Rapid developments in large language models (LLMs) have created new opportunities for their use in the energy sector, from forecasting renewable energy to power system operation and energy market analysis. These models help improve decision-making, anomaly detection, and optimization procedures in intricate energy systems by using vast amounts of structured and unstructured data. This study provides a comprehensive review of the LLM origins, evaluation, and fine-tuning techniques as well as their integration into energy systems, including their application in fault detection and diagnosis, energy forecasting, document automation, energy management, defect detection, and power system operation. Their performance in terms of explainability, generalization ability, and scalability for energy-related applications is critically examined in this paper. The report also emphasizes significant challenges to the adoption of LLMs, such as the need for computing power, the lack of data, and ethical issues like bias and false information. Power-efficient models, hybrid artificial intelligence (AI) platforms, and domain-specific fine-tuning are some of the solutions discussed. Future areas of interest include multi-modality to obtain maximal forecasting and operational intelligence, real-time adaptability, and explainable. This paper summarizes current developments and provides information on LLM-driven innovation in energy systems while maintaining transparency and dependability. Compared with prior LLM–energy surveys that either remain general-purpose or focus on a single subdomain, this review fills three concrete gaps: (i) a cross-domain synthesis of energy-specific LLM applications spanning power systems, buildings, and forecasting; (ii) a methods-oriented consolidation of evaluation and parameter-efficient fine-tuning practices tailored to energy tasks; and (iii) a deployment-centric analysis of real-time and edge constraints (energy, latency, hardware) with a practical reporting checklist for operational adoption.

Faults in distribution networks can result in severe transients, equipment failure, and power outages. The quick and accurate detection of the faulty section enables the operator to avoid prolonged power outages and economic losses by quickly retrieving the network. However, the occurrence of diverse fault types with various resistances and locations and the highly non-linear nature of distribution networks make fault section detection challenging for numerous conventional techniques. This study presents a cutting-edge deep learning-based algorithm to distinguish fault sections in distribution networks to address these issues. The proposed gated recurrent unit model utilizes only two samples of the angle between the voltage and current on either side of the feeders, which record by smart feeder meters, to detect faulty sections in real time. When a network fault occurs, the protection relays trigger the trip command for the breakers. Immediately, the angle data are obtained from all smart feeder meters of the network, which comprises a pre-fault sample and a post-fault sample. The data are then employed as an input to the pre-trained gated recurrent unit model to determine the faulted line. The performance of this novel algorithm was validated through simulations of various fault types in the IEEE-33 bus system. The model recognizes the faulty section with competitive performance in terms of accuracy.

Aging underground cables pose a threatening issue in distribution systems. Replacing all cables at once is economically unfeasible, making it crucial to prioritize replacements. Traditionally, age-based strategies have been used, but they are likely to fail to depict the real condition of cables. Insulation faults are influenced by electrical, mechanical, thermal, and chemical stresses, and partial discharges (PDs) often serve as early indicators and accelerators of insulation aging. The trends in PD activity provide valuable information about insulation condition, although they do not directly reveal the cable’s real age. Due to the absence of an established ranking methodology for such condition-based data, this paper proposes a fuzzy logic and fuzzy analytic hierarchy process (FAHP)-based cable vulnerability ranking framework that effectively manages uncertainty and expert-based conditions. The proposed framework requires only basic and readily accessible data inputs, specifically cable age, which utilities commonly maintain, and PD measurements, such as peak values and event counts, which can be acquired through cost-effective, noninvasive sensing methods. To systematically evaluate the method’s performance and robustness, particularly given the inherent uncertainties in cable age and PD characteristics, this study employs Monte Carlo simulations coupled with a Spearman correlation analysis. The effectiveness of the developed framework is demonstrated using real operational cable data from a Danish distribution network, meteorological information from the Danish Meteorological Institute (DMI), and synthetically generated PD data. The results confirm that the FAHP-based ranking approach delivers robust and consistent outcomes under uncertainty, thereby supporting utilities in making more informed and economical maintenance decisions.

The increasing failure rate of underground cable systems due to aging poses challenges for the reliable operation of power grids. Without proper asset management methods, the grid may not adequately support the ongoing climate-centered electrification of society. This highlights the need for strategic asset maintenance decision-making. While the existing literature includes inventive approaches for reliability modeling and maintenance strategy selection, there is still a gap in integrating detailed predictive analytics-based reliability modeling that accounts for common data deficiencies, alongside maintenance strategies that consider both reliability and cost-based performance indicators. In this paper, we present a comprehensive proactive cable replacement method using the Neural Weibull Proportional Hazard model, specifically designed to handle truncation and censoring mechanisms. The method employs a Bayesian nested Monte Carlo simulation to address the dual uncertainties associated with the probabilistic reliability model. We apply this maintenance simulation method to a case study of a real distribution system in Denmark. Our results indicate that the optimal replacement strategy reduces outage costs by up to 6.16 million DKK and can decrease the System Average Interruption Duration Index (SAIDI) by more than 50% compared to alternative replacement strategies. Finally, a sensitivity analysis of the maintenance budget shows that operators can reduce their annual budget for cable replacements by up to 50%, incurring only minor additional costs and an increase in peak SAIDI of just 1.5 minutes. The innovation lies in combining the NWPH model with Bayesian Monte Carlo simulation to handle incomplete records, quantify uncertainty, and guide cost-effective maintenance decisions.