Simulation of a closed-loop dc-dc converter using a physics-informed neural network-based model

2025-06-23

AI Summary

Commercial time-domain simulation software for power electronic systems—such as closed-loop DC–DC boost converters—faces a fundamental trade-off between simulation speed and accuracy. To address this, this paper proposes a physics-informed bidirectional long short-term memory neural network (BiLSTM-PINN) modeling framework. By embedding the underlying circuit differential equations directly into the BiLSTM architecture, the model achieves high-fidelity, computationally efficient dynamic response prediction across multiple operating points, parameter variations, and external disturbances. Quantitative evaluation shows that BiLSTM-PINN reduces median RMSE by 9× and prediction standard deviation by 2.6× compared to conventional fully connected neural networks (FCNNs), significantly improving both accuracy and robustness. It further outperforms purely data-driven BiLSTM by 4.5× in RMSE reduction and 1.7× in standard deviation reduction. This work establishes a verifiable, engineering-practical paradigm for deploying physics-informed deep learning in real-time power electronics simulation.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Hardware-aware ML
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

The growing reliance on power electronics introduces new challenges requiring detailed time-domain analyses with fast and accurate circuit simulation tools. Currently, commercial time-domain simulation software are mainly relying on physics-based methods to simulate power electronics. Recent work showed that data-driven and physics-informed learning methods can increase simulation speed with limited compromise on accuracy, but many challenges remain before deployment in commercial tools can be possible. In this paper, we propose a physics-informed bidirectional long-short term memory neural network (BiLSTM-PINN) model to simulate the time-domain response of a closed-loop dc-dc boost converter for various operating points, parameters, and perturbations. A physics-informed fully-connected neural network (FCNN) and a BiLSTM are also trained to establish a comparison. The three methods are then compared using step-response tests to assess their performance and limitations in terms of accuracy. The results show that the BiLSTM-PINN and BiLSTM models outperform the FCNN model by more than 9 and 4.5 times, respectively, in terms of median RMSE. Their standard deviation values are more than 2.6 and 1.7 smaller than the FCNN's, making them also more consistent. Those results illustrate that the proposed BiLSTM-PINN is a potential alternative to other physics-based or data-driven methods for power electronics simulations.

Problem

• Develop fast, accurate physics-informed neural network for dc-dc converter simulation
• Compare BiLSTM-PINN performance with FCNN in power electronics modeling
• Address speed-accuracy tradeoff in data-driven power electronics simulation tools

Innovation

• Physics-informed BiLSTM for DC-DC simulation
• Combines data-driven and physics-based methods
• Outperforms FCNN in accuracy and consistency