Operation guidance searching method and operation guidance searching system
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In liquefied natural gas (LNG) plants, it is challenging to identify efficient operation adjustments across numerous devices affected by various disturbances, such as outside air temperature, supply pressure, and gas composition, due to complex correlations between devices and disturbances.
Innovation Solution
An operation guidance searching method using machine learning to generate a plant model and reinforcement learning to optimize compressor power per unit production of LNG, minimizing compression power while maintaining an outlet temperature within a preset restriction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If traditional operation adjustment methods are used in LNG plants, then operation stability is maintained, but compression power consumption cannot be optimized under disturbance conditions
Solution Approach 1:
The patent introduces a computer as an intermediary that executes machine learning and reinforcement learning algorithms to optimize operation adjustments. The computer processes disturbance data and operation data to generate optimized manipulated variable values, resolving the contradiction by offloading complex optimization calculations from the control system while maintaining energy efficiency.
Solution Approach 2:
The patent replaces traditional mechanical control systems with information processing systems based on machine learning and reinforcement learning. By substituting physical control mechanisms with computational algorithms that analyze disturbance patterns and optimize operations, the system reduces compression power consumption without requiring complex mechanical modifications.
2Adaptability or versatility
If the number of devices in LNG plant is increased to handle various disturbances, then operational flexibility is improved, but identification of efficient operation adjustments becomes more difficult
Solution Approach 1:
The patent implements a universal operation optimization system that handles multiple disturbance types and device configurations through a single reinforcement learning framework. The system processes various disturbance data (outside air temperature, supply pressure, composition) and generates optimized adjustments for multiple devices, providing operational flexibility without increasing identification difficulty.
Solution Approach 2:
The patent employs feedback mechanisms where the reinforcement learning algorithm continuously monitors operation data and disturbance data, compares actual outcomes with expected outcomes, and adjusts manipulated variable values accordingly. This feedback loop enables the system to identify efficient operation adjustments automatically, even as the number of devices and disturbance types increases.
3Productivity
If machine learning and reinforcement learning are implemented for operation optimization, then compression power per unit production is minimized, but computational requirements and system complexity increase
Solution Approach 1:
The patent performs preliminary training of machine learning models using historical operation data and disturbance data before actual optimization is needed. By pre-training the reinforcement learning algorithm with extensive computational work beforehand, the system minimizes real-time computational requirements during actual operation, reducing the burden on computational resources while maintaining optimization performance.
Data Source
AI summary
Provided is a technology of searching for operation guidance for efficiently operating a liquefied natural gas plant. An operation guidance searching method for a liquefied natural gas plant includes: acquiring data sets of operation data of process variables for a plurality of target devices and disturbance data; generating, through machine learning, a plant model indicating correspondences of output values of process variables with respect to manipulated variables and input values of disturbances; and searching, through reinforcement learning, for input values of the manipulated variables operation variables for with which a compression power per unit production amount is minimized under a condition in which an outlet temperature of a liquefied natural gas is a preset restriction temperature or lower.