Multi-agent-based industrial Internet of Things production chain and industry chain collaboration method

Abstract

The invention provides a multi-agent-based industrial Internet of Things production chain and industry chain collaboration method, relates to the technical field of industrial Internet, and aims to provide a multi-agent-based industrial Internet of Things production chain and industry chain collaboration method under constraint conditions of elastic supply and dynamic demands of an external market, internal dynamic production conditions (equipment capability and process parameters) of an enterprise, external resource consumption, carbon emission and the like. In order to improve the comprehensive evaluation indexes of the whole production process including the indexes of product yield, quality, energy consumption, emission, cost and the like, a virtual manufacturing process is adopted to realize feedforward decision correction based on virtual simulation, and working condition identification and feedback are carried out through large-scale sensing data in industrial production. And self-optimization decision making and man-machine interaction dynamic optimization decision making are realized, so that an industrial chain and production chain cooperation mechanism based on multiple agents is established.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial Internet, and specifically relates to a multi-agent based industrial Internet of Things production chain and industrial chain collaboration method. Background Art

[0002] During the operation of the steel industry production chain and industrial chain, a large amount of multi-source heterogeneous data (existing data or data that will occur in the future) will be generated, and there will be a wide range of optimization and control decisions. To this end, it is necessary to build an intelligent system for steel enterprises, with the smart factory brain composed of data centers, business centers, and algorithm centers as the core. Through industrial chain collaboration, value chain extension, and ecological integration, the full process and full-link steel intelligent management from suppliers, production bases, production and manufacturing production lines to customers will be improved to achieve the goals of intelligent production, networked collaboration, personalized customization, and service extension of steel enterprises, so as to promote the optimization of the operation system, the coordination of resource advantages, and the innovation of production models, and enhance the competitiveness and advantages of the entire steel enterprise.

[0003] The Chinese patent "CN202310285877.9 A multi-agent collaboration method based on a hierarchical communication mechanism" provides a multi-agent collaboration method based on a hierarchical communication mechanism. First, the intra-group communication topology is established, then the inter-group communication topology is established, and then the intra-group information aggregation and inter-group information exchange are performed, and finally the action decision is made. By establishing a reasonable and efficient hierarchical communication structure between agents, the centralized data acquisition, aggregation and utilization are broken, and the dimensionality disaster problem of centralized training is alleviated.

[0004] The Chinese patent "CN202310895639.X Multi-agent Collaborative Computing Resource Scheduling Method, Device and System" discloses a multi-agent collaborative computing resource scheduling method. First, the task request queue is configured to obtain the historical operation state sequence of the edge cluster, which is input into the multi-layer superimposed LSTM neural network, and the system state matrix with time series characteristics is output. The state matrix is ​​then input into the Actor-Critic network configured by the edge cluster for multi-agent reinforcement learning to complete resource scheduling; then the loss function and gradient are calculated based on the task return to update the Actor-Critic network parameters. When facing large-scale service requests, it can successfully learn the periodic state between requests, improve the system throughput, and use a combination of time series network and strategy to schedule each task, which has faster convergence speed and less data required for training.

[0005] The above is the traditional multi-agent coordination mechanism. Although it can achieve coordination between agents at the same level, it cannot meet the needs of multi-level coordination of the steel industry's production chain and industrial chain and the high-end, intelligent and green development of the steel industry. Summary of the invention

[0006] In view of the shortcomings of the existing technology, the present invention designs a multi-agent-based industrial Internet of Things production chain and industrial chain collaboration method. Based on the three-layer intelligent agent, the production equipment intelligent agent adaptively makes optimization decisions, optimizes resource allocation, and reasonably allocates and recycles energy, which is the bottom-level intelligent agent. The production chain-level intelligent agent collaboratively controls the production equipment intelligent agent to achieve the optimal allocation of process resources and worry-free production in the workshop. The industrial chain-level intelligent agent and the production chain-level intelligent agent collaborate with each other to achieve global optimization of the manufacturing process;

[0007] A production chain and industrial chain collaboration method based on multi-agent industrial Internet of Things, which consists of three parts: production equipment agent, production chain-level collaborative agent, and industrial chain-level resource allocation agent;

[0008] The production equipment agent corresponds to the production equipment, including blast furnace agent, converter agent, and rolling mill agent, and completes their respective production goals; the production chain-level agent includes the workshop production plan agent and the workshop operation plan agent, which, on the one hand, generates the operation plan of each process production equipment, controls the production equipment agent to realize the production of the underlying production equipment, and on the other hand, obtains the production plan of the enterprise from the industrial chain-level agent and is controlled by the industrial chain-level agent; the industrial chain-level agent includes the material procurement plan agent, the product marketing plan agent, and the manufacturing resource allocation plan agent, which optimizes the overall production process of the enterprise, determines the production plan of each process, and guides each process to complete the production operation;

[0009] The production equipment agent, production chain-level collaborative agent, and industrial chain-level resource allocation agent adopt the contract net protocol and CNP mechanism for communication and decision-making; the contract net consists of n nodes, which are divided into two types: managers and agents. In the contract net, managers and agents exchange dynamic messages through negotiation. The decision-making of the contract net includes the following four basic steps:

[0010] Step 1: The manager releases the task for bidding;

[0011] Step 2: The agent submits its bidding information to the manager, including capacity and production cost;

[0012] Step 3: The manager selects an agent to perform the task based on the principle of maximizing the company's profits;

[0013] Step 4: Finally, the manager confirms the contract and finally completes the bidding-competitive decision-making process;

[0014] The functions of the production equipment intelligent agent are: to obtain production data information of each equipment through sensors, including: production efficiency, production benefits, product quality, operation status of production equipment, environmental changes, material types and quality, energy resources, human resources, and equipment resources; to intelligently perceive production working condition information, including: material input, energy consumption, and carbon emissions; and to use neural network models to adaptively make optimization decisions, optimize resource allocation, and reasonably allocate and recycle energy; production equipment intelligent agents adopt communication and information sharing strategies, and effective communication and information sharing are required between intelligent agents in order to understand each other's status, intentions, and action plans. Communication is achieved by directly transmitting messages, sharing knowledge bases, or sensor data.

[0015] The blast furnace agent, converter agent, and rolling mill agent have the following functions:

[0016] (1) Blast furnace agent: Each blast furnace agent bids for its own production task from the workshop operation planning agent, while sensing production data and production conditions;

[0017] (2) Converter Agent: Each converter agent bids for its own production task from the workshop operation planning agent and senses its own production data and production conditions.

[0018] (3) Rolling mill agent: Each rolling mill agent bids for its own production tasks from the workshop operation planning agent, and senses its own production data and production conditions;

[0019] The production chain-level intelligent agent receives the order production task and uses the contract network to make independent decisions and formulate cooperation strategies. The decision-making and cooperation strategy involve the division of tasks, which requires comprehensive consideration of the overall goal and individual interests; the workshop production planning agent and the workshop operation planning agent of the ironmaking, steelmaking and rolling processes of each process collaboratively control the production equipment intelligent agent to achieve worry-free production in the workshop; the workshop production planning agent formulates the workshop production plan according to the production capacity of each process transmitted by the production equipment through the JADE platform and the enterprise production model; the workshop operation planning agent gives full play to the production capacity of each process, organizes production reasonably, ensures the schedulability of the resources of each process, reasonably arranges the operations of each process, realizes the coordinated matching of the operation of each process, and ensures the smooth operation of the whole workshop production;

[0020] The workshop production planning agent and the workshop operation planning agent achieve the global goal through mutual cooperation, and their functions are as follows:

[0021] (1) Workshop production planning agent: Use the contract network protocol to bid to the manufacturing resource allocation planning agent to obtain the workshop production plan;

[0022] (2) Workshop operation plan agent: bid for each production equipment agent and select a production equipment agent to complete the workshop operation plan;

[0023] The industry chain-level agent achieves the overall goal by coordinating the production of each process. It coordinates with upstream and downstream enterprises to complete material procurement and product marketing plans through the material procurement plan agent, product marketing plan agent, and manufacturing resource allocation plan agent. At the same time, it prioritizes the orders accepted by the enterprise to improve the efficiency of the entire enterprise.

[0024] The functions of the material procurement planning agent, product marketing planning agent, and manufacturing resource allocation planning agent are as follows:

[0025] (1) Material Procurement Planning Agent: The production of an enterprise depends on the materials of upstream enterprises. Therefore, the enterprise bids for upstream enterprises according to the Material Procurement Planning Agent to complete the procurement of required materials.

[0026] (2) Product Marketing Plan Agent: The production of downstream enterprises depends on the enterprise's products. Therefore, the enterprise bids for downstream enterprises according to the product marketing plan agent to complete the enterprise's product marketing plan;

[0027] (3) Manufacturing resource allocation planning agent: responsible for generating production plans based on the production orders received by the enterprise, and sending the orders generated by the production plan to the production chain-level workshop production planning agent, and completing the bidding-bidding with the production chain-level intelligent agent;

[0028] The production chain and industrial chain collaboration method is implemented using the JADE framework of Java. The basic functions of the JADE framework include the agent AMS responsible for supervising and managing the access and use process of the platform, and the directory service agent DF and the message transmission system that provide the default yellow pages. Among them, AMS is used to supervise and control the Agent platform, and each agent must register with AMS to obtain a valid AID; DF provides the default yellow pages service for the Agent; the message transmission system is a software component in the platform that controls all message exchanges; Agent interaction and communication are carried out through the above functions, and the MySQL database is used to realize information access; the production equipment intelligent agent, the production chain-level intelligent agent, and the industrial chain-level intelligent agent are created through the basic functions of the JADE platform, and the information interaction between different Agents is completed through the MySQL database and the message transmission system in the JADE framework, and the identity identification between Agents is realized through the AID in the JADE framework;

[0029] A multi-agent-based industrial Internet of Things production chain and industrial chain collaboration method, which also includes the following contents:

[0030] Establishing an enterprise production model provides a basis for the decision-making behavior of each agent. Specifically, set the enterprise's planned production time to T, the product types include hot-rolled plates, hot-rolled flat coils, M types of wire rods, N types of bars, and two ironmaking equipments, a large blast furnace and a small blast furnace. The molten iron output of the large blast furnace on the tth day is F. Iron_b (t), the unit production cost of a large blast furnace is UC Iron_b , the molten iron output of the small blast furnace on day t is F Iron_s (t), the unit production cost of a small blast furnace is UC Iron_s There are two steelmaking equipments, converter No. 1 and converter No. 2. The steel output of converter No. 1 on day t is F Steel_o (t), the unit production cost of converter No. 1 is UC Steel_o , the steel output of converter No. 2 on day t is F Steel_t (t), the unit production cost of converter No. 2 is UC Steel_t ; There is one hot rolling mill production line, and the output of the hot rolling mill on day t is The unit production cost of hot rolled plate is UC HR There is one flattening and coiling machine production line, and the output of the flattening and coiling machine on day t is The unit production cost of hot-rolled flat-rolled plate is UC Smooth ; There are three bar production lines. The output of the nth specification bar of the jth bar production line on the tth day is The unit production cost of the nth specification bar is There are two wire production lines. The output of the wire of specification m on day t of the j-th wire production line is The unit production cost of the mth specification bar is The demand for hot rolled plates is D HR , the price of hot rolled plate is Up HR , the demand for hot-rolled flat coils is D Smooth , hot rolled flat coil price is Up Smooth , the demand for bars of the nth specification is The price of the nth specification wire is The demand for wire of the mth specification is The price of the mth specification wire is The objective function of the enterprise production model is to maximize production profit, that is:

[0031]

[0032] In the production process, the material balance of ironmaking and steelmaking, the material balance of hot rolling and flat-rolling, and the material balance of wire and bar rolling are used as the limiting conditions of the enterprise production model;

[0033] The material balance of ironmaking and steelmaking meets the following constraints: (1) The total daily molten iron output of the large and small blast furnaces is equal to the steel output of the No. 1 converter and the No. 2 converter, that is, F Iron_b (t)-F Steel_o (t) = F Steel_t (t)-F Iron_s (t)t∈T; (2) The daily output of billets and slabs is equal to the sum of the steel output of converter No. 1 and converter No. 2, that is, in is the output of slab on day t, which is used to produce hot-rolled plates and hot-rolled flat-coiled plates. is the output of billets on day t. Billets are used to produce wire and bars;

[0034] The material balance of hot-rolled and flat-rolled coils meets the following constraints: (1) The daily hot-rolled plate output and slab inventory are equal to the inventory of the previous day plus the slab output of the day, that is, in is the inventory of the slab warehouse on day t; (2) the daily hot-rolled plate output is less than or equal to its maximum output, that is, in is the maximum daily output of the hot rolling mill; (3) the daily hot-rolled flat-rolled plate output is less than or equal to its maximum output, that is, in is the maximum daily output of hot-rolled flat-rolled plates; (4) The demand for hot-rolled plates is the sum of the daily hot-rolled output and the daily hot-rolled flat-rolled plates, that is, (5) The demand for hot-rolled flat-rolled plates is the sum of the daily production of hot-rolled flat-rolled plates, that is,

[0035] The material balance of wire and bar rolling should meet the following constraints: (1) On the tth day, the jth bar production line only produces bars of one specification, that is, in represents whether the j-th bar production line produces the n-th rolling specification bar on the t-th day; (2) the j-th wire production line only produces one specification wire on the t-th day, that is, in represents whether the j-th wire rod production line produces the m-th rolling specification wire rod on the t-th day; (3) the bar production output of the j-th bar production line on the t-th day is less than or equal to its maximum daily output, in is the maximum daily output of the nth specification bar produced by the jth production line; (4) on the tth day, the output of wire rod produced by the jth wire rod production line is less than or equal to its maximum daily output, in is the maximum daily output of the mth specification wire rod produced by the jth production line; (5) The total output of n specifications of bars produced by the jth production line on the tth day plus the inventory on that day is equal to the inventory of bars on the previous day plus the supply of bars from billets, that is, in is the quantity of billets supplied to the j-th bar production line on day t, is the inventory of the jth billet-bar warehouse on the tth day; (6) The total output of m types of wire rods produced by the jth production line on the tth day plus the inventory on that day is equal to the inventory of wire rods on the previous day plus the supply of wire rods from billets, that is, in is the quantity of billets supplied to the j-th wire production line on day t, is the inventory of the jth billet-wire warehouse on the tth day; (7) The slab output on the tth day is the sum of the slabs supplied to the bar production line and the wire production line, that is, in is the quantity of billets supplied to the j-th bar production line on day t, is the quantity of billets supplied to the j-th wire production line on day t, is the production of billets on the tth day (8) The demand for bars of the nth specification is the sum of the daily production of the two production lines for bars of this specification, that is, (9) The demand for wire of the mth specification is the sum of the daily output of the two production lines of wire of this specification, that is,

[0036] For the industry chain-level intelligent entities, the manufacturing resource allocation planning agent completes order information exchange and bidding with other companies through the JADE platform and MySQL database, obtains the product production orders of the company, and puts the orders into the scheduling tasks according to the priority order. The priority calculation formula is as follows: Among them, p represents the task priority, processtime represents the processing time, endtime represents the latest start time, and nowtime represents the current time. The priority is written into the task queue from high to low. When a new task arrives or a task fails and returns, it will be inserted as an emergency order and the priority table needs to be updated. At this time, the emergency order has the highest priority; when the order is cancelled, all tasks of the order are deleted and the priority table is updated. At the same time, the high-priority order is tendered to the production chain-level Agent on the JADE platform; the order information includes the task quantity of each specification product, the order delivery time, and the price of each specification product in the order; the decision of the material procurement planning agent is determined by the enterprise production model, and is obtained through the message transmission system of the JADE platform and the manufacturing resource allocation planning agent. The order information is obtained, and the Gurobi solver is used to solve the enterprise production model under the condition of determining the order information to obtain the molten iron output required to complete the order. The material input and molten iron output historical data of the enterprise in the MySQL database are trained through machine learning to obtain the material-molten iron output neural network model to determine the material input ratio per unit molten iron output, and then the molten iron output required for the order is used to deduce the input amount of each material, generate the enterprise material procurement plan, and bid to upstream enterprises based on the material procurement amount in the material procurement plan, and select lower-priced materials with the goal of maximizing the enterprise's interests; the product marketing plan agent communicates with the production chain-level agent through the JADE platform to obtain the inventory information of each process product, bid to downstream enterprises with the goal of maximizing the enterprise's interests, and select the downstream enterprise with the highest bid to complete the contract;

[0037] For the production chain-level intelligent agent, the JADE platform and MySQL database are used to exchange information with the production equipment agent and the industrial chain-level agent and complete the bidding work. The workshop production planning agent uploads the production line information, daily maximum production capacity information, production cost information, and product inventory information of each production chain-level agent to the MySQL database and bids to the manufacturing resource allocation planning agent through the JADE platform, and uses the enterprise production model to use the enterprise profit value as the evaluation standard of the bid document to use the gurobi solver to determine the task volume completed by each process; the workshop operation planning agent bids to the production equipment agent through the JADE platform, and also uses the enterprise production model to use the enterprise profit value as the evaluation standard of the bid document to use the gurobi solver to determine the production output of different specifications of products produced by each production equipment every day. At the same time, the production data information and production conditions of the production equipment-level agent of each process are obtained through the MySQL database. If a production equipment has abnormal data, the workshop operation planning agent reports it to the manufacturing resource allocation planning agent and puts the task responsible for the equipment back into the task pool as a failed order;

[0038] For the production equipment agent, the JADE platform and MySQL database are used to bid to the production chain-level Agent. Each production equipment Agent uploads the production data information and production condition information to the MySQL database through sensor data. The JADE platform is used to bid with the production data information and production condition information as the bidding content. The workshop operation planning Agent uses the Gurobi solver to determine the output of different specifications of products produced by each production equipment every day with the enterprise profit value as the evaluation standard of the bid. During the production process, the equipment agent uses the historical material input information, carbon emission information, and product output information of each production equipment in the MySQL database to use deep learning training. The neural network model with the material input of each production equipment as input and carbon emissions and product output as output determines the relationship between the material input ratio of each production equipment and the carbon emission factor, and the unit product output. Each production equipment adaptively optimizes the material input of each process and optimizes the allocation of resources.

[0039] Beneficial technical effects of the present invention:

[0040] In view of the need for the steel industry production chain and industrial chain to meet the lean and green requirements of the steel industry's transformation and upgrading and the emergencies that may arise at any time, the intelligentization of steel enterprises requires changing the existing production chain and industrial chain coordination mechanism, and using the contract network protocol to build an industrial Internet production chain and industrial chain coordination method based on multi-agents composed of three-level intelligent entities.

[0041] The industrial Internet of Things production chain and industrial chain collaboration method based on multi-agents described in the present invention is a method suitable for the collaborative optimization of the steel industry industrial chain and production chain. Through the collaboration between agents at the same level and the collaboration between agents at multiple levels, a systematic multi-agent collaboration method is formed around production and operation, covering three levels of intelligent systems: production equipment agents, production chain-level agents, and industrial chain-level agents. The big data technology based on the industrial Internet and the artificial intelligence technology oriented to deep perception are used to achieve the goals of stable product quality, timely delivery of orders, and maximization of corporate benefits for steel enterprises. Not only can real-time production planning decisions be made in production, but also emergencies such as equipment damage and emergency orders can be effectively dealt with. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic diagram of a method for coordinating production and industrial chains based on a multi-agent industrial Internet of Things according to an embodiment of the present invention;

[0043] Figure 2 Flowchart of the contract network protocol mechanism according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The present invention will be further described below in conjunction with the accompanying drawings and embodiments;

[0045] Under the constraints of flexible supply and dynamic demand in the external market, dynamic production conditions within the enterprise (equipment capacity, process parameters), external resource consumption, carbon emissions, etc., the present invention aims to improve the comprehensive evaluation indicators of the entire production process including product output, quality, energy consumption, emissions, costs, etc., using a virtual manufacturing process to achieve feedforward decision correction based on virtual simulation, and to identify and feedback working conditions through large-scale sensory data in industrial production, so as to achieve self-optimization decision-making and dynamic optimization decision-making through human-computer interaction, thereby establishing a collaborative method for the industrial chain and production chain based on multi-agents;

[0046] A collaborative method for production and industrial chains based on multi-agent industrial Internet of Things, as shown in the attached Figure 1 As shown, it consists of three parts: production equipment agent, production chain-level collaborative agent, and industrial chain-level resource allocation agent;

[0047] The production equipment agent corresponds to the production equipment, including blast furnace agent, converter agent, and rolling mill agent, and completes their respective production goals; the production chain-level agent includes the workshop production plan agent and the workshop operation plan agent, which is responsible for completing the production goals of each process in the production chain. On the one hand, it generates the operation plan of the production equipment of each process, controls the production equipment agent to realize the production of the underlying production equipment, and on the other hand, obtains the production plan of the enterprise from the industrial chain-level agent and is controlled by the industrial chain-level agent; the industrial chain-level agent includes the material procurement plan agent, the product marketing plan agent, and the manufacturing resource allocation plan agent, which optimizes the overall production process of the enterprise, determines the production plan of each process, and guides each process to complete the production operation;

[0048] The production equipment agent, production chain-level collaborative agent, and industrial chain-level resource allocation agent use the contract net protocol and CNP mechanism for communication and decision-making. The process of the contract net protocol is as follows: Figure 2 As shown in the figure, the contract network consists of several nodes, which are divided into two types: managers and agents. In the contract network, managers and agents exchange messages dynamically through negotiation, rather than in advance. The decision-making of the contract network includes the following four basic steps:

[0049] Step 1: The manager releases the task for bidding;

[0050] Step 2: The agent submits its bidding information to the manager, including capacity and production cost;

[0051] Step 3: The manager selects an agent to perform the task based on the principle of maximizing the company's profits;

[0052] Step 4: Finally, the manager confirms the contract and finally completes the bidding-competitive decision-making process;

[0053] The functions of the production equipment intelligent agent are: to obtain production data information of each equipment through sensors, including: production efficiency, production benefits, product quality, operation status of production equipment, environmental changes, material types and quality, energy resources, human resources, and equipment resources; to intelligently perceive production working condition information, including: material input, energy consumption, and carbon emissions; and to use neural network models to adaptively make optimization decisions, optimize resource allocation, and reasonably allocate and recycle energy; production equipment intelligent agents adopt communication and information sharing strategies, and effective communication and information sharing are required between intelligent agents in order to understand each other's status, intentions, and action plans. Communication is achieved by directly transmitting messages, sharing knowledge bases, or sensor data.

[0054] The blast furnace agent, converter agent, and rolling mill agent have the following functions:

[0055] (1) Blast furnace agent: Each blast furnace agent bids for its own production task from the workshop operation planning agent, while sensing production data and production conditions;

[0056] (2) Converter Agent: Each converter agent bids for its own production task from the workshop operation planning agent and senses its own production data and production conditions.

[0057] (3) Rolling mill agent: Each rolling mill agent bids for its own production tasks from the workshop operation planning agent, and senses its own production data and production conditions;

[0058] The production chain-level intelligent agent receives the order production task and uses the contract network to make independent decisions and formulate cooperation strategies. The decision-making and cooperation strategy involve the division of tasks, which requires comprehensive consideration of the overall goal and individual interests; the workshop production planning agent and the workshop operation planning agent of the ironmaking, steelmaking and rolling processes of each process collaboratively control the production equipment-level intelligent agent to achieve worry-free production in the workshop; the workshop production planning agent formulates the workshop production plan according to the production capacity of each process transmitted by the production equipment through the JADE platform and the enterprise production model; the workshop operation planning agent gives full play to the production capacity of each process, organizes production reasonably, ensures the schedulability of the resources of each process, reasonably arranges the operations of each process, realizes the coordinated matching of the operation of each process, and ensures the smooth operation of the whole workshop production;

[0059] The workshop production planning agent and the workshop operation planning agent achieve the global goal through mutual cooperation, and their functions are as follows:

[0060] (1) Workshop production planning agent: Use the contract network protocol to bid to the manufacturing resource allocation planning agent to obtain the workshop production plan;

[0061] (2) Workshop operation plan agent: bid for each production equipment agent and select a production equipment agent to complete the workshop operation plan;

[0062] The industry chain-level intelligent agent achieves the overall goal by coordinating the production of each process. It coordinates with upstream and downstream enterprises to complete material procurement and product marketing plans through the material procurement planning agent, product marketing planning agent, and manufacturing resource allocation planning agent. At the same time, it prioritizes the orders accepted by the enterprise to improve the efficiency of the entire enterprise.

[0063] The functions of the material procurement planning agent, product marketing planning agent, and manufacturing resource allocation planning agent are as follows:

[0064] (1) Material Procurement Planning Agent: The production of an enterprise depends on the materials of upstream enterprises. Therefore, the enterprise bids for upstream enterprises according to the Material Procurement Planning Agent to complete the procurement of required materials.

[0065] (2) Product Marketing Plan Agent: The production of downstream enterprises depends on the enterprise's products. Therefore, the enterprise bids for downstream enterprises according to the product marketing plan agent to complete the enterprise's product marketing plan;

[0066] (3) Manufacturing resource allocation planning agent: responsible for generating production plans based on the production orders received by the enterprise, and sending the orders generated by the production plan to the production chain-level workshop production planning agent, and completing the bidding-bidding with the production chain-level intelligent agent;

[0067] The production chain and industrial chain collaboration method is implemented using the JADE framework of Java. The basic functions of the JADE framework include the agent AMS responsible for supervising and managing the access and use process of the platform, and the directory service agent DF and the message transmission system that provide the default yellow pages. Among them, AMS is used to supervise and control the Agent platform, and each agent must register with AMS to obtain a valid AID; DF provides the default yellow pages service for the Agent; the message transmission system is a software component in the platform that controls all message exchanges; Agent interaction and communication are carried out through the above functions, and the MySQL database is used to realize information access; the production equipment intelligent agent, the production chain-level intelligent agent, and the industrial chain-level intelligent agent are created through the basic functions of the JADE platform, and the information interaction between different Agents is completed through the MySQL database and the message transmission system in the JADE framework, and the identity identification between Agents is realized through the AID in the JADE framework;

[0068] A method for industrial Internet of Things production chain and industrial chain collaboration based on multi-agent, specifically including the following contents:

[0069] Establish an enterprise production model to provide a basis for the decision-making behavior of each agent. Set the enterprise's planned production time to T, the product types include hot-rolled plates, hot-rolled flat coils, M types of wire rods, N types of bars, and two ironmaking equipments, a large blast furnace and a small blast furnace. The molten iron output of the large blast furnace on the tth day is F. Iron_b (t), the unit production cost of a large blast furnace is UC Iron_b , the molten iron output of the small blast furnace on day t is F Iron_s (t), the unit production cost of a small blast furnace is UC Iron_s There are two steelmaking equipments, converter No. 1 and converter No. 2. The steel output of converter No. 1 on day t is F Steel_o (t), the unit production cost of converter No. 1 is UC Steel_o , the steel output of converter No. 2 on day t is F Steel_t (t), the unit production cost of converter No. 2 is UC Steel_t ; There is one hot rolling mill production line, and the output of the hot rolling mill on day t is The unit production cost of hot rolled plate is UC HR There is one flattening and coiling machine production line, and the output of the flattening and coiling machine on day t is The unit production cost of hot-rolled flat-rolled plate is UC Smooth ; There are three bar production lines. The output of the nth specification bar of the jth bar production line on the tth day is The unit production cost of the nth specification bar is There are two wire production lines. The output of the wire of specification m on day t of the j-th wire production line is The unit production cost of the mth specification bar is The demand for hot rolled plates is D HR , the price of hot rolled plate is Up HR , the demand for hot-rolled flat coils is D Smooth , hot rolled flat coil price is Up Smooth , the demand for bars of the nth specification is The price of the nth specification wire is The demand for wire of the mth specification is The price of the mth specification wire is The objective function of the enterprise production model is to maximize production profit, that is:

[0070]

[0071] In the production process, the material balance of ironmaking and steelmaking, the material balance of hot rolling and flat-rolling, and the material balance of wire and bar rolling are used as the limiting conditions of the enterprise production model;

[0072] The material balance for ironmaking and steelmaking should meet the following constraints: (1) The total daily molten iron output of the large and small blast furnaces is equal to the steel output of the No. 1 converter and the No. 2 converter, that is, F Iron_b (t)-F Steel_o (t) = F Steel_t (t)-F Iron_s (t)t∈T; (2) The daily output of billets and slabs is equal to the sum of the steel output of converter No. 1 and converter No. 2, that is, in is the output of slab on day t, which is used to produce hot-rolled plates and hot-rolled flat-coiled plates. is the output of billets on day t. Billets are used to produce wire and bars;

[0073] The material balance of hot-rolled and flat-rolled coils shall meet the following constraints: (1) The daily hot-rolled plate output and slab inventory shall be equal to the inventory of the previous day plus the slab output of the day, that is, in is the inventory of the slab warehouse on day t; (2) the daily hot-rolled plate output is less than or equal to its maximum output, that is, in is the maximum daily output of the hot rolling mill; (3) the daily hot-rolled flat-rolled plate output is less than or equal to its maximum output, that is, in is the maximum daily output of hot-rolled flat-rolled plates; (4) The demand for hot-rolled plates is the sum of the daily hot-rolled output and the daily hot-rolled flat-rolled plates, that is, (5) The demand for hot-rolled flat-rolled plates is the sum of the daily production of hot-rolled flat-rolled plates, that is,

[0074] The material balance of wire and bar rolling should meet the following constraints: (1) On the tth day, the jth bar production line only produces bars of one specification, that is, in represents whether the j-th bar production line produces the n-th rolling specification bar on the t-th day; (2) the j-th wire production line only produces one specification wire on the t-th day, that is, in represents whether the j-th wire rod production line produces the m-th rolling specification wire rod on the t-th day; (3) the bar production output of the j-th bar production line on the t-th day is less than or equal to its maximum daily output, in is the maximum daily output of the nth specification bar produced by the jth production line; (4) on the tth day, the output of wire rod produced by the jth wire rod production line is less than or equal to its maximum daily output, in is the maximum daily output of the mth specification wire rod produced by the jth production line; (5) The total output of n specifications of bars produced by the jth production line on the tth day plus the inventory on that day is equal to the inventory of bars on the previous day plus the supply of bars from billets, that is, in is the quantity of billets supplied to the j-th bar production line on day t, is the inventory of the jth billet-bar warehouse on the tth day; (6) The total output of m types of wire rods produced by the jth production line on the tth day plus the inventory on that day is equal to the inventory of wire rods on the previous day plus the supply of wire rods from billets, that is, in is the quantity of billets supplied to the j-th wire production line on day t, is the inventory of the jth billet-wire warehouse on the tth day; (7) The slab output on the tth day is the sum of the slabs supplied to the bar production line and the wire production line, that is, in is the quantity of billets supplied to the j-th bar production line on day t, is the quantity of billets supplied to the j-th wire production line on day t, is the production of billets on the tth day (8) The demand for bars of the nth specification is the sum of the daily production of the two production lines for bars of this specification, that is, (9) The demand for wire of the mth specification is the sum of the daily output of the two production lines of wire of this specification, that is,

[0075] For the industry chain-level intelligent entities, the manufacturing resource allocation planning agent completes order information exchange and bidding with other companies through the JADE platform and MySQL database, obtains the product production orders of the company, and puts the orders into the scheduling tasks according to the priority order. The priority calculation formula is as follows: Among them, p represents the task priority, processtime represents the processing time, endtime represents the latest start time, and nowtime represents the current time. The priority is written into the task queue from high to low. When a new task arrives or a task fails and returns, it will be inserted as an emergency order and the priority table needs to be updated. At this time, the emergency order has the highest priority; when the order is cancelled, all tasks of the order are deleted and the priority table is updated. At the same time, the high-priority order is tendered to the production chain-level Agent on the JADE platform; the order information includes the task quantity of each specification product, the order delivery time, and the price of each specification product in the order; the decision of the material procurement planning agent is determined by the enterprise production model, and is obtained through the message transmission system of the JADE platform and the manufacturing resource allocation planning agent. The order information is obtained, and the Gurobi solver is used to solve the enterprise production model under the condition of determining the order information to obtain the molten iron output required to complete the order. The material input and molten iron output historical data of the enterprise in the MySQL database are trained through machine learning to obtain the material-molten iron output neural network model to determine the material input ratio per unit molten iron output, and then the molten iron output required for the order is used to deduce the input amount of each material, generate the enterprise material procurement plan, and bid to upstream enterprises based on the material procurement amount in the material procurement plan, and select lower-priced materials with the goal of maximizing the enterprise's interests; the product marketing plan agent communicates with the production chain-level agent through the JADE platform to obtain the inventory information of each process product, bid to downstream enterprises with the goal of maximizing the enterprise's interests, and select the downstream enterprise with the highest bid to complete the contract;

[0076] For the production chain-level intelligent agent, the JADE platform and MySQL database are used to exchange information with the production equipment agent and the industrial chain-level agent and complete the bidding work. The workshop production planning agent uploads the production line information, daily maximum production capacity information, production cost information, and product inventory information of each production chain-level agent to the MySQL database and bids to the manufacturing resource allocation planning agent through the JADE platform, and uses the enterprise production model to use the enterprise profit value as the evaluation standard of the bid document to use the gurobi solver to determine the task volume completed by each process; the workshop operation planning agent bids to the production equipment agent through the JADE platform, and also uses the enterprise production model to use the enterprise profit value as the evaluation standard of the bid document to use the gurobi solver to determine the production output of different specifications of products produced by each production equipment every day. At the same time, the production data information and production conditions of the production equipment-level agent of each process are obtained through the MySQL database. If a production equipment has abnormal data, the workshop operation planning agent reports it to the manufacturing resource allocation planning agent and puts the task responsible for the equipment back into the task pool as a failed order;

[0077] For the production equipment agent, the JADE platform and MySQL database are used to bid to the production chain-level Agent. Each production equipment Agent uploads the production data information and production condition information to the MySQL database through sensor data. The JADE platform is used to bid with the production data information and production condition information as the bidding content. The workshop operation planning Agent uses the Gurobi solver to determine the output of different specifications of products produced by each production equipment every day with the enterprise profit value as the evaluation standard of the bid. During the production process, the equipment agent uses the historical material input information, carbon emission information, and product output information of each production equipment in the MySQL database to use deep learning training. The neural network model with the material input of each production equipment as input and carbon emissions and product output as output determines the relationship between the material input ratio of each production equipment and the carbon emission factor, and the unit product output. Each production equipment adaptively optimizes the material input of each process and optimizes the allocation of resources.

Claims

1. A multi-agent based industrial Internet of Things production chain and industrial chain collaboration method, characterized in that: It consists of three parts: production equipment agent, production chain-level collaborative agent, and industrial chain-level resource allocation agent; The production equipment agent corresponds to the production equipment, including blast furnace agent, converter agent, and rolling mill agent, to complete their respective production goals; the production chain-level agent includes the workshop production plan agent and the workshop operation plan agent, which, on the one hand, generates the operation plan for the production equipment of each process, controls the production equipment agent to realize the production of the underlying production equipment, and on the other hand, obtains the production plan of the enterprise from the industrial chain-level agent and is controlled by the industrial chain-level agent; the industrial chain-level agent includes the material procurement plan agent, the product marketing plan agent, and the manufacturing resource allocation plan agent, which optimizes the overall production process of the enterprise, determines the production plan of each process, and guides each process to complete the production operation.

2. According to the multi-agent-based industrial Internet of Things production chain and industrial chain collaboration method according to claim 1, it is characterized in that: The production equipment agent, production chain-level collaborative agent, and industrial chain-level resource allocation agent adopt the contract net protocol and CNP mechanism for communication and decision-making; the contract net consists of n nodes, which are divided into two types: managers and agents. In the contract net, managers and agents exchange dynamic messages through negotiation. The decision-making of the contract net includes the following four basic steps: Step 1: The manager releases the task for bidding; Step 2: The agent submits its bidding information to the manager, including capacity and production cost; Step 3: The manager selects an agent to perform the task based on the principle of maximizing the company's profits; Step 4: Finally, the manager confirms the contract and finally completes the bidding-competitive decision-making process.

3. According to the multi-agent-based industrial Internet of Things production chain and industrial chain collaboration method of claim 1, it is characterized in that: The functions of the production equipment intelligent agent are: to obtain production data information of each equipment through sensors, including: production efficiency, production benefits, product quality, operation status of production equipment, environmental changes, material types and quality, energy resources, human resources, and equipment resources; to intelligently perceive production working condition information, including: material input, energy consumption, and carbon emissions; and to use neural network models to adaptively make optimization decisions, optimize resource allocation, and reasonably allocate and recycle energy; production equipment intelligent agents adopt communication and information sharing strategies, and effective communication and information sharing are required between intelligent agents in order to understand each other's status, intentions, and action plans. Communication is achieved by directly transmitting messages, sharing knowledge bases, or sensor data. The blast furnace agent, converter agent, and rolling mill agent have the following functions: (1) Blast furnace agent: Each blast furnace agent bids for its own production task from the workshop operation planning agent, while sensing production data and production conditions; (2) Converter Agent: Each converter agent bids for its own production task from the workshop operation planning agent and senses its own production data and production conditions. (3) Rolling mill agent: Each rolling mill agent bids for its own production tasks from the workshop operation planning agent, and senses its own production data and production conditions; The production chain-level intelligent agent receives the order production task and uses the contract network to make independent decisions and formulate cooperation strategies. The decision-making and cooperation strategy involve the division of tasks, which requires comprehensive consideration of the overall goal and individual interests; the workshop production planning agent and the workshop operation planning agent of the ironmaking, steelmaking and rolling processes of each process collaboratively control the production equipment intelligent agent to achieve worry-free production in the workshop; the workshop production planning agent formulates the workshop production plan according to the production capacity of each process transmitted by the production equipment through the JADE platform and the enterprise production model; the workshop operation planning agent gives full play to the production capacity of each process, organizes production reasonably, ensures the schedulability of the resources of each process, reasonably arranges the operations of each process, realizes the coordinated matching of the operation of each process, and ensures the smooth operation of the whole workshop production; The workshop production planning agent and the workshop operation planning agent achieve the global goal through mutual cooperation, and their functions are as follows: (1) Workshop production planning agent: Use the contract network protocol to bid to the manufacturing resource allocation planning agent to obtain the workshop production plan; (2) Workshop operation plan agent: bid for each production equipment agent and select a production equipment agent to complete the workshop operation plan; The industry chain-level agent achieves the overall goal by coordinating the production of each process. It coordinates with upstream and downstream enterprises to complete material procurement and product marketing plans through the material procurement plan agent, product marketing plan agent, and manufacturing resource allocation plan agent. At the same time, it prioritizes the orders accepted by the enterprise to improve the efficiency of the entire enterprise. The functions of the material procurement planning agent, product marketing planning agent, and manufacturing resource allocation planning agent are as follows: (1) Material Procurement Planning Agent: The production of an enterprise depends on the materials of upstream enterprises. Therefore, the enterprise bids for upstream enterprises according to the Material Procurement Planning Agent to complete the procurement of required materials. (2) Product Marketing Plan Agent: The production of downstream enterprises depends on the enterprise's products. Therefore, the enterprise bids for downstream enterprises according to the product marketing plan agent to complete the enterprise's product marketing plan; (3) Manufacturing resource allocation planning agent: responsible for generating production plans based on the production orders received by the enterprise, and issuing the orders generated by the production plan to the production chain-level workshop production planning agent, and completing the bidding-bidding with the production chain-level intelligent agent.

4. According to the multi-agent-based industrial Internet of Things production chain and industrial chain collaboration method of claim 1, it is characterized in that: The production chain and industrial chain collaboration method is implemented using the JADE framework of Java. The basic functions of the JADE framework include the agent AMS responsible for supervising and managing the access and use process of the platform, and the directory service agent DF and the message transmission system that provide the default yellow pages. Among them, AMS is used to supervise and control the Agent platform, and each agent must register with AMS to obtain a valid AID; DF provides the default yellow pages service for the Agent; the message transmission system is a software component in the platform that controls all message exchanges; Agent interaction and communication are carried out through the above functions, and the MySQL database is used to realize information access; the production equipment intelligent agent, the production chain-level intelligent agent, and the industrial chain-level intelligent agent are created through the basic functions of the JADE platform, and the information interaction between different Agents is completed through the MySQL database and the message transmission system in the JADE framework, and the identity identification between Agents is realized through the AID in the JADE framework.

5. According to the multi-agent-based industrial Internet of Things production chain and industrial chain collaboration method of claim 1, it is characterized in that: Also included are the following: Establish an enterprise production model to provide a basis for the decision-making behavior of each agent. Set the enterprise's planned production time to T, the product types include hot-rolled plates, hot-rolled flat coils, M types of wire rods, N types of bars, and two ironmaking equipments, a large blast furnace and a small blast furnace. The molten iron output of the large blast furnace on the tth day is F. Iron_b (t), the unit production cost of a large blast furnace is UC Iron_b , the molten iron output of the small blast furnace on day t is F Iron_s (t), the unit production cost of a small blast furnace is UC Iron_s There are two steelmaking equipments, converter No. 1 and converter No.

2. The steel output of converter No. 1 on day t is F Steel_o (t), the unit production cost of converter No. 1 is UC Steel_o , the steel output of converter No. 2 on day t is F Steel_t (t), the unit production cost of converter No. 2 is UC Steel_t ; There is one hot rolling mill production line, and the output of the hot rolling mill on day t is The unit production cost of hot rolled plate is UC HR There is one flattening and coiling machine production line, and the output of the flattening and coiling machine on day t is The unit production cost of hot-rolled flat-rolled plate is UC Smooth ; There are three bar production lines. The output of the nth specification bar of the jth bar production line on the tth day is The unit production cost of the nth specification bar is There are two wire production lines. The output of the wire of specification m on day t of the j-th wire production line is The unit production cost of the mth specification bar is The demand for hot rolled plates is D HR , the price of hot rolled plate is Up HR , the demand for hot-rolled flat coils is D Smooth , hot rolled flat coil price is Up Smooth , the demand for bars of the nth specification is The price of the nth specification wire is The demand for wire of the mth specification is The price of the mth specification wire is The objective function of the enterprise production model is to maximize production profit, that is: During the production process, the material balance of ironmaking and steelmaking, the material balance of hot rolling and flat coiling, and the material balance of wire and bar rolling are used as constraints on the enterprise production model.

6. The method for industrial Internet of Things production chain and industrial chain collaboration based on multi-agent according to claim 5 is characterized in that: The material balance of ironmaking and steelmaking meets the following constraints: (1) The total daily molten iron output of the large and small blast furnaces is equal to the steel output of the No. 1 converter and the No. 2 converter, that is, F Iron_b (t)-F Steel_o (t) = F Steel_t (t)-F Iron_s (t)t∈T; (2) The daily output of billets and slabs is equal to the sum of the steel output of converter No. 1 and converter No. 2, that is, in is the output of slab on day t, which is used to produce hot-rolled plates and hot-rolled flat-coiled plates. is the output of billets on day t. Billets are used to produce wire and bars; The material balance of hot-rolled and flat-rolled coils meets the following constraints: (1) The daily hot-rolled plate output and slab inventory are equal to the inventory of the previous day plus the slab output of the day, that is, in is the inventory of the slab warehouse on day t; (2) the daily hot-rolled plate output is less than or equal to its maximum output, that is, in is the maximum daily output of the hot rolling mill; (3) the daily hot-rolled flat-rolled plate output is less than or equal to its maximum output, that is, in is the maximum daily output of hot-rolled flat-rolled plates; (4) The demand for hot-rolled plates is the sum of the daily hot-rolled output and the daily hot-rolled flat-rolled plates, that is, (5) The demand for hot-rolled flat-rolled plates is the sum of the daily production of hot-rolled flat-rolled plates, that is, The material balance of wire and bar rolling should meet the following constraints: (1) On the tth day, the jth bar production line only produces bars of one specification, that is, in represents whether the j-th bar production line produces the n-th rolling specification bar on the t-th day; (2) the j-th wire production line only produces one specification wire on the t-th day, that is, in represents whether the j-th wire rod production line produces the m-th rolling specification wire rod on the t-th day; (3) the bar production output of the j-th bar production line on the t-th day is less than or equal to its maximum daily output, in is the maximum daily output of the nth specification bar produced by the jth production line; (4) on the tth day, the output of wire rod produced by the jth wire rod production line is less than or equal to its maximum daily output, in is the maximum daily output of the mth specification wire rod produced by the jth production line; (5) The total output of n specifications of bars produced by the jth production line on the tth day plus the inventory on that day is equal to the inventory of bars on the previous day plus the supply of bars from billets, that is, in is the quantity of billets supplied to the j-th bar production line on day t, is the inventory of the jth billet-bar warehouse on the tth day; (6) The total output of m types of wire rods produced by the jth production line on the tth day plus the inventory on that day is equal to the inventory of wire rods on the previous day plus the supply of wire rods from billets, that is, in is the quantity of billets supplied to the j-th wire production line on day t, is the inventory of the jth billet-wire warehouse on the tth day; (7) The slab output on the tth day is the sum of the slabs supplied to the bar production line and the wire production line, that is, in is the quantity of billets supplied to the j-th bar production line on day t, is the quantity of billets supplied to the j-th wire production line on day t, is the production of billets on the tth day (8) The demand for bars of the nth specification is the sum of the daily production of the two production lines for bars of this specification, that is, (9) The demand for wire of the mth specification is the sum of the daily output of the two production lines of wire of this specification, that is, 7. The method for industrial Internet of Things production chain and industrial chain collaboration based on multi-agent according to claim 1 is characterized in that: For the industry chain-level intelligent entities, the manufacturing resource allocation planning agent completes order information exchange and bidding with other companies through the JADE platform and MySQL database, obtains the product production orders of the company, and puts the orders into the scheduling tasks according to the priority order. The priority calculation formula is as follows: Among them, p represents the task priority, processtime represents the processing time, endtime represents the latest start time, and nowtime represents the current time. The priority is written into the task queue from high to low. When a new task arrives or a task fails and returns, it will be inserted as an urgent order and the priority table needs to be updated. At this time, the urgent order has the highest priority; when the order is cancelled, all tasks of the order are deleted and the priority table is updated. At the same time, the high-priority orders are tendered to the JADE platform and the production chain-level Agent; the order information includes the task quantity of each specification product, the order delivery time, and the price of each specification product in the order; The decision of the material procurement plan agent is determined by the enterprise production model. It communicates with the manufacturing resource allocation plan agent through the message transmission system of the JADE platform to obtain order information. Under the condition of determining the order information, the Gurobi solver is used to solve the enterprise production model to obtain the molten iron output required to complete the order. The material input ratio per unit molten iron output is determined by training the enterprise material input and molten iron output historical data in the MySQL database through machine learning to obtain a material-molten iron output neural network model. Then, the input amount of each material is deduced from the molten iron output required for the order, and the enterprise material procurement plan is generated. The material procurement amount in the material procurement plan is used to bid to upstream enterprises, and the lower-priced materials are selected with the maximum benefit of the enterprise as the goal. The product marketing planning agent communicates with the production chain-level agent through the JADE platform to obtain the inventory information of each process product, bids to downstream enterprises with the goal of maximizing the interests of the enterprise, and selects the downstream enterprise with the highest bid to complete the contract; For the production chain-level intelligent agent, the JADE platform and MySQL database are used to exchange information with the production equipment agent and the industrial chain-level agent and complete the bidding work. The workshop production planning agent uploads the production line information, daily maximum production capacity information, production cost information, and product inventory information of each production chain-level agent to the MySQL database and bids to the manufacturing resource allocation planning agent through the JADE platform, and uses the enterprise production model to use the enterprise profit value as the evaluation standard of the bid document to use the gurobi solver to determine the task volume completed by each process; the workshop operation planning agent bids to the production equipment agent through the JADE platform, and also uses the enterprise production model to use the enterprise profit value as the evaluation standard of the bid document to use the gurobi solver to determine the production output of different specifications of products produced by each production equipment every day. At the same time, the production data information and production conditions of the production equipment-level agent of each process are obtained through the MySQL database. If a production equipment has abnormal data, the workshop operation planning agent reports it to the manufacturing resource allocation planning agent and puts the task responsible for the equipment back into the task pool as a failed order; For the production equipment agent, the JADE platform and MySQL database are used to bid to the production chain-level Agent. Each production equipment Agent uploads the production data information and production condition information to the MySQL database through sensor data. The JADE platform is used to bid with the production data information and production condition information as the bidding content. The workshop operation planning Agent uses the Gurobi solver to determine the output of different specifications of products produced by each production equipment every day with the enterprise profit value as the evaluation standard of the bid. During the production process, the equipment agent uses the historical material input information, carbon emission information, and product output information of each production equipment in the MySQL database to use deep learning training. The neural network model with the material input of each production equipment as input and carbon emissions and product output as output determines the relationship between the material input ratio of each production equipment and the carbon emission factor, and the unit product output. Each production equipment adaptively optimizes the material input of each process and optimizes the allocation of resources.

Images