Chemical Process Closed-Loop Control Using Simulated State Variables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large-scale chemical plants face challenges in optimizing operation, control, transparency, and maintenance, particularly during commissioning, with existing advanced process control systems not fully addressing the complexity and safety requirements of these plants.
Innovation Solution
A method involving sensory data acquisition, dynamic process simulation, and machine learning to determine manipulated variables for real-time control, using a client-server architecture and fieldbus communication to adjust actuators, with a machine learning module trained on simulated state variables and manipulated variables to enhance control precision and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a machine learning module is integrated into the control system to improve control precision and adaptability, then control performance and transparency are improved, but device complexity and difficulty of detecting and measuring increase
Solution Approach 1:
A simulation model serves as an intermediary between the machine learning module and the actual chemical process. The simulation model receives manipulated variables from the machine learning module and generates simulated process variables, which are then compared with actual measured values. This intermediary layer allows the complex machine learning operations to occur in a virtual environment, reducing the direct complexity burden on the physical control system while maintaining high control precision.
Solution Approach 2:
The system creates a virtual copy of the chemical process through a simulation model that replicates the behavior of the actual plant. This digital twin or copy allows the machine learning module to train and operate without directly interacting with the complex physical system, thereby improving control precision while containing device complexity within the virtual domain. The simulated state variables from this copy are used to determine manipulated variables that are then applied to the real process.
2Measurement precision
If dynamic process simulation is performed in real-time to determine state variables for control, then control precision and adaptability are improved, but computing time and device complexity increase
Solution Approach 1:
The simulation model is pre-configured with process models and parameters before real-time operation. During commissioning and operation, the simulation rapidly calculates simulated state variables based on current manipulated variables and process conditions. This preliminary preparation of the simulation environment enables fast real-time computation without requiring complex on-the-fly model development, thus achieving high state variable accuracy while minimizing computing time delays.
3Adaptability or versatility
If machine learning is used to replace control programs for determining manipulated variables, then adaptability and control precision are improved, but ease of operation and device complexity worsen
Solution Approach 1:
The system implements a feedback mechanism where simulated state variables from the simulation model are continuously compared with actual measured process variables. The differences (deviations) between simulated and actual values are fed back to the machine learning module, which uses this feedback to adjust and optimize its predictions of manipulated variables. This feedback loop enables the machine learning system to adapt to changing plant conditions while maintaining ease of operation through automated adjustment without requiring manual reconfiguration.
Data Source
Figure 1
Figure 2
AI summary
The present invention relates to a method for the closed-loop control of a chemical process carried out in an industrial-scale chemical installation (140), having the following steps of: - capturing process data relating to the chemical installation (140) using sensors, - transmitting the process data to a control system (110) via a field bus (100), - transmitting at least one subset of the process data from the control system (110) to a computer system (120), wherein the computer system (120) has a simulation program (128) for stationary and/or dynamic process simulation of the chemical process, a closed-loop control program (129) for implementing a closed-loop controller for the chemical process and a memory (122) for storing simulated state variables (124), - calculating the simulated state variables of the chemical process in a cyclically repeated manner from the subset of the process data (123) by means of the simulation program (128) and storing the simulated state variables (124) in the memory (122), - transmitting at least one desired value of a control variable of the chemical process to the closed-loop control program (129), - reading at least one subset of the simulated state variables (124) from the memory (122) for input to the closed-loop control program (129), - determining a manipulated variable in order to achieve the desired value (113) by means of the closed-loop control program (129) by processing the simulated state variables (124) read from the memory (122), - transmitting the calculated manipulated variable to the control system (110).