Applications of Machine Learning and Artificial Intelligence in Modern Power and Energy Systems

1. Introduction

The digital transformation of the energy sector has intensified the demand for intelligent, data-driven, and computationally efficient techniques to bolster the reliability, safety, and adaptability of modern electrical infrastructures. This Special Issue, entitled “Applications of Machine Learning and Artificial Intelligence in Modern Power and Energy Systems”, presents a peer-reviewed collection of 11 high-quality papers. These developments cover the entire spectrum of modern energy—from physically based simulations and diagnostics of high-voltage systems to advanced machine learning for predicting the load and stability of the power grid. Collectively, the papers reflect an evolving research landscape where hybrid modeling, explainable AI (XAI), and proactive maintenance are becoming central to next-generation intelligent power systems.

2. Physics-Based and Hybrid Modeling

Contribution 1 investigated the data-driven dynamic behavior of SF₆ HVDC-GIS conical solid insulators. Using COMSOL Multiphysics 6.2 simulations, the authors analyzed ion-pair generation and dielectric performance under specific radioactive conditions, providing a robust framework for assessing insulation reliability.

Contribution 2 evaluated the impact of voltage-source converters (VSCs) on power systems by developing a refined P–Q capability curve. The work offers vital insights for renewable-integrated grid planning by accounting for the iterative nature of converter–grid interactions.

3. AI for State Estimation and Load Forecasting

Contribution 3 proposed an interpretable state estimation framework based on Kolmogorov–Arnold Networks (KANs), enhancing the transparency and accuracy of grid stability assessments compared with traditional neural network architectures.

Contribution 4 introduced a novel multi-scale wind power prediction model that accounts for temporal dependencies and cross-scale variable relationships, significantly reducing uncertainty in wind energy integration.

Contribution 5 examined the advancements in household load forecasting through a deep learning model optimized with meta-heuristic hyperparameter tuning, demonstrating robust performance across diverse energy consumption profiles.

4. Equipment Diagnostics and Proactive Maintenance

Contribution 6 developed an innovative methodology for the proactive maintenance of distribution transformers. By utilizing machine learning algorithms for predictive classification, the study identified transformers vulnerable to failure, enabling system operators to shift from reactive to cost-effective proactive maintenance strategies.

Contribution 7 proposed an LSTM-driven HVDC fault diagnosis framework enhanced with knowledge graphs. This hybrid approach leverages both historical data and domain-specific knowledge to increase the accuracy and explainability of fault localization.

Contribution 8 contributed to predictive maintenance by developing a deep convolutional neural network (CNN) for the automated detection and classification of rolling bearing defects in electrical machinery.

Contribution 9 explored fuel-consumption prediction models for electrified transportation, incorporating dynamic parameters such as vehicular “jerk” to achieve higher predictive fidelity.

5. Digital Twins and Reinforcement Learning

Contribution 10 improved parameter extraction in 3D grid information models (GIMs) using an enhanced CNN architecture, streamlining the creation of automated digital twins for power system analysis.

Contribution 11 conducted a comprehensive review of reinforcement learning (RL) applications in energy systems. Their survey covers the state of the art in optimization, control, and operational decision-making for both renewable and conventional resources.

6. Conclusions

This Special Issue highlights the transformative role of data mining techniques and computational intelligence in modern energy systems. The collected materials demonstrate the maturity of hybrid modeling approaches based on physics and machine learning, along with significant advances in renewable energy production forecasting and electricity load prediction. At the same time, the studies highlight the growing importance of the interpretability, reliability, and system-level integration of models, which are essential for the practical implementation of smart solutions in real-world energy systems.

Furthermore, the increasing adoption of predictive maintenance strategies and digital twin concepts illustrates the shift towards proactive asset management and increased grid resilience. The wider adoption of reinforcement learning methods for energy optimization further highlights the potential of smart control and decision-making frameworks in addressing the complexity of future energy infrastructures. Overall, the papers in this Special Issue point to a future characterized by smart, adaptive, and sustainable electrical systems capable of meeting the technical, economic, and sustainability challenges of the evolving energy landscape.