Long-process iron and steel enterprise production and marketing closed-loop real-time optimization system and method

Abstract

The invention provides a production and marketing closed-loop real-time optimization system and method for a long-process iron and steel enterprise, and belongs to the technical field of iron and steel enterprise informatization management, and the system comprises a cloud server cluster which is integrated with an information subsystem and is configured with a data integration and analysis platform; the edge server is configured with a data collector and a plurality of algorithm models; the local terminal comprises a plurality of user terminal devices and a plurality of sensors; the sensor collects the state of production equipment and production process data, and the algorithm model monitors and controls the production process in real time according to the sensor data and business data in each information subsystem; and the data integration uniformly manages business data of each information subsystem, data collected by each algorithm model and each sensor, training and issuing of each algorithm model, and closed-loop optimization of a production link. According to the invention, integration and optimization of multiple links of enterprise production, sales and inventory are realized, the production efficiency of iron and steel enterprises is improved, and the operation cost is reduced.

Description

Technical Field

[0001] The present application belongs to the field of information management technology for steel enterprises, and specifically relates to a real-time optimization system and method for the production and sales closed loop of long-process steel enterprises. Background Art

[0002] For steel companies with long production processes, real-time optimization of the production and sales closed loop is a key factor in improving production efficiency, reducing operating costs, and enhancing market competitiveness. Traditional steel company management models rely on manual monitoring and decision-making, resulting in information lags and slow responses, failing to meet market demands for rapid response and flexible adjustments. Furthermore, information silos are a serious problem across various production processes, with independent information systems preventing effective data integration and sharing, hindering overall optimization capabilities.

[0003] While existing information systems for steel companies have achieved a degree of automated production, they still face numerous challenges. For example, while sensors can collect real-time data on production equipment status and production processes, this data is often localized, preventing seamless integration between the cloud and the edge. Furthermore, traditional PLC controllers can only execute preset rules and cannot dynamically adjust policies based on real-time data. For example, equipment anomalies require manual intervention to adjust production plans, resulting in extended downtime. Furthermore, model updates rely on offline batch training, making it impossible to optimize parameters in real time based on production execution results, thus forming an "open-loop" decision. Furthermore, existing data integration and analysis methods often only process and analyze data from a single information subsystem, failing to achieve cross-subsystem data integration and collaborative optimization. This results in companies being unable to fully utilize data resources across the entire supply chain when making decisions on production planning, inventory management, sales forecasting, and other areas, resulting in inefficient and inaccurate decisions. Summary of the Invention

[0004] In a first aspect, an embodiment of the present application provides a real-time optimization system for the production and sales closed loop of a long-process steel enterprise, comprising a cloud server cluster, an edge server, and a local terminal; The cloud server cluster integrates several information subsystems and is equipped with a data integration and analysis platform; The edge server is equipped with a data collector and several algorithm models; The local terminal includes several user terminal devices and several sensors; The sensors collect production equipment status and production process data, and the algorithm model monitors and controls the production process in real time based on the sensor data and business data in each information subsystem; The data integration and analysis platform is connected to the edge server, local terminal and each information subsystem, and is used to uniformly manage the business data of each information subsystem, each algorithm model and the data collected by each sensor, the training and distribution of each algorithm model, and the closed-loop optimization of the production link.

[0005] Furthermore, the information subsystem includes an ERP subsystem, an MES subsystem, an e-commerce subsystem, and a daily cost subsystem; The cloud server cluster is also configured with a database management subsystem for storing and maintaining business data in each sub-information system.

[0006] Furthermore, the edge server is also configured with an AI accelerator to accelerate the reasoning process of the algorithm model.

[0007] Furthermore, the algorithm model includes an available capacity calculation model, a sales-production conversion model, an optimal production scheduling model, a low inventory scheduling model, an intelligent production scheduling model and an intelligent warehousing and logistics management model.

[0008] Furthermore, the user terminal device responds to user requests, connects to the data integration and analysis platform to access various information subsystems, and performs production-related operations and task management.

[0009] In a second aspect, the present application also provides a method for real-time optimization of the production and sales closed loop of a long-process steel enterprise, comprising the following steps: S1. Build a data integration and analysis platform in the cloud, integrate various information subsystems, configure edge servers and data acquisition interfaces at the edge, and deploy sensors and user terminal devices at local terminals; S2. Centrally manage sensor data, business data, and algorithm models through a data integration and analysis platform, and train and optimize each algorithm model; S3. The edge server uses algorithm models based on sensor data and business data to monitor and control the production process in real time, achieving closed-loop optimization of production and sales.

[0010] Furthermore, the specific steps of step S1 are as follows: S11. Build a data integration and analysis platform in the cloud, establish API interfaces for each information subsystem, integrate business data, and store it in a unified database management subsystem; S12. Configure the data collection interface and edge server at the edge, and deploy the algorithm model in the edge server; S13. Deploy sensors and user terminal devices at the local terminal, configure the sensors to connect with the data acquisition interface, and configure the user terminal devices to connect with each information subsystem.

[0011] Furthermore, the specific steps of step S2 are as follows: S21. Obtain sensor data in real time through the data acquisition interface on the data integration and analysis platform and perform preprocessing; S22. Acquire business data from various information subsystems and standardize it on the data integration and analysis platform; S23. Train and optimize the algorithm model based on the processed sensor data and business data on the data integration and analysis platform, and send the algorithm model to the edge server.

[0012] Furthermore, the training of the algorithm model in step S23 includes: Establish corresponding algorithm models according to the needs of the production process, use actual constraints as constraints, and use decision content as output variables to train and optimize the algorithm models.

[0013] Furthermore, the specific steps of step S23 are as follows: S231. Use a multi-objective mixed integer programming algorithm to train an available capacity calculation model to optimize annual production planning and maintenance arrangements; S232. Use a two-stage hybrid optimization strategy to train the sales-production conversion model and generate a monthly production plan; S233. Use a multi-objective mixed integer programming framework to train an optimal production scheduling model to optimize the production schedule of rolled product batches and steelmaking heats; S234. Use time series analysis and machine learning to train low inventory scheduling models and dynamically adjust inventory levels; S235. Use adaptive control or reinforcement learning to train intelligent production scheduling models and optimize production schedules; S236. Use warehouse management and logistics scheduling algorithms to train smart warehousing and logistics management models to optimize supply chain processes; S237. Send each algorithm model to the edge server.

[0014] Furthermore, the specific steps of step S3 are as follows: S31. The edge server obtains real-time data on production equipment status and production processes collected by sensors, and combines it with business data from the information subsystem to monitor the production process in real time. S32. The edge server inputs the collected data and business data into the corresponding algorithm model, combines the collaborative optimization relationship between the algorithm models, and performs optimization calculations on each algorithm model; S33. Execute closed-loop optimization of production and marketing based on various algorithm models; S34. The edge server sends the optimization instructions output by each algorithm model to production equipment, warehousing, and logistics systems for execution, collects execution results in real time, and then feeds the execution results back to the corresponding algorithm model, triggering parameter adjustment or re-optimization; S35. Regularly summarize the execution results on the data integration and analysis platform, retrain each algorithm model, and send the updated model parameters to the edge server to complete the closed-loop optimization iteration.

[0015] Furthermore, the collaborative optimization relationship between the algorithm models in step S32 includes the input-output coupling relationship between the algorithm models, the constraint linkage relationship across the algorithm models, the reverse condition relationship of the model parameters, the collaborative response relationship of abnormal events between the algorithm models, and the weight distribution relationship of the algorithm models.

[0016] Furthermore, in step S33, the production and marketing closed-loop optimization based on each algorithm model is specifically as follows: Convert market demand data into executable production plan parameters through sales-production conversion model; Determine the capacity boundary constraints of each production line through the available capacity calculation model; Generate a specific steelmaking-rolling collaborative production plan under capacity constraints by adopting the optimal production scheduling model; Dynamically adjust raw material inventory strategies to match production requirements through low inventory scheduling models; Real-time monitoring and response to production anomalies through intelligent production scheduling models; Optimize material distribution solutions through smart warehousing and logistics management models; Continuously optimize the decision parameters of each model based on daily cost data.

[0017] Furthermore, the specific steps of step S33 are as follows: S331. Input the order data from the e-commerce subsystem business data into the sales-production conversion model, output the monthly production plan and plan adjustment factors, and provide the adjustment factors to the optimal production scheduling model; S332. Input the equipment utilization data collected by sensors, the energy cost and maintenance plan data of the ERP subsystem, and the industrial constraint data of the MES subsystem into the available capacity calculation model, and output the maximum available capacity of each production and sales unit; Combine the maximum available capacity, fixed-point priority, and process constraints with the optimal production scheduling model to obtain the steelmaking furnace plan and rolling batch production scheduling plan, and synchronize the production scheduling plan with the low inventory scheduling model and the intelligent warehousing and logistics management model; S333. Input the planned production output from the optimal production scheduling model, the real-time inventory data from the warehouse management subsystem, and the raw material quality collected by sensors into the low inventory scheduling model. The model then outputs the raw material replenishment quantity and safety stock threshold adjustment strategy, and generates inventory optimization instructions for the warehouse logistics model. S334. Input abnormal signals collected by sensors and current work order information from the MES subsystem into the intelligent production scheduling model, output a production reduction plan or equipment maintenance time window, trigger the recalculation of the optimal scheduling model, and link the low inventory scheduling model to adjust the spare parts strategy; S335. Input the production schedule adjusted by the optimal production scheduling model, the warehouse space occupancy rate collected by sensors, and the transportation resource status of the e-commerce subsystem into the smart warehousing and logistics management model, and output the optimal transportation route and loading priority. S336. Input the actual energy consumption data and production efficiency deviation of the MES subsystem into the daily cost subsystem to generate the capacity model correction coefficient and the optimal scheduling model cost weight adjustment, and feed them back to the available capacity calculation model and the optimal scheduling model.

[0018] It can be seen from the above technical solutions that this application has the following advantages: The real-time optimization system and method for the production and sales closed loop of long-process steel enterprises provided in this application, by building a collaborative architecture of cloud, edge and local terminals, and integrating multiple information subsystems and algorithm models, can efficiently process and analyze massive production data, accurately predict market demand, and achieve deep integration and real-time optimization of multiple links of enterprise production, sales, and inventory, optimize production scheduling and resource allocation, improve the production efficiency of steel enterprises, and reduce operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for the description. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] Figure 1 This is a schematic diagram of the real-time optimization system for the production and marketing closed loop of a long-process steel enterprise according to the present invention.

[0021] Figure 2 This is a flow chart of the real-time optimization method for the production and marketing closed loop of a long-process steel enterprise according to the present invention. DETAILED DESCRIPTION

[0022] Various embodiments of the present disclosure will be described more fully below in the detailed description of the closed-loop production and marketing real-time optimization system for a long-process steel enterprise. The present disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of the present disclosure to the specific embodiments disclosed herein, and that the present disclosure is to be construed as encompassing all adjustments, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of the present disclosure.

[0023] For example, in the current development of long-process steel enterprises, achieving real-time optimization of the production and sales closed loop has become a key to improving production efficiency, reducing operating costs, and enhancing market competitiveness. However, traditional steel enterprise management methods rely primarily on manual monitoring and decision-making. This model inevitably leads to problems such as delayed information transmission and slow response speed, and simply cannot meet the market's urgent demand for rapid response and flexible adjustment. In addition, serious information silos exist between various production links. Various information systems such as ERP and MES are independent of each other, making it difficult to effectively integrate and share data, which greatly limits the further improvement of the company's overall optimization capabilities.

[0024] While existing information systems in steel enterprises have promoted automated production to a certain extent, a number of pressing issues remain. Take sensors, for example. While they can collect real-time data on production equipment status and production processes, this data is often confined to local locations, preventing seamless integration and interaction between the cloud and the edge. Traditional PLC controllers, on the other hand, can only execute preset rules and are unable to dynamically adjust strategies based on real-time data. Furthermore, existing data integration and analysis methods are often limited to processing and analyzing data from a single information subsystem, making it difficult to achieve cross-subsystem data integration and collaborative optimization. This prevents companies from fully tapping into and utilizing data resources across the entire supply chain in key decision-making processes such as production planning, inventory management, and sales forecasting. This results in inefficient decision-making and makes it difficult to ensure accuracy.

[0025] To address the above issues, this embodiment provides a real-time optimization system for the production and sales closed loop of long-process steel enterprises. Through the close integration of cloud server clusters, edge servers and local terminals, it realizes the comprehensive collection, efficient processing and intelligent decision-making of production data. It not only improves the real-time monitoring and control capabilities of the production process, but also optimizes production-related processes through intelligent algorithm models, thereby improving production efficiency and reducing operating costs.

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0027] See also Figure 1 FIG2 is a schematic diagram of a real-time optimization system for a production and marketing closed loop of a long-process steel enterprise in a specific embodiment, wherein the system includes a cloud server cluster, an edge server, and a local terminal; The cloud server cluster integrates several information subsystems and is equipped with a data integration and analysis platform; The edge server is equipped with a data collector and several algorithm models; The local terminal includes several user terminal devices and several sensors; The sensors collect production equipment status and production process data, and the algorithm model monitors and controls the production process in real time based on the sensor data and business data in each information subsystem; The data integration and analysis platform is connected to the edge server, local terminal and each information subsystem, and is used to uniformly manage the business data of each information subsystem, each algorithm model and the data collected by each sensor, the training and distribution of each algorithm model, and the closed-loop optimization of the production link.

[0028] This embodiment achieves the collaboration of data processing, model calculation and real-time monitoring through a layered architecture of cloud server clusters, edge servers and local terminals. By integrating information subsystems on the same platform, it breaks down information silos and achieves seamless data flow and collaborative optimization. Through the PLC controller, sensor data and business data are combined to monitor the status of production equipment in real time and perform optimization control to improve the stability of the production process. The data integration and analysis platform is used to uniformly manage collected data and business data, providing a basis for the training and optimization of algorithm models.

[0029] Furthermore, as a refinement and expansion of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, another long-process steel enterprise production and sales closed-loop real-time optimization system is provided, which includes a cloud server cluster, an edge server, and a local terminal; The cloud server cluster integrates several information subsystems and is equipped with a data integration and analysis platform; The edge server is equipped with a data collector and several algorithm models; The local terminal includes several user terminal devices and several sensors; The sensors collect production equipment status and production process data, and the algorithm model monitors and controls the production process in real time based on the sensor data and business data in each information subsystem; The data integration and analysis platform is connected to the edge server, local terminal and each information subsystem, and is used to uniformly manage the business data of each information subsystem, each algorithm model and each sensor collected data, the training and distribution of each algorithm model, and the closed-loop optimization of the production process; The information subsystem includes ERP subsystem, MES subsystem, e-commerce subsystem and daily cost subsystem; The cloud server cluster is also equipped with a database management subsystem for storing and maintaining business data in each sub-information system; The edge server is also equipped with an AI accelerator to accelerate the inference process of the algorithm model; The algorithm model includes an available capacity calculation model, a sales-production conversion model, an optimal production scheduling model, a low inventory scheduling model, an intelligent production scheduling model, and an intelligent warehousing and logistics management model.

[0030] like Figure 2 As shown, the following is an embodiment of the method for real-time optimization of the production and marketing closed loop of a long-process steel enterprise provided by the embodiment of the present disclosure. This method and the real-time optimization system for the production and marketing closed loop of a long-process steel enterprise of the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiment of the method for real-time optimization of the production and marketing closed loop of a long-process steel enterprise, please refer to the embodiment of the real-time optimization system for the production and marketing closed loop of a long-process steel enterprise.

[0031] The method comprises the following steps: S1. Build a data integration and analysis platform in the cloud, integrate various information subsystems, configure edge servers and data acquisition interfaces at the edge, and deploy sensors and user terminal devices at local terminals; It should be noted that by building a data integration and analysis platform in the cloud and integrating information subsystems, and configuring corresponding devices and models on the edge and local terminals, a foundation is provided for achieving real-time optimization of the production and sales closed loop, enabling data to be circulated and processed between different levels; S2. Centrally manage sensor data, business data, and algorithm models through a data integration and analysis platform, and train and optimize each algorithm model; It should be noted that the data integration and analysis platform manages sensor data and business data in a unified manner, ensuring data accuracy and providing a data foundation for model training and production control. S3. The edge server uses algorithm models based on sensor data and business data to monitor and control the production process in real time, achieving closed-loop optimization of production and sales. It should be noted that the edge server (which can use a PLC controller) controls production-related processes based on data and algorithm models, realizes the automatic intelligence of the production process, and can adjust the production strategy in time according to real-time data and analysis results, improve production efficiency, reduce production costs, and improve product quality.

[0032] This embodiment realizes the comprehensive collection, processing and optimization of production data by building a data integration and analysis platform in the cloud, configuring edge servers and data acquisition interfaces on the edge, and configuring user terminal devices and sensors on the local terminal.

[0033] Furthermore, as a refinement and expansion of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, another method for real-time optimization of the production and sales closed loop of a long-process steel enterprise is provided, which includes: S1. Build a data integration and analysis platform in the cloud, integrate various information subsystems, configure edge servers and data acquisition interfaces at the edge, and deploy sensors and user terminal devices at local terminals. The specific steps of step S1 are as follows: S11. Build a data integration and analysis platform in the cloud, establish API interfaces for each information subsystem, integrate business data, and store it in a unified database management subsystem; S12. Configure the data collection interface and edge server at the edge, and deploy the algorithm model in the edge server; S13. Deploy sensors and user terminal devices in the local terminal, and configure the sensor and data acquisition interface connection, as well as configure the user terminal device to connect with each information subsystem; S2. Use the data integration and analysis platform to uniformly manage sensor data, business data, and algorithm models, and train and optimize each algorithm model. The specific steps of step S2 are as follows: S21. Obtain sensor data in real time through the data acquisition interface on the data integration and analysis platform and perform preprocessing; S22. Acquire business data from various information subsystems and standardize it on the data integration and analysis platform; S23. Train and optimize the algorithm model based on the processed sensor data and business data on the data integration and analysis platform, and send the algorithm model to the edge server; The training of the algorithm model in step S23 includes: Establish corresponding algorithm models according to production requirements, use actual constraints as constraints, and use decision content as output variables to train and optimize the algorithm models; S3. The edge server uses the algorithm model based on sensor data and business data to monitor and control the production process in real time, achieving closed-loop optimization of production and sales. The specific steps of step S3 are as follows: S31. The edge server obtains real-time data on production equipment status and production processes collected by sensors, and combines it with business data from the information subsystem to monitor the production process in real time. S32. The edge server inputs the collected data and business data into the corresponding algorithm model, combines the collaborative optimization relationship between the algorithm models, and performs optimization calculations on each algorithm model; S33. Execute closed-loop optimization of production and marketing based on various algorithm models; The production and marketing closed-loop optimization based on each algorithm model in step S33 is specifically as follows: Convert market demand data into executable production plan parameters through sales-production conversion model; Determine the capacity boundary constraints of each production line through the available capacity calculation model; Generate a specific steelmaking-rolling collaborative production plan under capacity constraints by adopting the optimal production scheduling model; Dynamically adjust raw material inventory strategies to match production requirements through low inventory scheduling models; Real-time monitoring and response to production anomalies through intelligent production scheduling models; Optimize material distribution solutions through smart warehousing and logistics management models; Continuously optimize the decision parameters of each model based on daily cost data; S34. The edge server sends the optimization instructions output by each algorithm model to production equipment, warehousing, and logistics systems for execution, collects execution results in real time, and then feeds the execution results back to the corresponding algorithm model, triggering parameter adjustment or re-optimization; S35. Regularly summarize the execution results on the data integration and analysis platform, retrain each algorithm model, and send the updated model parameters to the edge server to complete the closed-loop optimization iteration.

[0034] In an embodiment of the present invention, based on step S23, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner.

[0035] Step S23. In the data integration and analysis platform, the algorithm model is trained and optimized based on the processed sensor data and business data, and the algorithm model is sent to the edge server; It should be noted that the data integration and analysis platform is based on the processed sensor data and business data of each information subsystem, and establishes corresponding algorithm models according to the needs of different production links; Determine the algorithm of each algorithm model based on the corresponding business data, determine the generation target, set the actual restrictions in production as constraints, and use the decision content of the corresponding production link as the output decision variable. Train and optimize each algorithm model, and send the trained and optimized algorithm model to the edge server, while updating the model regularly. The specific steps of step S23 are as follows: S231. Use a multi-objective mixed integer programming algorithm to train an available capacity calculation model to optimize annual production planning and maintenance arrangements; Specifically, the data integration and analysis platform establishes an available capacity calculation model using a multi-objective mixed integer programming algorithm based on the total production capacity, maintenance period, maintenance period flexibility, production line-related production rules, and production cost data in the business data. This model is trained with the goal of minimizing total cost, with maintenance period flexibility, production line-related rules, and market demand matching as constraints, and with the optimal annual production plan for each production line and the specific maintenance schedule as output decision variables. The data integration and analysis platform uses a multi-objective mixed integer programming algorithm to calculate the optimal annual production plan and specific maintenance schedule for each iron, steel, and material production line; The objective function is constructed to minimize the total cost as follows:

[0036] Among them, c i prod represents the unit production cost, c i inv represents the unit inventory cost, c i maint represents the maintenance cost, Q i,t Denotes cumulative output, D i,t represents the cumulative market demand, x i,t represents the planned output of production line i in month t, in tons, y i,s Indicates whether production line i is under maintenance in month s, y i,s ∈{0,1}, for example, 1 can be used to represent maintenance and 0 to represent production; Set the constraints as follows: Flexible constraints on maintenance period:

[0037] Among them, d i It represents the maintenance period of production line i, which can be measured in days and satisfy the elastic range [d i min ,d i max ]; Production line association rule constraints: Take the production line connection from ironmaking to steelmaking and then to rolling as an example:

[0038] Among them, x represents the planned output of the steel production line in month T, x is the planned output of the iron production line in month T. The steelmaking output shall not exceed α times the molten iron output, which limits the supply of molten iron. Market demand matching constraints:

[0039] x i,t represents the planned output of production line i in month t, represents the total market demand for production line i; The output decision variables are: The planned output of production line i in month t is x i,t , whether production line i was overhauled in month s i,s And the maintenance period d of production line i i ; S232. Use a two-stage hybrid optimization strategy to train the sales-production conversion model and generate a monthly production plan; Specifically, the data integration and analysis platform uses historical sales data, customer order requirements, and market demand forecasts in business data, combined with daily costs and strategic product output indicator constraints, to establish a sales-to-production conversion model using a two-stage hybrid optimization strategy algorithm. This model is trained with market demand forecasting and capacity allocation optimization as its goals, product structure and production plans as constraints, and monthly production plan generation as its output decision variable. Generate monthly production plan using two-stage hybrid optimization strategy algorithm; Phase 1: Multi-product demand forecasting based on time series decomposition and deep learning: The first step is time series decomposition and feature fusion, which decomposes the original sales volume into three parts: trend, season, and residual:

[0040] in, is the trend term, which is modeled using a piecewise linear function; is the seasonal term, which is fitted using Fourier series; is the residual term; Then use LSTM and attention mechanism to predict the residual; Residual sequence , order volume and market indicators is the input; Update the LSTM state using the following formula:

[0041] Update the attention mechanism weights using the following formula:

[0042] The prediction output is as follows:

[0043] Finally, the final demand forecast formula is as follows:

[0044] Phase 2: Multi-product capacity allocation optimization using mixed integer nonlinear programming algorithm; S233. Use a multi-objective mixed integer programming framework to train an optimal production scheduling model to optimize the production schedule of rolled product batches and steelmaking heats; Specifically, the data integration and analysis platform uses customer order and order forecast information in business data, combined with the process scheduling sequence characteristics and variety-related production characteristics of the production lines for different specifications of products, to establish an optimal production scheduling model using a multi-objective mixed integer programming framework. This model is trained with demand satisfaction, optimal cost, and minimum residual material as the goals, and with billet hot delivery requirements and rolled product batch and steelmaking furnace scheduling as constraints, and with cut-to-length combination optimization, hot delivery connection, and variety mixing as output decision variables. The objective function is constructed as follows:

[0045] in, represents the output of production line p in month m, represents the inventory of production line p in month m, Indicates whether emergency production capacity is enabled in month m, c p prod represents the unit production cost of production line p, c p inv represents the unit inventory cost of production line p, represents the unit cost of emergency production capacity; Set the constraints as follows: Capacity Limit Constraints:

[0046] in, is the basic production capacity in month m, is the emergency capacity limit for month m, represents the resource consumption coefficient per unit product p, represents the output of product p in month m, Indicates whether emergency capacity is enabled in month m, for example =1 means enabled, =0 means not enabled; Inventory dynamic balance constraints:

[0047] in, represents the inventory of production line p in month m, represents the sales volume of production line p in month m, represents the output of product p in month m; Constraints on supply guarantee of strategic products:

[0048] in, represents the minimum annual output requirement of strategic product p, represents the output of strategic product p in month m; Process coupling constraints:

[0049] in, represents the output of product p in month m, represents the output of product q in month m, and product p is related to product q in terms of process. It represents the process coupling coefficient, which describes the degree of process correlation between product p and product q, that is, to produce a certain amount of product q, a corresponding amount of product p needs to be produced; It should be noted that the model integrates information such as customer orders and order forecasts, and combines the process scheduling sequence characteristics and variety-related production characteristics of the production line for products of different specifications to optimize the hot delivery demand of steel billets, so as to generate the optimal rolling batch and steelmaking furnace production plan with the goals of meeting demand, optimizing cost, and minimizing residual materials. The model realizes the collaborative optimization of the entire steelmaking-rolling process through a multi-objective mixed integer programming (MIP) framework. The core algorithm design includes: steel billet demand matching, using a dynamic programming algorithm with a penalty term to optimize the length combination and minimize residual materials; hot delivery connection, constructing a bipartite graph based on time window constraints, and maximizing the direct hot delivery matching of furnaces and rolling batches through the Hungarian algorithm; variety mixing, using the improved DBSCAN clustering algorithm to achieve the optimal furnace grouping under the isolation of conflicting varieties; layered solution strategy, the short-term layer uses the branch and bound method to accurately solve the MIP model, and the long-term prediction layer combines LSTM time series prediction with the tabu search algorithm for dynamic rolling optimization. S234. Use time series analysis and machine learning to train low inventory scheduling models and dynamically adjust inventory levels; Specifically, the data integration and analysis platform uses inventory information on raw materials, work-in-progress, and finished products from business data, combined with historical sales data, market forecast data, and external factors, to establish a low-inventory scheduling model using time series analysis and machine learning algorithms. This model is trained with the goal of dynamically adjusting inventory levels, using safety stock strategies and reorder point calculations as constraints, and using inventory replenishment plans and demand forecasts as output decision variables. S235. Use adaptive control or reinforcement learning to train intelligent production scheduling models and optimize production schedules; Specifically, the data integration and analysis platform uses the abnormal information of production plan execution in the collected data and combines it with the production schedule determined in the business data to establish an intelligent production scheduling model using adaptive control algorithms or reinforcement learning methods. The model is trained with the goals of meeting order requirements, minimizing the impact of abnormalities, and optimizing production scheduling, with real-time adjustment of production plans and abnormality handling as constraints, and production schedule optimization and task scheduling as output decision variables. S236. Use warehouse management and logistics scheduling algorithms to train smart warehousing and logistics management models to optimize supply chain processes; Specifically, the data integration and analysis platform establishes an intelligent warehousing and logistics management model based on inventory status, transportation information, and customer order data in business data and collected data, combined with historical sales data, seasonal fluctuations, and market trends. This model is trained with the goal of optimizing inventory management and logistics, with minimizing transportation costs, shortening delivery time, and maximizing customer satisfaction as constraints, and with the optimal transportation route and inventory replenishment strategy as output decision variables. S237. Send each algorithm model to the edge server.

[0050] In an embodiment of the present invention, based on step S32 and step S33, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner.

[0051] The collaborative optimization relationship between the algorithm models in step S32 includes the input-output coupling relationship between the algorithm models, the constraint linkage relationship across the algorithm models, the reverse condition relationship of the model parameters, the collaborative response relationship of abnormal events between the algorithm models, and the weight distribution relationship of the algorithm models; Input-output coupling relationship between each algorithm model: For example, in the collaborative scenario of production scheduling and inventory scheduling, The rolling stock batch plan output by the optimal production scheduling model needs to be synchronized to the low inventory scheduling model in real time. The low inventory scheduling model dynamically calculates the safety stock threshold based on the raw material consumption rate in the plan:

[0052] in, is the production consumption of the i-th type of raw materials, ρ i is the process loss rate, k The volatility factor is trained from historical data; It should be noted that the input-output coupling relationship between the optimal production scheduling model and the low inventory scheduling model can avoid raw material shortages or backlogs caused by the disconnection between production scheduling and inventory; The linkage relationship between constraints across algorithm models: Taking the scenario of coordination between capacity allocation and logistics scheduling as an example, The maximum available capacity output by the capacity calculation model can be used as an input constraint for the smart warehousing and logistics management model, and the transportation priority rules can be dynamically adjusted as follows: IF Production Line A Capacity Utilization ≥ 90% THEN The logistics model uses an emergency channel to transport finished steel coils from production line A. ELSE Delivery by conventional cost-optimal route It should be noted that this constraint linkage can ensure that finished products from high-load production lines are shipped out in a timely manner, avoiding production stoppages due to warehouse congestion; Inverse conditional relationship of model parameters: Taking the closed-loop optimization scenario of production cost and production scheduling as an example, The unit energy consumption cost deviation value Δ calculated in real time by the daily cost subsystem C Energy consumption is inputted into the optimal production scheduling model in reverse order to dynamically adjust the cost weight coefficient of the production scheduling plan:

[0053] This leads to the scheduling model giving priority to low-energy production sequences; By setting the inverse relationship of the model parameters, cost-sensitive scheduling can be achieved to respond to energy price fluctuations; The collaborative response relationship between abnormal events and algorithm models: Take the emergency dispatch scenario when equipment fails as an example; When the intelligent production scheduling model detects a production line fault, such as one triggered by sensor vibration spectrum analysis, the following linkage is executed: Optimal production scheduling model: recalculates the production schedule for the remaining production lines to ensure the delivery of key orders; Low inventory scheduling model: Initiate the backup supplier procurement process and freeze the inventory allocation of raw materials associated with the faulty production line; Smart warehousing and logistics management model: allocate dedicated transport queues for urgent procurement of raw materials; Through the coordinated response relationship between algorithm models for equipment failure anomalies, multi-model collaboration shortens the impact of failures from an average of 48 hours to 12 hours; Weight distribution relationship of the algorithm model: Take the model priority adjustment scenario when market demand changes suddenly as an example. When the sales-production conversion model detects that the market demand deviation rate exceeds 15%, it sends weight adjustment instructions to each algorithm model through the data integration and analysis platform: The order delivery timeliness weight of the optimal scheduling model increased by 30%; The safety stock threshold weight of the low inventory scheduling model is reduced by 20%; The logistics management model enables on-time delivery priority mode; The weight distribution relationship of the algorithm model can quickly respond to market changes and avoid decision delays caused by model target conflicts; Specifically, step S33 includes the following steps: S331. Input the order data from the e-commerce subsystem business data into the sales-production conversion model, output the monthly production plan and plan adjustment factors, and provide the adjustment factors to the optimal production scheduling model; Use the following sales-production conversion model to trigger production scheduling adjustments:

[0054] The order data of the e-commerce subsystem p,t and market demand forecasts Converted into monthly production plan and adjustment factors ; For example, the order data O 高强钢,t =1000 tons, e-commerce order data O 钢卷,t= 1000t, LSTM prediction residuals =50 tons input small production conversion model, the output is as follows: Monthly production plan X 钢卷,t =1050 tons, adjustment factor α =1.05; Thus, forecast errors can be compensated by dynamically adjusting factors to ensure that production plans meet actual needs; S332. Input the equipment utilization data collected by sensors, the energy cost and maintenance plan data of the ERP subsystem, and the industrial constraint data of the MES subsystem into the available capacity calculation model, and output the maximum available capacity of each production and sales unit; Use the following capacity boundary constraint formula of the available capacity calculation model

[0055] Equipment utilization Energy costs As input, calculate the maximum available capacity of each production line ; For example, the equipment utilization rate of production line A is =85%, energy cost =500 yuan / ton Input the capacity calculation model and get the output: Maximum available capacity = 800 tons (subject to equipment load and energy quota); Finally, it provides hard capacity boundaries for the scheduling model to avoid overload production; Combine the maximum available capacity, fixed-point priority, and process constraints with the optimal production scheduling model to obtain the steelmaking furnace plan and rolling batch production scheduling plan, and synchronize the production scheduling plan with the low inventory scheduling model and the intelligent warehousing and logistics management model; Steelmaking-rolling coordination formula using optimal production scheduling model

[0056] Calculate the optimal rolling batch under capacity constraints and steelmaking heats Production scheduling plan; For example, the production capacity of production line A is =800 tons, billet demand =500 tons Input the optimal production scheduling model and get the output: steelmaking heat plan = 2 furnaces, rolling batch =2 batches; Finally, through hot delivery connection and mixed casting strategy, the waste of residual materials was reduced (the residual material rate was reduced from 8% to 3%). S333. Input the planned production output from the optimal production scheduling model, the real-time inventory data from the warehouse management subsystem, and the raw material quality collected by sensors into the low inventory scheduling model. The model then outputs the raw material replenishment quantity and safety stock threshold adjustment strategy, and generates inventory optimization instructions for the warehouse logistics model. Safety stock dynamic adjustment formula using low inventory scheduling model

[0057] According to the production schedule and raw material consumption rate Dynamically adjust inventory strategies; For example, the iron ore consumption will be scheduled =5000 tons / week, fluctuation factor =1.2 Input the low inventory scheduling model and get the output: safety stock threshold Ssafe = 12,000 tons (20% increase over the baseline value); Finally, cope with fluctuations in raw material prices and avoid production stoppages due to supply disruptions; S334. Input abnormal signals collected by sensors and current work order information from the MES subsystem into the intelligent production scheduling model, output a production reduction plan or equipment maintenance time window, trigger the recalculation of the optimal scheduling model, and link the low inventory scheduling model to adjust the spare parts strategy; Abnormal response formula using intelligent production scheduling model

[0058] Sensor-based vibration spectrum and ticket priority , triggering production reduction or maintenance instructions; For example, the rolling mill vibration value =95dB (exceeds the threshold of 80dB), current work order priority =High-input intelligent production scheduling model, output: triggering a production reduction plan (20% reduction) and dispatching a spare rolling mill to take over the order; Ultimately, unplanned downtime was reduced (downtime was reduced from 4 hours to 1 hour); S335. Input the production schedule adjusted by the optimal production scheduling model, the warehouse space occupancy rate collected by sensors, and the transportation resource status of the e-commerce subsystem into the smart warehousing and logistics management model, and output the optimal transportation route and loading priority. S336. Input the actual energy consumption data and production efficiency deviation of the MES subsystem into the daily cost subsystem to generate the capacity model correction coefficient and the optimal scheduling model cost weight adjustment, and feed them back into the available capacity calculation model and the optimal scheduling model; It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0059] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A real-time optimization system for production and marketing closed loop of long-process steel enterprises, characterized by: Includes cloud server clusters, edge servers, and local terminals; The cloud server cluster integrates several information subsystems and is equipped with a data integration and analysis platform; The edge server is equipped with a data collector and several algorithm models; The local terminal includes several user terminal devices and several sensors; The sensors collect production equipment status and production process data, and the algorithm model monitors and controls the production process in real time based on the sensor data and business data in each information subsystem; The data integration and analysis platform is connected to the edge server, local terminal and each information subsystem, and is used to uniformly manage the business data of each information subsystem, each algorithm model and the data collected by each sensor, the training and distribution of each algorithm model, and the closed-loop optimization of the production link.

2. The long-process steel enterprise production and marketing closed-loop real-time optimization system according to claim 1 is characterized in that: The information subsystem includes ERP subsystem, MES subsystem, e-commerce subsystem and daily cost subsystem; The cloud server cluster is also configured with a database management subsystem for storing and maintaining business data in each sub-information system.

3. The real-time optimization system for production and marketing closed loop of a long-process steel enterprise according to claim 1 is characterized in that: The edge server is also equipped with an AI accelerator to accelerate the reasoning process of the algorithm model.

4. The long-process steel enterprise production and marketing closed-loop real-time optimization system according to claim 1 is characterized in that: The algorithm model includes an available capacity calculation model, a sales-production conversion model, an optimal production scheduling model, a low inventory scheduling model, an intelligent production scheduling model, and an intelligent warehousing and logistics management model.

5. The long-process steel enterprise production and marketing closed-loop real-time optimization system according to claim 1 is characterized in that: User terminal devices respond to user requests, connect to the data integration and analysis platform to access various information subsystems, and perform production-related operations and task management.

6. A real-time optimization method for the production and marketing closed loop of a long-process steel enterprise, characterized in that: The steps include: S1. Build a data integration and analysis platform in the cloud, integrate various information subsystems, configure edge servers and data acquisition interfaces at the edge, and deploy sensors and user terminal devices at local terminals; S2. Centrally manage sensor data, business data, and algorithm models through a data integration and analysis platform, and train and optimize each algorithm model; S3. The edge server uses algorithm models based on sensor data and business data to monitor and control the production process in real time, achieving closed-loop optimization of production and sales.

7. The method for real-time optimization of the production and marketing closed loop of a long-process steel enterprise according to claim 6, characterized in that: The specific steps of step S1 are as follows: S11. Build a data integration and analysis platform in the cloud, establish API interfaces for each information subsystem, integrate business data, and store it in a unified database management subsystem; S12. Configure the data collection interface and edge server at the edge, and deploy the algorithm model in the edge server; S13. Deploy sensors and user terminal devices at the local terminal, configure the sensors to connect with the data acquisition interface, and configure the user terminal devices to connect with each information subsystem.

8. The method for real-time optimization of the production and marketing closed loop of a long-process steel enterprise according to claim 6, characterized in that: The specific steps of step S2 are as follows: S21. Obtain sensor data in real time through the data acquisition interface on the data integration and analysis platform and perform preprocessing; S22. Acquire business data from various information subsystems and standardize it on the data integration and analysis platform; S23. Train and optimize the algorithm model based on the processed sensor data and business data on the data integration and analysis platform, and send the algorithm model to the edge server.

9. The method for real-time optimization of the production and marketing closed loop of a long-process steel enterprise according to claim 8, characterized in that: The training of the algorithm model in step S23 includes: Establish corresponding algorithm models according to the needs of the production process, use actual constraints as constraints, and use decision content as output variables to train and optimize the algorithm models.

10. The method for real-time optimization of the production and marketing closed loop of a long-process steel enterprise according to claim 8, characterized in that: The specific steps of step S23 are as follows: S231. Use a multi-objective mixed integer programming algorithm to train an available capacity calculation model to optimize annual production planning and maintenance arrangements; S232. Use a two-stage hybrid optimization strategy to train the sales-production conversion model and generate a monthly production plan; S233. Use a multi-objective mixed integer programming framework to train an optimal production scheduling model to optimize the production schedule of rolled product batches and steelmaking heats; S234. Use time series analysis and machine learning to train low inventory scheduling models and dynamically adjust inventory levels; S235. Use adaptive control or reinforcement learning to train intelligent production scheduling models and optimize production schedules; S236. Use warehouse management and logistics scheduling algorithms to train smart warehousing and logistics management models to optimize supply chain processes; S237. Send each algorithm model to the edge server.

11. The method for real-time optimization of the production and marketing closed loop of a long-process steel enterprise according to claim 8, characterized in that: The specific steps of step S3 are as follows: S31. The edge server obtains real-time data on production equipment status and production processes collected by sensors, and combines it with business data from the information subsystem to monitor the production process in real time. S32. The edge server inputs the collected data and business data into the corresponding algorithm model, combines the collaborative optimization relationship between the algorithm models, and performs optimization calculations on each algorithm model; S33. Execute closed-loop optimization of production and marketing based on various algorithm models; S34. The edge server sends the optimization instructions output by each algorithm model to production equipment, warehousing, and logistics systems for execution, collects execution results in real time, and then feeds the execution results back to the corresponding algorithm model, triggering parameter adjustment or re-optimization; S35. Regularly summarize the execution results on the data integration and analysis platform, retrain each algorithm model, and send the updated model parameters to the edge server to complete the closed-loop optimization iteration.

12. The method for real-time optimization of the production and marketing closed loop of a long-process steel enterprise according to claim 11, characterized in that: The collaborative optimization relationship between the algorithm models in step S32 includes the input-output coupling relationship between the algorithm models, the constraint linkage relationship across the algorithm models, the reverse condition relationship of the model parameters, the collaborative response relationship of abnormal events between the algorithm models, and the weight distribution relationship of the algorithm models.

13. The method for real-time optimization of the production and marketing closed loop of a long-process steel enterprise according to claim 11, characterized in that: The production and marketing closed-loop optimization based on each algorithm model in step S33 is specifically as follows: Convert market demand data into executable production plan parameters through sales-production conversion model; Determine the capacity boundary constraints of each production line through the available capacity calculation model; Generate a specific steelmaking-rolling collaborative production plan under capacity constraints by adopting the optimal production scheduling model; Dynamically adjust raw material inventory strategies to match production requirements through low inventory scheduling models; Real-time monitoring and response to production anomalies through intelligent production scheduling models; Optimize material distribution solutions through smart warehousing and logistics management models; Continuously optimize the decision parameters of each model based on daily cost data.

14. The method for real-time optimization of the production and marketing closed loop of a long-process steel enterprise according to claim 11, characterized in that: The specific steps of step S33 are as follows: S331. Input the order data from the e-commerce subsystem business data into the sales-production conversion model, output the monthly production plan and plan adjustment factors, and provide the adjustment factors to the optimal production scheduling model; S332. Input the equipment utilization data collected by sensors, the energy cost and maintenance plan data of the ERP subsystem, and the industrial constraint data of the MES subsystem into the available capacity calculation model, and output the maximum available capacity of each production and sales unit; Combine the maximum available capacity, fixed-point priority, and process constraints with the optimal production scheduling model to obtain the steelmaking furnace plan and rolling batch production scheduling plan, and synchronize the production scheduling plan with the low inventory scheduling model and the intelligent warehousing and logistics management model; S333. Input the planned production output from the optimal production scheduling model, the real-time inventory data from the warehouse management subsystem, and the raw material quality collected by sensors into the low inventory scheduling model. The model then outputs the raw material replenishment quantity and safety stock threshold adjustment strategy, and generates inventory optimization instructions for the warehouse logistics model. S334. Input abnormal signals collected by sensors and current work order information from the MES subsystem into the intelligent production scheduling model, output a production reduction plan or equipment maintenance time window, trigger the recalculation of the optimal scheduling model, and link the low inventory scheduling model to adjust the spare parts strategy; S335. Input the production schedule adjusted by the optimal production scheduling model, the warehouse space occupancy rate collected by sensors, and the transportation resource status of the e-commerce subsystem into the smart warehousing and logistics management model, and output the optimal transportation route and loading priority. S336. Input the actual energy consumption data and production efficiency deviation of the MES subsystem into the daily cost subsystem to generate the capacity model correction coefficient and the optimal scheduling model cost weight adjustment, and feed them back to the available capacity calculation model and the optimal scheduling model.

Images