Technology Category
Machine Learning: Hardware-aware MLSearch and Optimization: Mixed Discrete/Continuous SearchCognitive Modeling & Cognitive Systems: Neural Spike Coding
Neural Substitute Solver for Efficient Edge Inference of Power Electronic Hybrid Dynamics
🤖 AI Summary
Real-time inference of hybrid continuous–discrete dynamics in power electronic systems (PES) remains challenging on resource-constrained edge devices due to computational bottlenecks in traditional solvers. To address this, we propose the Neural Surrogate Solver (NSS) framework, which replaces computationally intensive matrix operations and high-order numerical integration with lightweight, parallelizable neural network modules. This design overcomes sequential computation limitations while preserving modeling fidelity. To our knowledge, NSS is the first approach enabling high-fidelity, real-time PES dynamic inference directly on edge hardware. Experimental deployment on a multilevel DC–DC converter demonstrates a 23× speedup in inference latency and a 60% reduction in FPGA resource utilization compared to conventional solvers. The framework thus significantly expands the feasibility of high-accuracy dynamic simulation and closed-loop control at the edge.
Technology Category
Application Category
📝 Abstract
Advancing the dynamics inference of power electronic systems (PES) to the real-time edge-side holds transform-ative potential for testing, control, and monitoring. How-ever, efficiently inferring the inherent hybrid continu-ous-discrete dynamics on resource-constrained edge hardware remains a significant challenge. This letter pro-poses a neural substitute solver (NSS) approach, which is a neural-network-based framework aimed at rapid accurate inference with significantly reduced computational costs. Specifically, NSS leverages lightweight neural networks to substitute time-consuming matrix operation and high-order numerical integration steps in traditional solvers, which transforms sequential bottlenecks into highly parallel operation suitable for edge hardware. Experimental vali-dation on a multi-stage DC-DC converter demonstrates that NSS achieves 23x speedup and 60% hardware resource reduction compared to traditional solvers, paving the way for deploying edge inference of high-fidelity PES dynamics.
Problem
Research questions and friction points this paper is trying to address.
Efficient edge inference for power electronic hybrid dynamics
Reducing computational costs in resource-constrained edge hardware
Replacing traditional solvers with lightweight neural networks
Innovation
Methods, ideas, or system contributions that make the work stand out.
Neural-network-based substitute solver for edge inference
Lightweight networks replace matrix operations and integration
Enables parallel operations on resource-constrained hardware
🔎 Similar Papers
No similar papers found.
