Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DT-MPC: Synthesizing Derivation-Free Model Predictive Control from Power Converter Netlists via Physics-Informed Neural Digital Twins

Jialin Zheng, Haoyu Wang, Yangbin Zeng, Han Xu, Di Mou, Hong Li, Patrick Wheeler, Sergio Vazquez, Leopoldo G. Franquelo

TL;DR

DT-MPC replaces manual, topology-specific MPC modeling with a directly synthesized, high-fidelity digital twin derived from circuit netlists. It combines a physics-informed neural surrogate predictor for real-time speed, a priority-based multi-objective cost, and a simplex-based optimizer to navigate continuous, high-dimensional control spaces, demonstrated on a 1500 W dual-active-bridge converter. The approach achieves faster-than-real-time inference, superior dynamic response, and competitive steady-state efficiency while drastically reducing engineering design time, representing a paradigm shift from model derivation to automated synthesis. This framework enables generalized, scalable control design for advanced power converters via cloud-edge-plant collaboration and hardware-accelerated computation.

Abstract

Model Predictive Control (MPC) is a powerful control strategy for power electronics, but it highly relies on manually-derived and topology-specific analytical models, which is labor-intensive and time-consuming in practical designs. To overcome this bottleneck, this paper introduces a Digital-Twin-based MPC (DT-MPC) framework for generic power converters that can systematically translate a high-level circuit into an objective-aware control policy by leveraging a DT as a high-fidelity system model. Furthermore, a physics-informed neural surrogate predictor is proposed to accelerate predictions by DT and enable real-time operation. A gradient-free simplex search optimizer is also introduced to efficiently handle complex multi-objective optimization. The efficacy of the framework has been validated through a cloud-to-edge deployment on a 1500 W dual active bridge converter. Experimental results show that the synthesized predictive model achieves an inference speed over 7 times faster than real time, the DT-MPC controller outperforms several human-designed counterparts, and the overall framework reduces engineering design time by over 95\%, verifying the superiority of DT-MPC on generalized power converters.

DT-MPC: Synthesizing Derivation-Free Model Predictive Control from Power Converter Netlists via Physics-Informed Neural Digital Twins

TL;DR

DT-MPC replaces manual, topology-specific MPC modeling with a directly synthesized, high-fidelity digital twin derived from circuit netlists. It combines a physics-informed neural surrogate predictor for real-time speed, a priority-based multi-objective cost, and a simplex-based optimizer to navigate continuous, high-dimensional control spaces, demonstrated on a 1500 W dual-active-bridge converter. The approach achieves faster-than-real-time inference, superior dynamic response, and competitive steady-state efficiency while drastically reducing engineering design time, representing a paradigm shift from model derivation to automated synthesis. This framework enables generalized, scalable control design for advanced power converters via cloud-edge-plant collaboration and hardware-accelerated computation.

Abstract

Model Predictive Control (MPC) is a powerful control strategy for power electronics, but it highly relies on manually-derived and topology-specific analytical models, which is labor-intensive and time-consuming in practical designs. To overcome this bottleneck, this paper introduces a Digital-Twin-based MPC (DT-MPC) framework for generic power converters that can systematically translate a high-level circuit into an objective-aware control policy by leveraging a DT as a high-fidelity system model. Furthermore, a physics-informed neural surrogate predictor is proposed to accelerate predictions by DT and enable real-time operation. A gradient-free simplex search optimizer is also introduced to efficiently handle complex multi-objective optimization. The efficacy of the framework has been validated through a cloud-to-edge deployment on a 1500 W dual active bridge converter. Experimental results show that the synthesized predictive model achieves an inference speed over 7 times faster than real time, the DT-MPC controller outperforms several human-designed counterparts, and the overall framework reduces engineering design time by over 95\%, verifying the superiority of DT-MPC on generalized power converters.

Paper Structure

This paper contains 22 sections, 1 theorem, 25 equations, 15 figures, 5 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Problem Definition and DT-MPC Framework
  3. Problem Definition of MPC in Power Converters
  4. DT-MPC: The Direct Synthesis Paradigm
  5. Core Components and Implementation of DT-MPC
  6. Neural Surrogate Predictor for Real-Time Execution
  7. Priority-Based Cost Function for Multi-Objective Control
  8. Simplex Search Optimizer for Control Space Exploration
  9. Overall Implementation Workflow
  10. Case Setup and DT-MPC Deployment
  11. Case Study: DAB Converter
  12. System Architecture: Cloud, Edge, and Plant
  13. End-to-End Implementation Workflow
  14. Phase 1: Framework Configuration and Offline Training
  15. Phase 2: Model Preparation and Hardware Integration
  16. ...and 7 more sections

Key Result

Theorem 1

Let the cost function $J: \mathbb{R}^n \to \mathbb{R}$ be continuously differentiable ($C^1$). Assume the sequence of iterates $\{U_k\}$ generated by the algorithm remains within a bounded set $\mathcal{D}$, and that the gradient $\nabla J$ is Lipschitz continuous on $\mathcal{D}$, i.e., there exist Then, any cluster point of the sequence $\{U_k\}$ is a stationary point of $J$.

Figures (15)

  • Figure 1: The basic components of MPC
  • Figure 2: The proposed DT-MPC direct synthesis workflow.
  • Figure 3: Core components of the DT-MPC framework. (a) Block diagram of the programmatically generated Digital Twin model. (b) Illustration of the event-driven solver. (c) DT-MPC framework. (d) The process of formulating the priority-based multi-objective cost function.
  • Figure 4: (a) Traditional high-order mathematic solver. (b) Proposed neural surrogate predictor.
  • Figure 5: Geometric steps of one simplex search iteration. (a) Initial simplex. (b) Identify worst vertex ($J_{max}$) and centroid ($g$). (c) Reflect $J_{max}$ through $g$. (d) Expand simplex if reflection is very good. (e) contract if reflection is poor. (f) If no improvement, poll directions around best vertex ($J_{min}$). (g) If polling fails, shrink simplex towards $J_{min}$.
  • ...and 10 more figures

Theorems & Definitions (2)

  • Theorem 1: Convergence to Stationary Points
  • proof