NMR-Based Yield Prediction for Hydrocarbon Process Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current hydrocarbon processing systems rely on approximate human decisions due to the complexity of variables affecting product yield and quality, with existing NMR spectroscopy applications limited to optimizing specific processes like desalter operations, lacking comprehensive end-to-end optimization across hydrocarbon processing systems.
Innovation Solution
A system and method utilizing NMR spectroscopy data to generate yield prediction models that optimize hydrocarbon processing system configurations through a trained yield prediction model and optimization module, incorporating feedstock and process characteristics to achieve improved objective metrics such as revenue, efficiency, and product throughput across all process units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If NMR spectroscopy is used to analyze crude feedstock, then feedstock characteristics can be determined, but the analysis is limited to determining only the concentration of various components, lacking comprehensive optimization capability
Solution Approach 1:
The patent extends NMR spectroscopy applications from single-component concentration analysis to comprehensive multi-parameter optimization. The system uses NMR data to predict multiple product yields across different process units (desalter, atmospheric distillation, vacuum distillation, hydrocracking, coking) simultaneously, enabling the same analytical technique to serve multiple optimization functions across the entire refining process chain.
Solution Approach 2:
The patent transitions from one-dimensional concentration measurement to multi-dimensional optimization by incorporating temporal and process-unit dimensions. The system provides time-course predictions for multiple process units and integrates these predictions with operational constraints to optimize multiple objectives (revenue, efficiency, environmental impact) simultaneously, adding temporal and multi-objective dimensions to the traditional single-point concentration analysis.
2Ease of operation
If human operators configure processing processes based on experience, then certain process configurations can be identified, but the number of variables and complexity of relationships makes decisions crude and approximate
Solution Approach 1:
The patent replaces human operator expertise with an automated computational system. The system uses machine learning models trained on historical data to predict product yields and optimize process configurations automatically. This substitution of mechanical human decision-making with automated algorithmic processing eliminates the crudity and approximation inherent in experience-based operations while maintaining ease of use through automated decision support.
Solution Approach 2:
The system implements feedback mechanisms by continuously monitoring actual process performance against predicted yields and using this information to refine optimization recommendations. The machine learning models learn from historical operational data and update their predictions, creating a feedback loop that improves optimization accuracy over time while maintaining ease of operation through automated iterative refinement.
3Adaptability or versatility
If existing optimization systems are limited to single process units like desalter, then specific process objectives can be achieved, but comprehensive end-to-end optimization across the entire hydrocarbon processing system cannot be realized
Solution Approach 1:
The patent divides the complex hydrocarbon processing system into discrete process units (desalter, atmospheric distillation, vacuum distillation, hydrocracking, coking), each with its own optimization model. This segmentation allows the system to handle complexity by breaking it down into manageable components that can be optimized independently but are coordinated through integrated constraints and objectives, enabling comprehensive end-to-end optimization without overwhelming complexity.
Solution Approach 2:
The system merges the optimization capabilities for individual process units into a unified end-to-end optimization framework. By combining the predictions from multiple process unit models and integrating them with shared constraints (feedstock quality, product demands, operational limits) and objectives (revenue, efficiency, environmental impact), the system achieves comprehensive optimization across the entire refining chain while managing complexity through integrated coordination.
4Adaptability or versatility
If existing NMR applications focus on single objective optimization like mitigating hydrochloric acid formation, then specific process problems can be addressed, but multi-objective optimization for revenue, efficiency, and environmental impact cannot be achieved
Solution Approach 1:
The patent creates a universal optimization framework that can simultaneously address multiple objectives (revenue maximization, efficiency improvement, environmental impact reduction) within a single integrated system. The machine learning models are configured to evaluate and optimize for all these objectives concurrently, allowing the same system to serve multiple optimization purposes without requiring separate specialized models for each objective.
Solution Approach 2:
The system manages model configuration complexity by dynamically adjusting optimization parameters and weights based on user priorities and operational constraints. Rather than requiring complex manual configuration for each objective, the system allows flexible parameter modification through standardized interfaces, enabling multi-objective optimization while keeping the underlying model complexity managed through automated parameter tuning and standardized workflows.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables comprehensive end-to-end optimization of hydrocarbon processing systems, improving product yields and system efficiency by adjusting operating parameters based on predicted yields and constraints, thereby enhancing objective metrics like revenue and environmental impact.
Implementation Method 1
A nuclear magnetic resonance (NMR) scan of a sample of a hydrocarbon feedstock material is performed to generate NMR data
Data Source
AI summary
There are provided systems, methods, and processor-readable media for optimizing the end to end operation of a hydrocarbon processing system using NMR spectroscopy data. The hydrocarbon processing system is configured to process hydrocarbon feedstock material, via a configuration defined by a set of configuration parameters, such that one or more final products are produced. A yield prediction model is used to process NMR data obtained from an NMR scan of the feedstock material to generate yield prediction data. The yield prediction data includes predictions of product yields of various intermediate products and/or final products. An optimization module is used to process the yield prediction data predicted by the trained yield prediction model such that optimized configuration parameter data is obtained. The optimized configuration parameter data is effective for establishing an optimized configuration for operating the hydrocarbon processing system that optimizes or improves an objective metric.