission intensity of the target steel production line is as follows: ； in, represents the carbon emission intensity of the target steel production line, and Indicates the density of input and output materials, and Indicates the carbon content of the input and output materials, Indicates temperature and pressure Dependent reaction rate term.

5. The carbon monitoring method for the steel industry based on IoT sensing according to claim 4 is characterized in that: The graph neural network is used to correct the carbon emission intensity based on the material flow relationship between the steel production equipment and the fusion data.

6. The carbon monitoring method for the steel industry based on IoT sensing according to claim 5 is characterized in that: The calculation formula for the actual carbon emission intensity is as follows: ； in, Indicates the actual carbon emission intensity, Indicates the carbon emission correction value.

7. A carbon monitoring system for the steel industry based on IoT sensing, which is applicable to a carbon monitoring method for the steel industry based on IoT sensing as described in any one of claims 1 to 6, characterized in that: include: A data monitoring unit, the data monitoring unit is used to perform data monitoring on the target steel production line according to multiple sensors to obtain the observation value, time stamp and spatial position code of each sensor, and construct a spatiotemporal difference matrix according to the spatial distance and time difference between the sensors; A data fusion unit, the data fusion unit is used to construct a spatiotemporal graph between multiple sensors according to the spatiotemporal difference matrix, and perform information aggregation on the spatiotemporal graph to obtain fused data; A process modeling unit, wherein the process modeling unit is used to construct a process adjacency matrix according to the process flow, wherein the process adjacency matrix represents the material and energy flow relationship between steel production equipment; A carbon monitoring unit, the carbon monitoring unit is used to monitor the carbon emission intensity of the target steel production line according to the fusion data to obtain the carbon emission intensity of the target steel production line; A correction calculation unit, the correction calculation unit is used to train the process adjacency matrix, the fusion data and the carbon emission intensity according to a pre-trained graph neural network to obtain a carbon emission correction value; An emission correction unit, wherein the emission correction unit is used to correct the carbon emission intensity according to the carbon emission correction value to obtain the actual carbon emission intensity of the target steel production line.

Images