Review and Outlook on Power Converters Exploiting Artificial Neural Networks: Recent Advances and Perspectives

Power converters are highly nonlinear systems that require complex design, analysis, and control efforts. Artificial Neural Networks learn from and adapt to training data, resulting in a valuable solution to match the nonlinear behavior without explicit mathematical models. Based on the most relevant studies from the previous five years, this paper thoroughly reviews neural network applications in power conversion systems. This paper proposes a thorough comparison of existing solutions and an open discussion on significant and emerging subtopics, such as black-box modeling, converter parameter identification, advanced control methods, and reliability-related issues. Several classifications and statistics are carried out and graphically outlined for a better comprehension of the state-of-the-art. Final remarks of each sub-topic are provided, including a critical analysis of existing solutions, performances, and perspectives. The real-time implementations, challenges, and research trends are also discussed. This study emphasizes promising fields, such as hybrid models that combine knowledge-based and data-driven methodologies, to outline future research perspectives. Reviewing the most recent and relevant state-of-the-art neural network applications in power electronics, this study provides insight into future developments for industrial applications as well.