Reformer Reactor Temperature Control Using Real-Time Property Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for controlling reactors in naphtha reforming processes struggle to operate in real time, leading to difficulties in predicting and maximizing the properties of products, particularly the content of aromatic components, due to reliance on daily measurement data and operator experience.
Innovation Solution
A method and apparatus that utilize trained prediction models to predict feed and product properties in real time, allowing for real-time operation control by calculating temperature adjustments based on predicted product properties, thereby optimizing aromatic component content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If real-time prediction models are implemented, then productivity and control responsiveness are improved, but device complexity increases
Solution Approach 1:
The patent replaces the mechanical/manual measurement and control system with a computational prediction model system. Instead of relying on physical measurement devices and operator experience, the invention uses trained prediction models (neural networks or machine learning algorithms) to predict product properties in real-time based on operating conditions, thereby achieving real-time control without proportional increases in physical device complexity
Solution Approach 2:
The prediction models create virtual copies of the physical reforming process by learning from historical data. The models replicate the complex chemical transformation relationships without requiring physical measurement of every parameter, allowing real-time prediction and control while maintaining manageable system complexity
2Measurement precision
If measurement frequency is increased to daily or real-time, then measurement precision and control accuracy are improved, but loss of time and operational cost increase
Solution Approach 1:
The prediction models are trained in advance on historical data to learn the relationships between operating conditions and product properties. This preliminary training action enables the models to provide real-time predictions during operation without requiring actual real-time measurement of product properties, thus achieving fast response without continuous measurement overhead
Solution Approach 2:
The prediction model acts as an intermediary between operating conditions and product properties. Instead of directly measuring product properties (which takes time), the model mediates by predicting properties from easily measurable operating parameters, providing fast and accurate information for control decisions
Data Source
AI summary
Disclosed are method and apparatus for controlling a reactor in a reformer including training a first prediction model for predicting properties of feed and a second prediction model for predicting properties of products; predicting the properties of feed being currently supplied to a reactor set in real time by allowing a first prediction unit including the trained first prediction model to receive a current operating condition of the reactor set; predicting the properties of products being produced in the reactor set in real time by allowing a second prediction unit including the trained second prediction model to receive the current operating condition and the predicted properties of feed; calculating amount of temperature fluctuation for each reactor as a control signal for controlling each of the reactors; and controlling an operating temperature of each of the reactors based on the calculated amount of temperature fluctuation.