Reformer Feed and Product Property Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current naphtha reforming processes face challenges in real-time control of reactor operations due to the difficulty in predicting properties of feed and products, leading to inefficient process management.
Innovation Solution
A method and apparatus that utilize trained predictive models to forecast the properties of feed and products in real-time, using current operating conditions and historical data, enabling precise control of reactor operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If real-time prediction of feed and product properties is implemented, then reactor operation control is improved, but device complexity increases due to the need for predictive models and computing resources
Solution Approach 1:
The patent replaces physical measurement devices and manual analysis methods with a computational prediction system. Predictive models (neural networks, random forests, support vector machines) substitute for traditional online analyzers and manual property determination, reducing hardware complexity while improving operational control capability
Solution Approach 2:
The patent introduces predictive models as intermediary components between operating conditions and product properties. These models act as virtual sensors that translate readily measurable operating parameters into predicted feed and product properties, avoiding the need for complex direct measurement systems
2Measurement precision
If multiple predictive models are trained and deployed, then prediction accuracy is improved, but loss of time increases during model training and deployment
Solution Approach 1:
The patent performs model training in advance using historical data before deployment. Multiple predictive models are pre-trained offline with comprehensive training sets, allowing accurate predictions during real-time operation without time loss. The training phase is separated from the prediction phase, so operational time is not affected
Solution Approach 2:
The patent trains and deploys multiple predictive models (exceeding the minimum single model) to achieve higher prediction accuracy. By using an ensemble of models (neural networks, random forests, support vector machines), the system achieves superior accuracy while the excessive training effort is performed offline rather than during operation
Data Source
AI summary
Disclosed are a method and apparatus of predicting properties of feed and products in a reformer. The method of predicting properties of feed and products in a reformer includes training a first predictive model for predicting the properties of feed in the reformer and a second predictive model for predicting the properties of products in the reformer; predicting the properties of feed being currently supplied to the reactor in real time by allowing a first prediction unit including the trained first prediction model to receive a current operating condition of the reactor in the reformer; and predicting the properties of products being produced in the reactor 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.