Automated Physical System Control with Long-Horizon Forecasts and MILP
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current process control methods face challenges in managing system inertia, optimizing non-linear parameter spaces, and integrating multi-variate forecasting models with optimization formulation, particularly in long-horizon trajectory optimization and multi-step set-point recommendations.
Innovation Solution
A system that trains multivariate time series forecasting models to predict long-horizon action trajectories, linearizes the state-based action response model, and reformulates it into a mixed-integer linear program (MILP) for efficient optimization, enabling multi-step set-point recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multivariate time series forecasting models are used to predict long-horizon action trajectories, then forecasting accuracy is improved, but device complexity increases
Solution Approach 1:
The patent segments the complex multivariate time series forecasting model into multiple linear models, each responsible for predicting a specific time step in the trajectory. This segmentation allows the system to handle long-horizon forecasting through a series of simpler linear predictions rather than one complex non-linear model, thereby maintaining forecasting accuracy while reducing individual model complexity.
Solution Approach 2:
The patent introduces linear models as intermediary components between the control inputs and the final trajectory prediction. These linear models serve as mediators that approximate the behavior of the complex non-linear system at each time step, enabling the overall system to achieve accurate long-horizon forecasting through composition of simpler intermediate predictions.
2Manufacturing precision
If the system optimizes non-linear parameter spaces, then optimization precision is improved, but computational complexity increases
Solution Approach 1:
The patent replaces complex non-linear optimization mechanisms with linear programming formulations. By substituting the non-linear parameter space optimization with a series of linear models and mixed-integer linear programming (MILP) formulations, the system achieves comparable optimization precision while significantly reducing computational complexity and enabling efficient solution through standard linear optimization solvers.
3Reliability
If multi-step set-point recommendations are generated, then control effectiveness is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary linearization of the system dynamics and pre-computes the linear models for each time step before generating the actual trajectory prediction. This preliminary action allows the system to quickly generate multi-step set-point recommendations by simply composing the pre-computed linear predictions, rather than performing complex non-linear calculations in real-time, thus maintaining control effectiveness while reducing processing time.
Data Source
AI summary
One or more systems, devices, computer program products and/or computer-implemented methods of use provided herein relate to automated control for a physical system with generic forecasting models. The computer-implemented system can comprise a memory that can store computer executable components. The computer-implemented system can further comprise a processor that can execute the computer executable components stored in the memory, wherein the computer executable components can comprise a training component that trains one or more multivariate time series forecasting models, and a prediction component that forecasts long-horizon action trajectories for a set of state variables over a defined range of time. Furthermore, an analysis component can linearize the state-based action response model, wherein the linearized the state-based action response model can be formulated into an MILP optimization problem and solved to obtain a multi-step set-point recommendation.