A Review of Recent AI Applications in Next-Generation Power Electronics

Abstract

This comprehensive review examines the potential applications of generative and non-metaheuristic AI-driven algorithms in the control, maintenance, and design of power electronics and systems (PELS). The work covers diverse models including GANs, neural networks, quantum neural networks (QNNs), fuzzy systems, and reinforcement learning. The authors analyze implications for efficiency, reliability, and sustainability while documenting recent funding initiatives in the U.S. and Europe. The paper provides a development timeline of AI strategies and their initial applications to power electronics systems.

Key Contributions

• Assessment of big data's significance in power electronics systems and implications for AI-driven control
• Survey of state-of-the-art AI implementations across control, maintenance, and design phases of PELS
• Evaluation of both established (reinforcement learning, neural networks, fuzzy logic) and emerging (GANs, quantum neural networks) AI methodologies
• Documentation of recently funded research projects across U.S. and European sectors
• Development timeline mapping when key AI strategies were first applied to power electronics
• Analysis of implications for efficiency, reliability, and sustainability in renewable energy and EV infrastructure

Methodology

Systematic literature review synthesizing recent publications on AI-driven approaches for power electronics, covering both practical implementations and emerging theoretical advances. The review spans control algorithms, predictive maintenance applications, and design optimization methods. Covers funded initiatives at national and international level to identify research momentum and investment priorities.

Results

• GANs identified as emerging tools for synthetic data generation and fault simulation in power electronics testing
• Quantum neural networks highlighted as a frontier methodology with early-stage PELS applications
• Reinforcement learning demonstrated strong results for adaptive control in variable-load power converter scenarios
• Fuzzy logic systems shown to provide interpretable control layers for hybrid AI-physical PELS designs
• Review identifies transformative potential of AI for enhancing power electronics performance and sustainability
• Documentation of funded projects signals significant institutional investment in AI-PELS convergence (U.S. DOE and EU Horizon programmes)

Limitations

• Specific numerical benchmarks and head-to-head quantitative comparisons were limited in the available excerpt; full results in the 31-page article
• Quantum neural network applications remain largely theoretical with limited real-world PELS deployment demonstrated
• Long-term reliability and certification pathway for AI-controlled power electronics not fully addressed
• Generalization across diverse grid topologies and power ratings requires further empirical validation

Source: A Review of Recent AI Applications in Next-Generation Power Electronics by Safari, Oshnoei, Blaabjerg, Applied Energy Vol. 402 (2025)