Model Predictive Control Computing Graph for Real-Time Parallel Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current model prediction control methods face significant computational bottlenecks due to serial computing approaches, leading to limited speed upgrades and inefficiencies in processing complex optimization problems, particularly in dynamic systems like electronic power and automobile electronics.
Innovation Solution
A method for real-time optimization and parallel computing of model prediction control using a computing chart, which involves building a prediction model and target function, employing a parallel computing architecture, and utilizing back propagation and gradient descent to optimize control amounts, enabling parallel processing and reducing computational time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If serial computing approach is used for model prediction control, then computing accuracy is maintained, but computing speed is limited and computing time is excessive
Solution Approach 1:
The patent segments the computing process into independent parallel tasks by introducing a computing chart that divides the prediction model and target function computations into separate, concurrently executable segments. This segmentation enables multiple computing operations to proceed simultaneously, directly addressing the speed limitation of serial computing while maintaining computational accuracy through structured task division.
Solution Approach 2:
The patent transitions from one-dimensional serial computing to multi-dimensional parallel computing by implementing a computing chart architecture that executes multiple computing tasks across different dimensions simultaneously. This dimensional transformation allows the system to perform prediction model computations, target function evaluations, and optimization iterations in parallel, dramatically reducing total computing time while preserving result accuracy.
2Productivity
If standard optimization algorithms are used, then solution accuracy is achieved, but computing complexity increases and parallel acceleration space is limited
Solution Approach 1:
The patent segments the optimization process into distinct, independently computable modules within the computing chart framework. By dividing the optimization iterations into separate tasks that can be executed in parallel across multiple computing nodes, the system reduces overall computing complexity while maintaining solution accuracy. Each segment handles specific computational aspects, allowing for efficient parallel implementation.
Solution Approach 2:
The computing chart serves as an intermediary structure that coordinates and manages parallel optimization computations. It acts as a mediator between the prediction model, target function, and optimization algorithm, organizing complex computations into manageable parallel tasks. This intermediary framework simplifies the overall computing complexity by providing a structured approach to parallel optimization while preserving the accuracy of standard optimization algorithms.
Data Source
AI summary
The disclosure relates to a method for real time optimization and parallel computing of model prediction control based on a computing chart, comprising the following steps: building a prediction model of a system state amount and building a target function of a system; building a parallel computing architecture for model prediction control of a prediction model and the target function and employing a triggering parallel computing method by the parallel computing architecture to synchronously compute the prediction model and the target function; and solving and computing a gradient with a manner of back propagation and using a gradient descent method to optimize a control amount of the system and realize real time optimal control of the system. Compared with the prior art, the present disclosure greatly improves a computing efficiency, ensures real time property of a model prediction controller, and extends application fields of model prediction control.