Hot metal scheduling management system and method for steel production

Abstract

The present application discloses a molten iron scheduling management system and method for steel production, which uses data analysis and processing technology based on deep learning to perform local time series feature analysis and significant aggregation on the production information (iron tapping information, iron usage information, scrap steel adding information and locomotive transportation operation information), thereby automatically decoding the molten iron scheduling value based on the time series characteristics of iron consumption obtained by inferring the time series between the iron tapping information, the iron usage information and the locomotive transportation operation information, and the time series representation between the time series characteristics of the scrap steel adding information. In this way, the production information in the steel production process can be automatically analyzed and processed using a deep learning model to obtain the molten iron scheduling value, thereby adjusting the production process in real time, reducing the manual operation time of the dispatcher, and significantly improving the scheduling efficiency and intelligence level.

Description

Technical Field

[0001] The present application relates to the field of intelligent management, and more specifically, to a molten iron scheduling management system and method for steel production. Background Art

[0002] In the steel production process, hot metal scheduling plays a vital role. As the key material connecting the ironmaking and steelmaking processes, hot metal scheduling directly affects production efficiency and cost control. Reasonable hot metal scheduling can not only ensure the continuous and stable flow of hot metal to the steelmaking process, but also effectively reduce waiting time and energy consumption, thereby improving overall production efficiency.

[0003] However, in the traditional hot metal dispatching process, the dispatcher relies on personal experience and the mastery of complex dispatching information, and adopts a dispatching mode of walking and watching. This is not only time-consuming and labor-intensive, inefficient, and increases the workload of dispatchers, but also difficult to consider all the dispatching operations in a stage due to the large number of dispatching operations and the lack of efficiency. In addition, due to the uneven level of dispatchers themselves, there is a lack of predictability of cross-interference between potential hot metal dispatching operations, so it is easy to conflict with each other, affecting the efficiency and quality of hot metal dispatching operations, and then affecting the efficiency of the overall production tasks of the iron zone, causing a great impact on the stable and orderly production of steel enterprises.

[0004] Therefore, an optimized molten iron scheduling management system for steel production is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a molten iron scheduling management system and method for steel production, which performs local time series feature analysis and significant aggregation on the production information (iron tapping information, iron usage information, scrap steel adding information and locomotive transportation operation information) by adopting data analysis and processing technology based on deep learning, so as to automatically decode the molten iron scheduling value based on the time series characteristics of iron consumption obtained by inferring the time series between the iron tapping information, the iron usage information and the locomotive transportation operation information and the time series representation between the time series characteristics of the scrap steel adding information. In this way, the production information in the steel production process can be automatically analyzed and processed by the deep learning model to obtain the molten iron scheduling value, so as to adjust the production process in real time, reduce the manual operation time of the dispatcher, and significantly improve the scheduling efficiency and intelligence level.

[0006] According to one aspect of the present application, a molten iron scheduling management system for steel production is provided, comprising:

[0007] A production information collection module, used to collect production information, wherein the production information includes a time queue of iron-out information, a time queue of iron-use information, a time queue of scrap steel adding information, and a time queue of locomotive transport operation information;

[0008] A steel scheduling instruction generation module, used for performing parameter time series feature capture and significant gated feature aggregation processing on the production information to generate steel production scheduling instructions;

[0009] The molten iron scheduling execution module is used to perform molten iron scheduling based on the steel production scheduling instruction.

[0010] According to another aspect of the present application, a molten iron scheduling management method for steel production is provided, comprising:

[0011] Collecting production information, wherein the production information includes a time queue of iron-out information, a time queue of iron-use information, a time queue of scrap steel adding information, and a time queue of locomotive transport operation information;

[0012] Capturing parameter time series features and performing significant gated feature aggregation processing on the production information to generate steel production scheduling instructions;

[0013] Based on the steel production scheduling instructions, molten iron scheduling is performed.

[0014] Compared with the prior art, the present application provides a molten iron scheduling management system and method for steel production, which uses data analysis and processing technology based on deep learning to perform local time series feature analysis and significant aggregation on the production information (iron tapping information, iron usage information, scrap steel adding information and locomotive transportation operation information), thereby automatically decoding the molten iron scheduling value based on the time series characteristics of iron consumption obtained by inferring the time series between the iron tapping information, the iron usage information and the locomotive transportation operation information, and the time series representation between the time series characteristics of the scrap steel adding information. In this way, the deep learning model can be used to automatically analyze and process the production information in the steel production process to obtain the molten iron scheduling value, so as to adjust the production process in real time, reduce the manual operation time of the dispatcher, and thus significantly improve the scheduling efficiency and intelligence level. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1is a block diagram of a molten iron scheduling management system for steel production according to an embodiment of the present application;

[0017] Figure 2 A data flow diagram of a molten iron scheduling management system for steel production according to an embodiment of the present application;

[0018] Figure 3 A block diagram of a steel scheduling instruction generation module in a molten iron scheduling management system for steel production according to an embodiment of the present application;

[0019] Figure 4 It is a flow chart of a molten iron scheduling management method for steel production according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0021] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0022] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0023] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0024] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0025] In the traditional hot metal dispatching process, the dispatcher relies on personal experience and the mastery of complex dispatching information to adopt a dispatching mode of walking and watching. This is not only time-consuming and labor-intensive, but also inefficient, and increases the workload of the dispatchers. In addition, due to the numerous dispatching operations and the lack of efficiency, it is difficult to comprehensively consider all the dispatching operations within the stage. In addition, due to the uneven level of the dispatchers themselves, there is a lack of predictability of the cross-interference between potential hot metal dispatching operations, so it is easy to conflict with each other, affecting the efficiency and quality of the hot metal dispatching operations, and then affecting the efficiency of the overall production tasks of the iron zone, causing a great impact on the stable and orderly production of steel enterprises. Therefore, an optimized hot metal dispatching management system for steel production is desired.

[0026] In the technical solution of the present application, a molten iron scheduling and management system for steel production is proposed. Figure 1 It is a block diagram of a molten iron scheduling management system for steel production according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of molten iron scheduling management system for steel production according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to an embodiment of the present application, a molten iron scheduling management system 300 for steel production includes: a production information collection module 310, used to collect production information, wherein the production information includes a time queue of iron output information, a time queue of iron usage information, a time queue of scrap steel adding information and a time queue of locomotive transportation operation information; a steel scheduling instruction generation module 320, used to perform parameter timing feature capture and significance gated feature aggregation processing on the production information to generate steel production scheduling instructions; a molten iron scheduling execution module 330, used to perform molten iron scheduling based on the steel production scheduling instructions.

[0027] In particular, the production information collection module 310 is used to collect production information, wherein the production information includes a time queue of iron-out information, a time queue of iron-usage information, a time queue of scrap steel adding information, and a time queue of locomotive transportation operation information. It should be understood that the iron-out information, the iron-usage information, and the locomotive transportation operation information can reflect the iron consumption situation. In the technical solution of the present application, the iron consumption situation is combined with the scrap steel adding information for analysis to obtain the molten iron scheduling value. Therefore, in a specific example of the present application, first, the production information is collected.

[0028] In particular, the steel scheduling instruction generation module 320 is used to perform parameter time series feature capture and significant gate feature aggregation processing on the production information to generate steel production scheduling instructions. In particular, in a specific example of the present application, Figure 3As shown, the steel scheduling instruction generation module 320 includes: a steel production information time series association encoding unit 321, which is used to input the time queue of the iron-out information, the time queue of the iron-using information, the time queue of the scrap steel adding information and the time queue of the locomotive transportation operation information into a sequence encoder respectively to obtain a time queue of the iron-out information time series association feature vector, a time queue of the iron-using information time series association feature vector, a time queue of the scrap steel adding information time series association feature vector and a time queue of the locomotive transportation operation information time series association feature vector; a steel production information time series association feature significant unit 322, which is used to input the time queue of the iron-out information time series association feature vector, the time queue of the iron-using information time series association feature vector, the time queue of the scrap steel adding information time series association feature vector and the time queue of the locomotive transportation operation information time series association feature vector into a time queue based on the feature energy local significance gating A feature aggregation network is used to obtain a time series significant aggregation representation vector of iron tapping information, a time series significant aggregation representation vector of iron usage information, a time series significant aggregation representation vector of scrap steel adding information and a time series aggregation representation vector of locomotive transportation operation information; an iron consumption time series multimodal fusion unit 323 is used to perform time series reasoning based on iron consumption on the time series significant aggregation representation vector of iron tapping information, the time series significant aggregation representation vector of iron usage information and the time series aggregation representation vector of locomotive transportation operation information to obtain a time series multimodal comprehensive representation vector of iron consumption; a molten iron shortage time series representation unit 324 is used to calculate a molten iron shortage time series representation vector between the time series multimodal comprehensive representation vector of iron consumption and the time series significant aggregation representation vector of scrap steel adding information as a molten iron shortage time series representation feature; a molten iron scheduling unit 325 is used to obtain a molten iron scheduling value as the steel production scheduling instruction based on the molten iron shortage time series representation feature.

[0029] Specifically, the steel production information time series association encoding unit 321 is used to input the time queue of the iron-making information, the time queue of the iron-using information, the time queue of the scrap steel adding information and the time queue of the locomotive transportation operation information into the sequence encoder respectively to obtain the time queue of the iron-making information time series association feature vector, the time queue of the iron-using information time series association feature vector, the time queue of the scrap steel adding information time series association feature vector and the time queue of the locomotive transportation operation information time series association feature vector. Considering that each of the time queues of the iron-making information, the time queue of the iron-using information, the time queue of the scrap steel adding information and the time queue of the locomotive transportation operation information has different degrees of implicit association relationships between different time spans, the Bi-LSTM model can capture the dynamic characteristics in the time series and the long-term dependencies between different time scales. Based on this, in the technical solution of the present application, the time queue of the iron-out information, the time queue of the iron-usage information, the time queue of the scrap steel adding information and the time queue of the locomotive transport operation information are respectively input into a sequence encoder based on a Bi-LSTM model to obtain a time queue of the iron-out information timing-related feature vector, a time queue of the iron-usage information timing-related feature vector, a time queue of the scrap steel adding information timing-related feature vector and a time queue of the locomotive transport operation information timing-related feature vector.

[0030] Specifically, the steel production information time series correlation feature significant unit 322 is used to input the time queue of the iron-making information time series correlation feature vector, the time queue of the iron-using information time series correlation feature vector, the time queue of the scrap steel adding information time series correlation feature vector and the time queue of the locomotive transportation operation information time series correlation feature vector into a feature aggregation network based on feature energy local significant gating to obtain the iron-making information time series significant aggregation representation vector, the iron-using information time series significant aggregation representation vector, the scrap steel adding information time series significant aggregation representation vector and the locomotive transportation operation information time series aggregation representation vector. Considering that the time queue of the time series correlation feature vector of each production information has different influences based on the overall time series at different time points, that is, each time series correlation feature has different correlation and dependence on the surrounding time series. Therefore, in order to strengthen the time series correlation features of each time point, so as to aggregate them into a comprehensive representation vector that captures the key information of the entire time series, in the technical solution of the present application, the time queue of the iron-out information time series correlation feature vector, the time queue of the iron-use information time series correlation feature vector, the time queue of the scrap steel adding information time series correlation feature vector and the time queue of the locomotive transportation operation information time series correlation feature vector are input into a feature aggregation network based on feature energy local significance gating to obtain the iron-out information time series significant aggregation representation vector, the iron-use information time series significant aggregation representation vector, the scrap steel adding information time series significant aggregation representation vector and the locomotive transportation operation information time series aggregation representation vector. In particular, the feature aggregation network based on feature energy local significance gating is a network that quantifies the energy distribution information of each time series correlation feature and aggregates the local energy distribution.

[0031] In an embodiment of the present application, taking the time queue of the iron-out information timing-related feature vector as an example, first, based on the maximum value, minimum value, mean value and variance of each iron-out information timing-related feature vector in the time queue of the iron-out information timing-related feature vector, the characteristic energy level coefficient of each iron-out information timing-related feature vector is determined to obtain a set of iron-out information timing-related feature energy level coefficients; then, the scale of the local neighborhood is determined, and the local neighborhood significance descriptor of each iron-out information timing-related feature energy level coefficient in the set of iron-out information timing-related feature energy level coefficients is calculated. A set of significant descriptors of energy levels of the time-series correlation features of the iron-out information is obtained, so as to capture and mine the energy distribution characteristics of each time-series correlation feature vector of the iron-out information in the local neighborhood, so as to identify significant time-series correlation features in the local range; and then the set of significant descriptors of energy levels of the time-series correlation features of the iron-out information is input into a mask module based on a gating function to obtain a set of significant descriptors of energy levels of the mask-gated probabilistic time-series correlation features of the iron-out information; that is, through nonlinear probabilistic transformation and adaptive feature screening, the dynamic adjustment of the significance of the time-series correlation features of the iron-out information is realized. Subsequently, the set of significant descriptors of energy levels of the mask-gated probabilistic time-series correlation features of the iron-out information is used as a sequence of weights, and the position-weighted sum of the time queue of the time-series correlation feature vector of the iron-out information is calculated to obtain the significant aggregation representation vector of the time-series correlation of the iron-out information. Here, each mask-gated probabilistic iron-out information timing-related feature energy level significance descriptor obtained by the gating function and the mask function is used as a weight to perform weighted summation on the time queue of the iron-out information timing-related feature vector, so as to strengthen the key significance features of the iron-out information timing association in the local time series segment, and obtain the iron-out information timing-related significant aggregation representation vector with rich representation capabilities that contains the significant features of the time points.

[0032] Among them, based on the maximum value, minimum value, mean value and variance of each iron-out information timing-related feature vector in the time queue of the iron-out information timing-related feature vector, the process of determining the characteristic energy level coefficient of each iron-out information timing-related feature vector to obtain a set of iron-out information timing-related characteristic energy level coefficients includes: respectively extracting the maximum value and the minimum value in the iron-out information timing-related feature vector to obtain the maximum value of the iron-out information timing-related feature and the minimum value of the iron-out information timing-related feature; respectively calculating the mean value and the variance of the iron-out information timing-related feature vector to obtain the mean value of the iron-out information timing-related feature and the variance of the iron-out information timing-related feature; adding the iron-out information timing-related variance and the preset hyperparameter to obtain the iron-out information timing-related feature. The first energy level coefficient of the timing correlation characteristic of the iron-out information is calculated; the square of the difference between the maximum value of the timing correlation characteristic of the iron-out information and the mean value of the timing correlation characteristic of the iron-out information is calculated to obtain the maximum difference value of the timing correlation of the iron-out information; the square of the difference between the minimum value of the timing correlation characteristic of the iron-out information and the mean value of the timing correlation characteristic of the iron-out information is calculated to obtain the minimum difference value of the timing correlation of the iron-out information; the maximum difference value of the timing correlation of the iron-out information is added to the minimum difference value of the timing correlation of the iron-out information and the preset hyperparameter to obtain the second energy level coefficient of the timing correlation characteristic of the iron-out information; the first energy level coefficient of the timing correlation characteristic of the iron-out information is divided by the second energy level coefficient of the timing correlation characteristic of the iron-out information to obtain the energy level coefficient of the timing correlation characteristic of the iron-out information.

[0033] More specifically, the process of inputting the set of iron-out information timing-associated feature energy level significance descriptors into a mask module based on a gating function to obtain a set of mask-gated probabilistic iron-out information timing-associated feature energy level significance descriptors includes: inputting each iron-out information timing-associated feature energy level significance descriptor in the set of iron-out information timing-associated feature energy level significance descriptors into a Sigmoid function for normalization processing to obtain a set of normalized iron-out information timing-associated feature energy level significance descriptors; inputting each normalized iron-out information timing-associated feature energy level significance descriptor in the set of normalized iron-out information timing-associated feature energy level significance descriptors into a gating function for masking processing to obtain the set of mask-gated probabilistic iron-out information timing-associated feature energy level significance descriptors.

[0034] In summary, in the above embodiment, the time queue of the iron-out information timing-associated feature vector, the time queue of the iron-use information timing-associated feature vector, the time queue of the scrap steel adding information timing-associated feature vector, and the time queue of the locomotive transportation operation information timing-associated feature vector are input into a feature aggregation network based on feature energy local significance gating to obtain an iron-out information timing-significant aggregation representation vector, an iron-use information timing-significant aggregation representation vector, a scrap steel adding information timing-significant aggregation representation vector, and a locomotive transportation operation information timing-significant aggregation representation vector, including: inputting the time queue of the iron-out information timing-associated feature vector into a feature aggregation network based on feature energy local significance gating, and processing it with the following feature aggregation formula to obtain the iron-out information timing-significant aggregation representation vector; wherein the feature aggregation formula is: ;in, is the time queue of the characteristic vector associated with the timing of the iron-making information, represents the number of feature vectors in the time queue of the iron-tapping information timing-related feature vectors, and Respectively The mean and variance of the time series associated feature vectors of the iron tapping information, and Respectively The maximum and minimum values ​​in the time series correlation feature vector of the iron-out information, is the preset hyperparameter, It is The characteristic energy level coefficient of the tapping information time series associated characteristic vector, It is The characteristic energy level coefficient of the tapping information time series associated characteristic vector, Indicates the The number of characteristic energy coefficients in the local neighborhood centered at is the first in the set of significant descriptors of the energy level of the temporal correlation feature of the iron-making information. a significant descriptor of the energy level of the temporal correlation feature of the iron-out information, yes function, It is the first in the set of normalized iron-out information temporal correlation feature energy level significant descriptors. Normalized iron-out information temporal correlation feature energy level significance descriptor, For masking, is the first in the set of the mask-gated probabilistic iron information temporal correlation feature energy level significance descriptors. The mask gated probabilistic iron information temporal correlation feature energy level significance descriptor, is the predetermined threshold, is the significant aggregation representation vector of the iron-out information time series. Similarly, the encoding method of the time series of the iron-using information time series associated feature vectors, the time series of the scrap steel adding information time series associated feature vectors and the time series of the locomotive transportation operation information time series associated feature vectors is as shown in the above formula.

[0035] Specifically, the iron consumption time series multimodal fusion unit 323 is used to perform time series reasoning based on iron consumption on the iron production information time series significant aggregation representation vector, the iron consumption information time series significant aggregation representation vector and the locomotive transportation operation information time series aggregation representation vector to obtain the iron consumption time series multimodal comprehensive representation vector. Considering that the iron production information time series significant aggregation representation vector, the iron consumption information time series significant aggregation representation vector and the locomotive transportation operation information time series aggregation representation vector are all important parameter information in the steel production process, and these features are mutually correlated in time series. In order to better infer and predict the consumption of steel based on the information of these three, so as to more accurately generate the scheduling instructions of steel, in the technical solution of the present application, the iron production information time series significant aggregation representation vector, the iron consumption information time series significant aggregation representation vector and the locomotive transportation operation information time series aggregation representation vector are input into the iron consumption time series reasoner based on the Bayesian probability network to obtain the iron consumption time series multimodal comprehensive representation vector. It should be understood that the Bayesian probability network is good at modeling complex causal relationships and can effectively capture the interdependence between multi-source data. That is, the interrelated information between the three in the time dimension can be better integrated through the iron consumption time series reasoner based on the Bayesian probability network, thereby obtaining more accurate iron consumption prediction features.

[0036] Specifically, the molten iron shortage time series representation unit 324 is used to calculate the molten iron shortage time series representation vector between the iron consumption time series multimodal comprehensive representation vector and the scrap steel addition information time series significant aggregation representation vector as the molten iron shortage time series representation feature. Considering that the iron consumption time series multimodal comprehensive representation vector and the scrap steel addition information time series significant aggregation representation vector respectively represent the time series aggregation characteristics of iron consumption and the time series characteristics after adding scrap steel, they represent the relationship between supply and demand. Therefore, in order to more intuitively see the gap between the actual demand and supply of molten iron at different time points, it is helpful to predict and adjust the supply of molten iron to meet the iron demand in the production process. In the technical solution of the present application, the molten iron shortage time series representation vector between the iron consumption time series multimodal comprehensive representation vector and the scrap steel addition information time series significant aggregation representation vector is calculated. In particular, the difference between the two can be calculated here as the molten iron shortage time series representation vector, so that the future supply and demand contradictions can be predicted more intuitively and accurately, providing support for molten iron scheduling decisions.

[0037] Specifically, the molten iron scheduling unit 325 is used to obtain the molten iron scheduling value as the steel production scheduling instruction based on the molten iron shortage timing representation characteristics. In a specific example of the present application, the molten iron shortage timing representation vector is input into the molten iron scheduling module based on the decoder to obtain the molten iron scheduling decoded value. That is, the molten iron shortage timing representation characteristics between the iron consumption timing multimodal comprehensive representation vector and the scrap steel addition information timing significant aggregation representation vector are used for decoding processing to automatically obtain the molten iron scheduling value. In this way, the deep learning model can be used to automatically analyze and process the production information in the steel production process to obtain the molten iron scheduling value, so as to adjust the production process in real time, reduce the manual operation time of the dispatcher, and thus significantly improve the scheduling efficiency and intelligence level.

[0038] In a preferred example, the applicant of the present application takes into account that the iron-producing information time series significant aggregation representation vector, the iron-usage information time series significant aggregation representation vector, the scrap steel adding information time series significant aggregation representation vector and the locomotive transportation operation information time series aggregation representation vector respectively represent the global time domain aggregation characteristics of the iron-producing information, the iron-usage information, the scrap steel adding information and the locomotive transportation operation information based on the local time domain distribution energy significance of the short-range-long-range bidirectional time series association. However, when further performing time series position-by-time series position Bayesian time series reasoning on the iron-producing information time series significant aggregation representation vector, the iron-usage information time series significant aggregation representation vector and the locomotive transportation operation information time series aggregation representation vector, and calculating the time series position difference between the iron consumption time series multimodal comprehensive representation vector and the scrap steel adding information time series significant aggregation representation vector, the full time domain aggregation difference caused by the source time series distribution difference will further cause time series reasoning and differential aggregation alignment conflicts, thereby causing the loss of aggregation key information and affecting the expression effect of the molten iron shortage time series representation vector.

[0039] Based on this, in the preferred example, when the molten iron shortage timing representation vector is input into the molten iron scheduling module based on the decoder to obtain the molten iron scheduling decoded value, the molten iron shortage timing representation vector is optimized. Specifically, the optimization process includes the following steps:

[0040] Clustering all the eigenvalues ​​of the molten iron shortage time series representation vector based on the distance between the eigenvalues, such as the absolute value distance of the difference, and determining the ratio of the number of clustering features to the number of features of the molten iron shortage time series representation vector;

[0041] Arranging the clustering features into a molten iron shortage time series clustering vector, and dividing the second norm of the molten iron shortage time series clustering vector by the second norm of the molten iron shortage time series representation vector to obtain a first molten iron shortage time series clustering related weight value;

[0042] Dividing a first power value of a norm of the molten iron shortage time series clustering vector with the ratio value as an exponent by a second power value of a norm of the molten iron shortage time series representation vector with the ratio value as an exponent to obtain a second molten iron shortage time series clustering related weight value;

[0043] For each eigenvalue of the molten iron shortage time series clustering vector, multiply it by the inverse of the difference between the first molten iron shortage time series clustering related weight value and the second molten iron shortage time series clustering related weight value to obtain an optimized eigenvalue of the molten iron shortage time series clustering vector;

[0044] For each eigenvalue outside the cluster in the molten iron shortage time series representation vector, multiply it by the inverse of the sum of the weight value associated with the first molten iron shortage time series cluster and the weight value associated with the second molten iron shortage time series cluster to obtain an optimized out-of-cluster eigenvalue of the molten iron shortage time series representation vector; and

[0045] The optimized eigenvalues ​​of the molten iron shortage time series clustering vector and the optimized out-of-class eigenvalues ​​of the molten iron shortage time series representation vector are combined to form an optimized molten iron shortage time series representation vector.

[0046] That is, the optimization process is expressed as: ;in, is the vector representing the time series of molten iron shortage, is the characteristic number of the hot metal shortage time series representation vector, is the clustering vector of the molten iron shortage time series, is the characteristic number of the hot metal shortage time series clustering vector, represents the clustering feature set corresponding to the molten iron shortage time series clustering vector, and Represents the binary norm and the uninorm of the vector respectively. Power.

[0047] Here, in order to avoid the loss of key suffix semantic information relative to the original feature set as a whole due to aggregation conflict based on the aggregated features of the molten iron shortage timing representation vector, the clustering ratio of the number of eigenvalues ​​of the molten iron shortage timing clustering vector relative to the number of features of the molten iron shortage timing representation vector is used as a decision function to perform an adversarial decision on the set absolute representation of the molten iron shortage timing clustering vector and the one-norm of the molten iron shortage timing representation vector, and positive and negative interactions are performed with the clustering intrinsic conflict representation of the two-norm of the molten iron shortage timing clustering vector and the molten iron shortage timing representation vector respectively to construct a solid alignment guardrail between the optimized molten iron shortage timing representation vector and the original feature set as a whole based on the aggregated features, thereby achieving the mitigation of harmful information loss intention of the optimized molten iron shortage timing representation vector based on the aggregation risk transferability, improving the expression effect of the optimized molten iron shortage timing representation vector, and thus improving the accuracy of the molten iron scheduling decoding value obtained by inputting the molten iron shortage timing representation vector into the decoder-based molten iron scheduling module. In this way, the deep learning model can be used to automatically analyze and process production information in the steel production process to obtain the molten iron scheduling value, so as to adjust the production process in real time, reduce the manual operation time of the dispatchers, and thus significantly improve the scheduling efficiency and intelligence level.

[0048] In particular, the molten iron scheduling execution module 330 is used to perform molten iron scheduling based on the steel production scheduling instruction. In this way, it is possible to ensure that there is sufficient molten iron supply when needed, and optimize the molten iron supply to improve production efficiency.

[0049] As described above, the molten iron scheduling management system 300 for steel production according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a molten iron scheduling management algorithm for steel production. In a possible implementation, the molten iron scheduling management system 300 for steel production according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the molten iron scheduling management system 300 for steel production can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the molten iron scheduling management system 300 for steel production can also be one of the many hardware modules of the wireless terminal.

[0050] Alternatively, in another example, the molten iron scheduling and management system 300 for steel production and the wireless terminal may also be separate devices, and the molten iron scheduling and management system 300 for steel production may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0051] Furthermore, a molten iron scheduling management method for steel production is also provided.

[0052] Figure 4 Flow chart of the molten iron scheduling management method for steel production according to an embodiment of the present application. Figure 4 As shown, the molten iron scheduling management method for steel production according to the embodiment of the present application includes the steps of: S1, collecting production information, wherein the production information includes a time queue of iron output information, a time queue of iron usage information, a time queue of scrap steel adding information, and a time queue of locomotive transportation operation information; S2, performing parameter timing feature capture and significance gated feature aggregation processing on the production information to generate a steel production scheduling instruction; S3, performing molten iron scheduling based on the steel production scheduling instruction.

[0053] In summary, the molten iron scheduling management method for steel production according to the embodiment of the present application is explained, which uses data analysis and processing technology based on deep learning to perform local time series feature analysis and significant aggregation on the production information (iron-producing information, iron-using information, scrap steel adding information and locomotive transportation operation information), so as to automatically decode the molten iron scheduling value based on the time series characteristics of iron consumption obtained by reasoning between the iron-producing information, the iron-using information and the locomotive transportation operation information, and the time series representation between the time series characteristics of the scrap steel adding information. In this way, the deep learning model can be used to automatically analyze and process the production information in the steel production process to obtain the molten iron scheduling value, so as to adjust the production process in real time, reduce the manual operation time of the dispatcher, and thus significantly improve the scheduling efficiency and intelligence level.

[0054] In a specific example of this application, the iron zone MES hot metal dispatching system uses a unified graphical operating platform as the core to achieve collaborative operations of hot metal dispatching-related blast furnace central control, transportation dispatching, steelmaking central control, tank car operation and maintenance, and other positions, and dynamically optimize the entire hot metal dispatching. The specific contents are as follows:

[0055] 1. Hot metal management

[0056] With a unified graphical operating platform as the core, collaborative operations of molten iron-related positions such as blast furnace central control, molten iron scheduling, transportation scheduling, steelmaking central control, and tank truck operation and maintenance are achieved.

[0057] 2. Steelmaking and Steelmaking Plan

[0058] Ironmaking dispatchers need to know the steelmaking production plan in order to better control the ironmaking rhythm.

[0059] 3. Iron-Required Plan for Steelmaking

[0060] Ironmaking scheduling requires iron ladle to provide iron demand plan to guide the iron receiving under the furnace.

[0061] 4. Casting rules configuration

[0062] Maintain the necessary iron-making rules, and the system will make iron-making predictions based on these rules.

[0063] Iron rules are used to meet the basic conditions for realizing the iron tapping prediction function.

[0064] 5. Blast furnace iron production prediction

[0065] According to the configured iron-tapping rules, the system automatically predicts the next iron-tapping information when the blast furnace is opened, providing a reference for iron-making scheduling and tank allocation. Iron-making requires predicting the next iron-tapping status.

[0066] 6. Blast furnace air allocation plan

[0067] Ironmaking dispatch maintains the space allocation plan. After it is issued, the system informs the logistics of the space allocation demand. Logistics allocates space according to the plan to meet production needs. Ironmaking issues the space allocation plan to guide logistics to dispatch vehicles to allocate tanks.

[0068] 7. Recommended air supply for blast furnace

[0069] According to the specific allocation rules, the system generates an allocation recommendation after the molten iron ladle is emptied for the dispatcher to refer to when allocating the space. The system generates an allocation recommendation for the dispatcher to refer to when allocating the space.

[0070] 8. Blast furnace iron production performance

[0071] In the case of data collection, the system automatically obtains the iron and pot information, and provides backup input function in abnormal situations to ensure the smooth tapping of the blast furnace. By maintaining the tapping and iron information and collecting and recording key elements, it can provide a data basis for subsequent production statistics and related indicators, and provide a prerequisite for the smooth settlement of molten iron output.

[0072] 9. Graphical operation of hot metal scheduling

[0073] The ironmaking dispatcher specifies the allocation plan according to the steelmaking steelmaking plan, and the steelmaking dispatcher is responsible for the weight allocation. The system forwards the corresponding information to the logistics system, and the logistics system implements transportation scheduling based on this. The system will record the relevant operation time node information and provide data support for the whole process tracking of molten iron.

[0074] 10. Check the iron receiving history in front of the furnace

[0075] With the iron receiving ID as the primary key, track and query the iron receiving history of the molten iron tank uploaded by blast furnace L2.

[0076] In particular, in the tracking and management of molten iron, it receives the steelmaking tapping plan and the steelmaking iron demand plan, generates an emptying plan, guides the blast furnace tapping and the molten iron tank receiving iron, tracks the opening of the iron, slag discharge, plugging time, the start and end of the tank receiving iron, the weight of the iron, and other complete information, and provides a multi-dimensional summary query function to fully record the molten iron information in front of the furnace. The scheduling of empty tanks and heavy tanks is realized through graphical operation. After receiving the scheduling instructions, the logistics department performs the operation, and at the same time tracks the inspection and measurement results of the quality control department to realize the collection of information on the whole process elements of the molten iron tank transportation process. Combined with the feedback of the emptying, upper and lower shelves and other signals on the steelmaking side, the data tracking of the whole process and all elements of the molten iron from the opening in front of the furnace to the final entry into the steelmaking converter is realized, laying the foundation for the subsequent molten iron data statistics.

[0077] In the maintenance of the molten iron ladle, the system provides basic information of the molten iron ladle, maintenance, ladle operation and maintenance management, and ladle life tracking management functions. During the blast furnace tapping, steelmaking pouring, and other operations, the status of the online molten iron ladle is automatically tracked. According to the relevant performance of the molten iron, the information related to the life of the molten iron ladle is tracked to provide a reference for the timely inspection and maintenance of the molten iron ladle. For the offline repair of the online molten iron ladle or the online operation of the offline molten iron ladle, the status adjustment is achieved through the ladle operation and maintenance management.

[0078] In the maintenance of hot metal ladle, through the information interaction and integration with steelmaking production execution system, process railway vehicle-ground joint control system, iron ladle tracking system, blast furnace L2, metering system, inspection and testing system, the operation information such as iron tapping, iron usage, scrap steel addition, locomotive transportation, etc. from blast furnace iron tapping to steelmaking iron pouring can be tracked. The position, weight, composition and status of hot metal ladle can be tracked in all aspects of iron-steel collaboration.

[0079] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A molten iron scheduling management system for steel production, characterized in that: include: A production information collection module, used to collect production information, wherein the production information includes a time queue of iron-out information, a time queue of iron-use information, a time queue of scrap steel adding information, and a time queue of locomotive transport operation information; A steel scheduling instruction generation module, used for performing parameter time series feature capture and significant gated feature aggregation processing on the production information to generate steel production scheduling instructions; A molten iron scheduling execution module, used for performing molten iron scheduling based on the steel production scheduling instruction; Wherein, the steel scheduling instruction generation module includes: The steel production information time series association encoding unit is used to input the time series of the iron-making information, the time series of the iron-using information, the time series of the scrap steel adding information and the time series of the locomotive transportation operation information into a sequence encoder respectively to obtain a time series of the iron-making information time series association feature vector, a time series of the iron-using information time series association feature vector, a time series of the scrap steel adding information time series association feature vector and a time series of the locomotive transportation operation information time series association feature vector; A steel production information time series correlation feature significant unit, used for inputting the time queue of the iron-out information time series correlation feature vector, the time queue of the iron-use information time series correlation feature vector, the time queue of the scrap steel adding information time series correlation feature vector and the time queue of the locomotive transportation operation information time series correlation feature vector into a feature aggregation network based on feature energy local significance gating to obtain an iron-out information time series significant aggregation representation vector, an iron-use information time series significant aggregation representation vector, a scrap steel adding information time series significant aggregation representation vector and a locomotive transportation operation information time series aggregation representation vector; The iron consumption time series multimodal fusion unit is used to perform time series reasoning based on iron consumption on the iron production information time series significant aggregation representation vector, the iron consumption information time series significant aggregation representation vector and the locomotive transportation operation information time series aggregation representation vector to obtain an iron consumption time series multimodal comprehensive representation vector; A molten iron shortage time series representation unit, used to calculate a molten iron shortage time series representation vector between the multimodal comprehensive representation vector of the iron consumption time series and the significant aggregation representation vector of the scrap steel addition information time series as a molten iron shortage time series representation feature; A molten iron scheduling unit, configured to obtain a molten iron scheduling value as the steel production scheduling instruction based on the molten iron shortage time series representation characteristics; Among them, the steel production information time series association encoding unit is used to: input the time queue of the iron-making information, the time queue of the iron-using information, the time queue of the scrap steel adding information and the time queue of the locomotive transportation operation information into a sequence encoder based on the Bi-LSTM model respectively to obtain the time queue of the iron-making information time series association feature vector, the time queue of the iron-using information time series association feature vector, the time queue of the scrap steel adding information time series association feature vector and the time queue of the locomotive transportation operation information time series association feature vector.

2. The molten iron scheduling and management system for steel production according to claim 1, characterized in that: The steel production information temporal correlation feature significant unit comprises: A subunit for calculating characteristic energy level coefficients of the timing correlation characteristic of the iron-out information, for determining characteristic energy level coefficients of each timing correlation characteristic of the iron-out information based on the maximum value, minimum value, mean value and variance of each timing correlation characteristic of the iron-out information in the time queue of the timing correlation characteristic of the iron-out information to obtain a set of characteristic energy level coefficients of the timing correlation characteristic of the iron-out information; The iron-tapping information time series correlation feature energy level significance descriptor calculation subunit is used to determine the scale of the local neighborhood, calculate the local neighborhood significance descriptor of each iron-tapping information time series correlation feature energy level coefficient in the set of the iron-tapping information time series correlation feature energy level coefficients to obtain the set of the iron-tapping information time series correlation feature energy level significance descriptors; A mask subunit of the energy level significance descriptor of the iron-tapping information timing-related feature is used to input the set of the energy level significance descriptors of the iron-tapping information timing-related feature into a mask module based on a gating function to obtain a set of mask-gated probabilistic energy level significance descriptors of the iron-tapping information timing-related feature; The iron-out information timing significant aggregation subunit is used to calculate the position-weighted sum of the time queue of the iron-out information timing associated feature vector using the set of mask-gated probabilistic iron-out information timing associated feature energy level significance descriptors as a sequence of weights to obtain the iron-out information timing significant aggregation representation vector.

3. The molten iron scheduling management system for steel production according to claim 2, characterized in that: The iron-out information time-series correlation characteristic energy level coefficient calculation subunit is used to: Respectively extracting the maximum value and the minimum value in the iron-tapping information timing-related feature vector to obtain the maximum value of the iron-tapping information timing-related feature and the minimum value of the iron-tapping information timing-related feature; Respectively calculating the mean and variance of the iron-tapping information time-series associated feature vector to obtain the iron-tapping information time-series associated feature mean and the iron-tapping information time-series associated feature variance; Adding the iron-tapping information time series correlation variance and a preset hyperparameter to obtain a first energy level coefficient of the iron-tapping information time series correlation feature; Calculate the square of the difference between the maximum value of the iron-out information timing correlation feature and the mean value of the iron-out information timing correlation feature to obtain the maximum difference value of the iron-out information timing correlation; Calculate the square of the difference between the minimum value of the iron-out information timing correlation feature and the mean value of the iron-out information timing correlation feature to obtain the minimum difference value of the iron-out information timing correlation; Adding the maximum difference value associated with the timing of the iron-out information, the minimum difference value associated with the timing of the iron-out information, and the preset hyperparameter to obtain the second energy level coefficient of the timing-associated feature of the iron-out information; The first energy level coefficient of the iron-out information timing correlation characteristic is divided by the second energy level coefficient of the iron-out information timing correlation characteristic to obtain the energy level coefficient of the iron-out information timing correlation characteristic.

4. The molten iron scheduling management system for steel production according to claim 3, characterized in that: The iron-out information temporal correlation feature energy level significance descriptor mask subunit is used to: Inputting each iron-tapping information time-series associated feature energy level significance descriptor in the set of iron-tapping information time-series associated feature energy level significance descriptors into a Sigmoid function for normalization processing to obtain a set of normalized iron-tapping information time-series associated feature energy level significance descriptors; Each normalized iron-out information timing-related feature energy level significance descriptor in the set of normalized iron-out information timing-related feature energy level significance descriptors is input into a gating function for masking processing to obtain the set of mask-gated probabilistic iron-out information timing-related feature energy level significance descriptors.

5. The molten iron scheduling and management system for steel production according to claim 4, characterized in that: The iron consumption time series multimodal fusion unit is used to: input the iron production information time series significant aggregation representation vector, the iron consumption information time series significant aggregation representation vector and the locomotive transportation operation information time series aggregation representation vector into the iron consumption time series reasoner based on the Bayesian probability network to obtain the iron consumption time series multimodal comprehensive representation vector.

6. The molten iron scheduling and management system for steel production according to claim 5, characterized in that: The molten iron scheduling unit is used to: input the molten iron shortage timing representation vector into a molten iron scheduling module based on a decoder to obtain a molten iron scheduling decoding value.

7. A method for managing molten iron scheduling for steel production, using the molten iron scheduling management system for steel production according to claim 1, characterized in that: include: Collecting production information, wherein the production information includes a time queue of iron-out information, a time queue of iron-use information, a time queue of scrap steel adding information, and a time queue of locomotive transport operation information; Capturing parameter time series features and performing significant gated feature aggregation processing on the production information to generate steel production scheduling instructions; Based on the steel production scheduling instructions, molten iron scheduling is performed.

Images