Artificial Intelligence Applications in Electric Power and Energy Systems: A PRISMA-Bibliometric Review of Tools, Application Domains, Quantified Impacts, and Deployment Barriers from 2021 to 2025

Artificial intelligence (AI) is increasingly used in electric power and energy systems for forecasting, fault diagnosis, stability assessment, control, renewable energy integration, and real-time monitoring. However, evidence remains fragmented across algorithms, applications, performance metrics, and deployment conditions. This study reviews recent AI applications in power and energy systems and examines their technological, economic, and operational contributions and limitations. A PRISMA-based systematic and bibliometric review was conducted using Scopus and Web of Science records published from 2021 to 2025. Bibliometrix/Biblioshiny 5.5.0 and VOSviewer 1.6.20 were used to analyze publication trends, citation patterns, geographical distribution, influential studies, and keyword networks, while 40 studies were retained for systematic evidence synthesis. In the reconciled 4854-record corpus, annual scientific production increased from 610 publications in 2021 to 1305 in 2025, with research concentrating on machine learning, deep learning, smart grids, stability, renewable energy, forecasting, fault diagnosis, optimization, and cybersecurity. The reviewed studies reported improvements in predictive accuracy, computational efficiency, fault localization, voltage control, energy savings, and real-time decision support. Nevertheless, data dependence, computational cost, limited interpretability, simulation-to-field transfer, and validation requirements remain major barriers. Future research should prioritize robust, explainable, physics-aware, and field-validated AI for critical power-system applications.