Whether due to climate change, cyber attacks, or physical attacks, outages in the electrical grid can cause economic and physical harm. In this work we investigate how Distributed Energy Resources (DERs) could be leveraged to enhance resilience by using demand response and energy storage to mitigate line outage events. In order to provide rapid decision support, in this work continuous off-policy Deep Reinforcement Learning (DRL) is applied on a portfolio of DERs after an extreme event has impacted. With the help of distributed training, hyperparameter optimisation, and a safety shield the resulting agents were able to reduce Energy Not Supplied (ENS) by up to 52.9%, on average, in unseen scenarios. By varying the composition of the DER portfolio, a weak linear relationship between portfolio size and scenario performance was found, alongside a strong impact of where demand response was applied. Nevertheless, based on this sensitivity analysis, a DSO could temporarily contract DERs based on the expected reduction in the cost of ENS, when compared to the cost of acquiring the DER.

With the expansion of machine learning, many failure detection algorithms have been developed. However, existing models are not applicable in the field of power systems, because practical limitations of grid operators are not considered. For instance, part of the models rely on component-specific sensor data. Owing to the size of power systems among others, it is unrealistic to install sensors on all grid components. We investigate the possibility to predict the failure of a grid component without using component-specific sensors. To this goal, an automatization algorithm is developed, that selects pre-processing methods of input data and optimal hyperparameters of machine learning models. It leverages commonly available operational data to generate warnings when there is a risk of failure of a specific grid component. Autoencoders provide satisfactory results when suitable choices are made concerning the performance criteria, and the data pre-processing, including the number of lags and sliding window steps and the use of binary variables. Finally, we discuss ways to improve the algorithm, notably methods for data augmentation, cross validation, and strategies to cope with the trade-off between false alarms and failure detection.

Artificial intelligence has become a prominent approach for predictive maintenance in power systems, driven by the need to anticipate failures and improve grid reliability. However, advanced models are often adopted without systematically assessing whether data quality, maintenance objectives, and operational constraints make them feasible or advantageous. In high-stakes power-system applications, misaligned methodological choices can produce solutions that are difficult to deploy, inflate costs, and delay actionable results. This motivates the need for clear, data-informed criteria for selecting suitable predictive-maintenance methods among complex artificial intelligence, conventional data-driven, and non-data-driven approaches. This paper presents GUIDE–PdM, a five-phase decision-support framework that aligns modeling strategies with data readiness and asset-management objectives. GUIDE–PdM supports practitioners through problem definition, data characterization, method selection, model development, and adaptive deployment, in a structured process where the use of complex artificial intelligence is justified by evidence rather than assumed. To complement the framework, we provide a data–method mapping table that links data modalities, constraints, and application requirements to methods reported in the literature. The framework is formalized as pseudo-code alongside the flowchart, with a fully worked numeric instantiation for one case in Appendix A. Its applicability is shown through four case studies in grid operation and asset management, illustrating how data readiness, system complexity, and operational constraints jointly shape method selection. While the case studies focus on power-system assets, the same approach should transfer to other industrial domains with similar constraints and a comparable need to justify method selection.

As Europe accelerates toward a more interconnected and decarbonized power system, cross-border electricity exchange and HVDC transmission have become essential for grid stability, renewable integration, and market coupling. This paper presents a Markov-based reliability assessment of Point-To-Point (P2P) and Multi-Terminal (MT) HVDC topologies, focusing on how network configuration impacts overall system reliability. Using continuous-Time Markov processes, the reliability of Voltage Source Converter (VSC) stations and DC transmission lines is evaluated, with a detailed analysis of subsystem vulnerabilities, particularly the critical AC↔DC conversion unit. Results show that partial capacity states are rare, with failures typically leading to complete link shutdowns. The study compares the availability of P2P monopolar and parallel-monopolar configurations with MT ring and meshed topologies, revealing that MT meshed systems offer superior fault tolerance, despite their complexity. The findings highlight the importance of preventive maintenance, configuration choices, and multi-vendor interoperability in enhancing HVDC reliability, providing a foundation for guiding future investments in reliable, resilient, and sustainable power grids.

Despite ongoing efforts to decarbonize society, climate change continues to result in more extreme events that can reduce the electrical grid’s ability to reliably supply power to end users. At the same time, across the distribution grid, Distributed Energy Resources (DERs), such as renewable generation, energy storage systems, and flexible resources, offer the possibility of operating the grid in novel ways. Distribution System Operator (DSOs) could employ these DERs to improve their ability to mitigate extreme events. This work therefore demonstrates how DERs, in the form of flexible loads, can be controlled by a Deep Reinforcement Learning (DRL) agent to minimize Energy Not Supplied (ENS) in the immediate aftermath of an extreme event. To obtain near-optimal performance on unseen scenarios, an enhanced Implicit Q Network (IQN+) architecture is proposed, trained, and evaluated on a modified CIGRE MV benchmark grid. The resulting IQN+ agent can outperform a passive baseline policy, a trained Rainbow DQN policy, and a single-timestep Optimal Power Flow (OPF) based policy on the test set. Sensitivity analysis reveals that the location and quantity of available DERs also impact the efficacy of the IQN+ agent, with the agent preferentially using actions on loads that result in greater average episode durations. These results highlight the potential for RL to rapidly provide decision support to operators by suggesting potential remedial actions to mitigate the impact of extreme events.

The growing adoption of electric vehicles (EVs) creates opportunities to support power system reliability by using EV batteries as decentralized storage resources. This paper proposes an optimization framework for EV aggregators participating in ancillary service markets through unidirectional smart charging (V1G) and vehicle-to-grid (V2G) operation, explicitly accounting for battery lifetime degradation. Eleven scenarios are evaluated in the context of the Swedish electricity market to examine different market participation strategies, degradation modeling approaches, and degradation compensation schemes. Two degradation models are considered: a simplified empirical model and a detailed lithium-ion degradation model based on solid electrolyte interphase (SEI) formation — a passivation layer that forms on the negative electrode during battery operation and contributes to capacity fade and internal resistance increase. Unlike most prior studies that focus only on capacity loss, the proposed framework captures both capacity fade and power capability fade, enabling more realistic scheduling and cost estimation. Results demonstrate that controlled EV charging and participation in ancillary services significantly improve economic outcomes compared to uncontrolled charging. In a case study with 55 EVs, the EV aggregator achieves a daily net revenue of up to €100, even for a relatively small-scale system. At the same time, EV owners receive charging cost reductions of up to 40% compared to the baseline uncontrolled charging scenario. Scenarios with explicit degradation compensation achieve fairer cost allocation at minimal profit reduction for the EV aggregator. A sensitivity analysis further demonstrates that the main economic conclusions remain robust across a wide range of degradation cost assumptions. Overall, the study confirms that battery degradation-aware V1G/V2G strategies can deliver significant economic benefits while respecting operational constraints.

To cater to the advancement of electric and hydrogen-powered aircraft, airports are increasingly motivated to transition to green multi-energy airport microgrids (MEAM) for efficient operation and integration of stochastic renewable energy sources (RES). This paper presents a temporally coordinated (two-stage) stochastic programming (SP) model to minimize the energy supply cost of MEAMs while enabling efficient operations with electricity, green hydrogen and thermal energy. From the MEAM perspective, first, the multi-energy loads of MEAMs are considered flexible and modeled in details; second, the electricity-to-hydrogen-and-heat (E2HH) model is applied considering the influence of climatic conditions on the electrolyzers' efficiencies. With regard to the SP model, for one thing, the copula method is employed to capture correlations between climatic parameters related to RES generation and loads to enhance the accuracy and fidelity of the generated scenarios; for another, the Jensen wake model is applied to improve the estimation accuracy of available wind power in the generated scenarios. Furthermore, an adapted-penalty Progressive Hedging (PH) model is proposed to decompose the SP model, reducing the computational burden. Finally, case studies indicate that the proposed approach can effectively coordinate the operation of MEAM to balance system security and cost minimization.

The operational challenges of the integration of electric vehicles (EV), air conditioning and photovoltaic panels (PV) are prompting the upgrade of distribution grids, seen here as cyber-physical infrastructures. An important upgrading feature of the cyber-side is the electrical grid monitoring, which needs to expand both in data coverage and granularity. The challenge is to decide the data strategy, or in other words, which level of granularity is actually needed in active distribution networks. This work proposes a framework to assist grid planners in selecting the level of data expansion needed, by quantifying the impact of extended data granularity on control capabilities, and corresponding grid performance. The framework combines machine learning with AC optimal power flow and state estimation to select incremental upgrades of the cyber-physical infrastructure. Grid planning and operation are simulated and tested for the IEEE 33-bus test system over a 5-year span to assess the role of granularity in grid performance for different cyber-infrastructures. The results show that extending data granularity is critical for mitigating voltage violations under high penetration of EVs, air conditioning and PVs. By modeling the relationships between data, grid planning and operation, and grid performance, the framework supports efficient cyber system upgrades to mitigate operational violations while accounting for budget limitations.

High Voltage Circuit Breakers (HVCBs) are critical components in power systems to maintain reliable operation. Accurate condition monitoring of HVCBs is vital to reduce maintenance costs and consequently to enhance grid reliability. However, achieving this with low-cost measurement devices, which often provide noisy signals, poses a significant challenge. In this paper, a novel defect classification framework for HVCBs is proposed that uses the Savitzky-Golay filter to preprocess the most common condition monitoring signal, which is the trip/close coil current. This filter is well-known for denoising while preserving critical signal features. Following signal preprocessing, a robust defect detection and classification methodology is introduced, combining time series similarity assessment techniques, such as Euclidean distance and dynamic time warping, with machine learning algorithms. Moreover, an experimental setup is designed to emulate the behavior of an HVCB's coil mechanism. To further enhance model transparency, Shapley additive explanations analysis is applied, providing interpretability into feature contributions toward model decisions. The obtained results validate the effectiveness of the proposed hybrid approach, demonstrating its potential to provide a cost-effective, accurate, and reliable solution for HVCB condition monitoring.

Power system communication networks enable operators to remotely monitor and control field equipment. The sophistication of these networks is also increasing as operators continue the trend towards digitization, which is beneficial in integrating distributed energy resources. However, as the attack surface increases in size so too does the risk of cyberattacks. The topology, configuration and composition of communication net-works is therefore confidential since this can provide information to attackers. As a result, the number of benchmarks available for research purposes is limited.A tool for procedurally generating communication network topologies is therefore proposed. While primarily intended as an enabler for public research into communication networks, this tool also allows general insights to be gained into the effect of communication network design on the vulnerability of networks to cyberattacks. The tool includes the ability to encapsulate network characteristics in JSON specification files, which is demonstrated with example Advanced Metering Infrastructure (AMI) and Supervisory Control and Data Acquisition (SCADA) networks. The SCADA network generation is then compared to a real-world case. Finally, the effect of network redundancy on the networks' cyber resilience is investigated.