Neural Network Predictive Control for Optimal Process Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional approaches to controlling continuous processes using neural networks lack assurance that the control signals generated will be optimal in an online environment, as they are typically trained to mimic historical controllers without guaranteeing optimal performance.
Innovation Solution
A predictive control system that uses a neural network model to predict controlled variables, evaluate an objective function, and perform a predictive optimization process to generate optimal manipulated variables, iteratively adjusting inputs to converge on optimal control objectives, while also considering disturbance variables and scenario-based optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a controller neural network is trained to mimic the behavior of a traditional controller using historical training data, then the controller neural network can be substituted for the online controller, but there is no guarantee that the control signals generated are optimal when implemented in the online environment
Solution Approach 1:
The patent creates a predictor neural network that copies and learns the dynamic behavior of the plant from historical data, rather than copying the controller's decision-making process. This predictor model is then used by an optimization algorithm to generate optimal control signals, ensuring both adaptability and optimality guarantees.
Solution Approach 2:
The predictor neural network serves as an intermediary between the historical process data and the optimization algorithm. It translates complex plant dynamics into predicted outputs that the optimizer can use to determine optimal control actions, bridging the gap between data-driven modeling and optimization-based control.
2Speed
If a neural network model is used to predict controlled variables in real-time, then the prediction speed is fast, but the model requires extensive training data and computational resources for accurate predictions
Solution Approach 1:
The predictor neural network is trained offline using historical process data before deployment. This preliminary training phase allows the model to learn complex plant dynamics in advance, so that during online operation, predictions can be made rapidly without requiring extensive real-time computation or additional training data.
3Ease of manufacture
If traditional control methods are used, then the control system is simple to implement, but the control signals generated are not optimal for improving yield and quality of hydrocarbon products
Solution Approach 1:
The patent replaces traditional optimization methods with a data-driven approach using neural networks for prediction combined with optimization algorithms. This substitution enables the system to handle complex nonlinear relationships in hydrocracking and hydrotreating processes, achieving optimal control for improved product yield and quality while maintaining practical implementability through modular system architecture.
Data Source
AI summary
A predictive control system includes controllable equipment and a controller. The controller is configured to use a neural network model to predict values of controlled variables predicted to result from operating the controllable equipment in accordance with corresponding values of manipulated variables, use the values of the controlled variables predicted by the neural network model to evaluate an objective function that defines a control objective as a function of at least the controlled variables, perform a predictive optimization process to generate optimal values of the manipulated variables for a plurality of time steps in an optimization period using the neural network model and the objective function, and operate the controllable equipment by providing the controllable equipment with control signals based on the optimal values of the manipulated variables generated by performing the predictive optimization process.