Agent-oriented method for integrated modeling of chemical process systems

The modeling methods commonly used in chemical process systems are structured analysis methods and object-oriented methods. The object-oriented method overcomes the shortcomings of the structured method to decompose the system according to its function. However, it has certain disadvantages in describing objects that have initiative activities such as fault diagnosis, simulation optimization, and production scheduling in the chemical production process.

1.2 Running the Task Model Process There is a class of active entity systems in the running system, such as real-time simulation, online optimization, scheduling, fault diagnosis, and security analysis. This article abstracts these operational tasks into agents. In order to realize integration of multiple operational tasks, the established multi-agent system includes: Agent structure, message transmission and exchange platform between agents, communication language and protocol of agent, and data format of communication language.

Agent structure As shown, the internal knowledge base including the internal knowledge base body consists of data or rules, including the description of the agent itself, the description of the agent to the environment, and the description of the agent to other agents. The function of the actuator is the management function module.

Each module is executed in parallel after being started by the execution machine and coordinated through the blackboard. The communication module is responsible for receiving external information and sending messages.

(Public Object Request Broker Architecture) As the "soft bus" for message transmission and exchange among agents, its role is to transfer requests across network address spaces, and to connect a group of agents with different functions so that the agents have Plug and play features. The communication language and protocol of the agent are (knowledge inquiry and operation language), the data format is adopted (product data expression and exchange standard), and the protocols in the process industry are AP221, AP227 and AP231. These protocols can be used as the content of the KQML communication language. Data Format 16.2 Integration of Chemical Process Systems Process models and operational task models have been established for process operation systems in terms of agent analysis and modeling methods. The integration of the process operation system can be achieved through 3 aspects: 1 information integration; 2 task integration; 3 people's participation.

2.1 Information integration Information integration refers to the integration of process object data, models, and related rule knowledge, which can be known through a neutral process model. The process model knowledge base consists of three parts: neutral data files, neutral model interface files, and process knowledge rules.

Neutral data files are for some of their own process engineering software such as Aspenplus, Pro-, etc. These softwares have specific input and output files, the data formats are not the same, but for the same process model parameters such as material properties parameters, equipment The design parameters and the operating parameters of the process are the same. These parameters can be standardized by the STEP protocol. The standardized process data is stored in neutral data files. These neutral data files can generate input/output files for different applications according to certain application protocols, so that the process model parameters can be efficiently stored in different applications. Take and exchange.

In addition to the parameters of process models that can be shared, process models can also be shared in different applications. The process model interface is defined and described in a neutral format, such as the CAPE-OPEN method to analyze the flow of messages between different agents. At the same time, an evaluation indicator reflecting the status of the entire process operation system is formulated. According to different process operating system states, through the message flow chart to find out which agents should be negotiated to determine the contents of the negotiations and the corresponding decisions.

The rule knowledge constructed according to the above method is stored in the internal knowledge base of the agent, and these rule knowledge can be written in advance, or a function module of the agent can be used to experience the change of the external environment and solve the problem in actual operation. Learn to produce new rules.

The participation of 23 participants means that the decisions to run the agents are not closed, and manual interventions and modifications can be made. Humans can be regarded as a special type of intelligent operation tasks. The integration of intelligent agents and humans becomes the process of designing and coordinating the ergonomics of a task agent and a special agent through an interactive interface. The key to the supporter's participation is also to construct the internal knowledge base of the agent so that the agent can interact with people according to the corresponding rule knowledge.

3 Framework structure of process operation system integration On the basis of the above work, the author proposes a framework for the integration of process operation systems in which the agent and process model knowledge bases are connected via (3) RBA with factory devices and DCS systems, as shown. Each agent connects through the network, and exchanges messages in the (3) RBA message transmission and exchange platform according to the IKQML communication language and protocol and STEP data format. Commercial software (AspenPlisGAMS, G2, etc. in the figure) can be called as a function module of the agent. Each agent judges the status of the process operation system, negotiates with other agents to avoid conflicts, and integrates the entire process operation.

4 Analysis of the conclusions, the establishment of a corresponding process model and operating task model. When a new task appears in the production process, it is added to the original system through the agent or object model. Agent-oriented system analysis and modeling methods can integrate different tasks of the process system, realize the openness of the system, and provide the theoretical basis and system structure for the integration of the entire process enterprise.

71 According to the agent-oriented modeling method, the process system is divided