Centralized reinforcement learning-based voltage regulation in distribution networks is becoming increasingly difficult due to the growing penetration of distributed energy resources, high computational burden, repeated power flow calculations, and increasing privacy concerns. This paper proposes a physics-informed fully distributed reinforcement learning framework that enables autonomous voltage regulation using only local smart meter data. A Thevenin-equivalent-based local voltage estimation model and a hybrid correction mechanism are developed to support accurate local decision-making without synchronized global measurements or centralized power flow solvers. A lightweight coordination mechanism is further introduced to refine the actions of independently trained local agents. Case studies show that the proposed framework reduces voltage violations by approximately 80%, achieves performance close to that of power flow-based training environments, and achieves a training speedup of about 6×[jls-end-space/]. The results also indicate that the relaxation factors in the reward function and the coordination scaler are critical to voltage regulation efficiency, whereas the discount factor has a smaller impact. These findings demonstrate the practicality of the proposed framework for privacy-aware fully distributed voltage regulation.

With the growing integration of Modular Multilevel Converters (MMCs) in Multi-Terminal Direct Current (MTDC) transmission systems, there is a growing need for control strategies that balance economic efficiency with robust dynamic performance. This paper presents an enhanced Optimal Power Flow (OPF)-based framework for hybrid AC-MTDC systems, incorporating a novel droop control strategy that jointly coordinates DC-voltage and AC-frequency regulation. By embedding frequency control loops into the MMCs, the method enables system-wide coordination that enhances power sharing and improves resilience under disturbances. The proposed strategy dynamically adjusts converter operating points to minimize generation costs and DC-voltage deviations, balancing economic objectives with system stability. A modified Nordic test system integrated with a four-terminal MTDC grid is used to validate the approach. Optimization is performed using Julia, while the system's dynamic performance is evaluated through electromagnetic transient simulations with the EMTP software. Case studies across multiple scenarios demonstrate that the proposed droop control achieves markedly improved frequency and voltage robustness over active power control, while incurring lower generation costs than the adaptive droop benchmark. The results highlight the ability of the proposed strategy to deliver cost-effective operation without compromising performance, offering a promising solution for the coordinated control of future hybrid AC-DC transmission networks.

Rapidly expanding Electric Vehicle (EV) adoption necessitates robust, large-scale charging strategies to meet global decarbonization targets. Traditional methods, such as heuristics and mathematical programming-based approaches, struggle to scale effectively and adapt to EV dispatch’s complexity, uncertainty, and variability. Reinforcement Learning (RL) offers a promising alternative due to its ability to handle complex optimization problems, process substantial real-time data, and learn continuously without explicit retraining for every scenario. This study proposes a novel end-to-end RL framework that leverages the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, enhanced by Graph Neural Networks (GNN), to capture spatial interactions among charging nodes. A key contribution is integrating a safety layer designed to ensure grid stability, preserve EV charging requirements, and enforce power limits. The RL agent was trained and evaluated using real EV charging sessions, offering a realistic assessment of its performance. The results indicate that the proposed method can efficiently coordinate large fleets of EVs, ensuring stable power grid operation and fair distribution of charging resources.

Electric-vehicle smart charging requires quick decision-making under uncertainty while enforcing strict electricity grid and user requirements. Mathematical optimization becomes too slow at scale, while online reinforcement learning struggles with sparse rewards and safety. This paper proposes GNN-DT, a topology-aware Decision Transformer that combines graph neural network embeddings with sequence modeling to learn charging policies from offline trajectories. The method operates over variable numbers of vehicles and chargers without retraining. Evaluated on realistic smart charging scenarios, GNN-DT achieves near-optimal performance, reaching rewards within 5 percent of an oracle solver while using up to 10× fewer training trajectories than baseline methods. It consistently outperforms online and offline reinforcement learning approaches and generalizes to unseen fleet sizes and network topologies. Inference runs in milliseconds, making the approach suitable for real-time deployment in large-scale charging systems.

To develop an optimal operational scheme for distribution networks capable of addressing asymmetric uncertainties associated with renewable energy and load demands, this paper presents a confidence level-based information gap decision theory (CL-IGDT) framework. Building on IGDT, the proposed framework utilizes the confidence level to capture the asymmetric characteristics of uncertainties and maximize the risk-averse capability of the solution in a probabilistic manner. To facilitate such probabilistic consideration, the imprecise Dirichlet model is employed to construct the ambiguity sets of uncertainties. Consequently, a two-stage robust optimal operation model for distribution networks using CL-IGDT is developed. An iterative method is proposed to solve the model and determine the upper and lower bounds of the objective function. Case study demonstrates that the proposed approach yields a more robust and statistically optimized solution with required accuracy compared to existing methods, contributing to a reduction in first-stage cost by 0.84%, second-stage average cost by 6.7%, and significantly increasing the reliability of the solution by 8%.

Integrating electric vehicles (EVs) into the power grid can revolutionize energy management strategies, offering both challenges and opportunities for creating a more sustainable and resilient grid. In this context, model predictive control (MPC) emerges as a powerful tool for addressing the complexities of Grid-to-Vehicle (G2V) and Vehicle-to-Grid (V2G) enabled demand response management. By leveraging advanced optimization techniques, MPC algorithms can anticipate future grid conditions and dynamically adjust EV charging and discharging schedules. However, no standard tools exist to evaluate novel energy management strategies based on MPC approaches. This work focuses on harnessing the potential of MPC in G2V and V2G applications by providing open-source algorithms that allow the maximization of EV flexibility and support demand response initiatives while mitigating the impact on EV battery health. Through extensive simulation and analysis, we demonstrate the efficacy of our approach in maximizing the benefits of G2V and V2G while assessing the impact on the longevity and reliability of EV batteries. Specifically, the proposed methods enable the optimization of EV charging and discharging schedules in real-time, taking into account fluctuating energy prices, grid constraints, and EV user preferences.

We introduce a physics-informed neural network for power flow (PINN4PF) that effectively captures the nonlinear dynamics of large-scale modern power systems. The proposed neural network (NN) architecture consists of two important advancements in the training pipeline: (A) a double-head feed-forward NN that aligns with power flow (PF), including an activation function that adjusts to the net active and reactive power injections patterns, and (B) a physics-based loss function that partially incorporates power system topology information through a novel hidden function. The effectiveness of the proposed architecture is illustrated through 4-bus, 15-bus, 290-bus, and 2224-bus test systems and is evaluated against two baselines: a linear regression model (LR) and a black-box NN (MLP). The comparison is based on (i) generalization ability, (ii) robustness, (iii) impact of training dataset size on generalization ability, (iv) accuracy in approximating derived PF quantities (specifically line current, line active power, and line reactive power), and (v) scalability. Results demonstrate that PINN4PF outperforms both baselines across all test systems by up to two orders of magnitude, not only in terms of direct criteria, e.g., generalization ability, but also in terms of derived physical quantities.

Electricity Consumption Profiles (ECPs) are crucial for operating and planning power distribution systems, especially with the increasing number of low-carbon technologies such as solar panels and electric vehicles. Traditional ECP modeling methods typically assume the availability of sufficient ECP data. However, in practice, the accessibility of ECP data is limited due to privacy issues or the absence of metering devices. Few-shot learning (FSL) has emerged as a promising solution for ECP modeling in data-scarce scenarios. Nevertheless, standard FSL methods, such as those used for images, are unsuitable for ECP modeling because (1) these methods usually assume several source domains with sufficient data and several target domains. However, in the context of ECP modeling, there may be thousands of source domains, e.g., households with a moderate amount of data, and thousands of target domains, e.g., households that ECP are required to be modeled. (2) Standard FSL methods usually involve cumbersome knowledge transfer mechanisms, such as pre-training and fine-tuning. To address these limitations, this paper proposes a novel FSL framework that integrates Transformers with Gaussian Mixture Models (GMMs) for ECP modeling. The proposed approach is fine-tuning-free, computationally efficient, and robust even with extremely limited data. Results show that our method can accurately restore the complex ECP distribution with a minimal amount of ECP data (e.g., only 1.6% of the complete domain dataset) and outperforms state-of-the-art time series modeling methods in the context of ECP modeling.

Digital substations, which replace traditional analog infrastructure, are essential to power grid operation but are facing growing vulnerability to cyber attacks. Existing anomaly detection in substation communication requires labeled datasets for supervised training and fails to incorporate temporal characteristics, which cannot detect unknown persistent attacks. Setting arbitrary thresholds for outlier detection leads to high false positives and low detection rates. This paper addresses cyber security challenges related to IEC 61850 Generic Object Oriented Substation Event (GOOSE) protocol within digital substations. We propose a novel unsupervised Transformer-based Distribution Fitting Anomaly Detection (TF-DiFAD) method for time series GOOSE frames with a robust thresholding technique. Deep packet inspection is used to extract features from GOOSE frames, which are processed through the proposed TF-DiFAD model. TF-DiFAD combines the deep learning transformer model with statistical distribution fitting techniques to accurately detect anomalous GOOSE frames. Specifically, reconstruction errors are generated using a state-of-the-art transformer model. A novel model-agnostic solution is applied for setting anomaly thresholds and calculating anomaly probabilities. The Kolmogorov-Smirnov test is employed to select the best-fitting distribution for these errors. TF-DiFAD is benchmarked against other state-of-the-art models using two distinct test datasets, demonstrating superior performance. The results indicate that TF-DiFAD detects anomalies with Receiver Operating Characteristics Area Under Curve (ROC AUC) scores of 96.84% and 95.73% respectively for both datasets.

The integration of distributed energy resources (DERs) has escalated the challenge of voltage magnitude regulation in distribution networks. Model-based approaches, which rely on complex sequential mathematical formulations, cannot meet the real-time demand. Deep reinforcement learning (DRL) offers an alternative by utilizing offline training with distribution network simulators and then executing online without computation. However, DRL algorithms fail to enforce voltage magnitude constraints during training and testing, potentially leading to serious operational violations. To tackle these challenges, we introduce a novel safe-guaranteed reinforcement learning algorithm, the DistFlow safe reinforcement learning (DF-SRL), designed specifically for real-time voltage magnitude regulation in distribution networks. The DF-SRL algorithm incorporates a DistFlow linearization to construct an expert-knowledge-based safety layer. Subsequently, the DF-SRL algorithm overlays this safety layer on top of the agent policy, recalibrating unsafe actions to safe domains through a quadratic programming formulation. Simulation results show the DF-SRL algorithm consistently ensures voltage magnitude constraints during training and real-time operation (test) phases, achieving faster convergence and higher performance, which differentiates it apart from (safe) DRL benchmark algorithms.

Rising energy expenses, the shift towards renewable sources, and grid congestion considerably affect the operations of container terminals. To tackle these challenges, it is necessary to implement energy-aware integrated operational planning which considers related uncertainties. This work proposes a two-stage stochastic mixed integer programming model to optimize container terminal operations planning and demand-responsive energy management. To this end, energy consumption is shifted whenever operationally possible and economically beneficial. We solve the proposed model by developing a dedicated progressive hedging algorithm. Operations considered in this model include vessel scheduling at berths, temperature control of refrigerated containers, and allocation of handling capacity of quay cranes, yard cranes, and automated guided vehicles to serve each vessel. Various scenarios for vessel arrival times and electricity prices are explored representing the uncertainty of energy demand and supply, respectively, based on a case study of the Altenwerder container terminal in Hamburg. Our results suggest potential cost savings of 5.9 per cent on average with a single energy price based on a long-term contract and 13.2 per cent when applying varying real-time electricity prices based on wholesale market rates. These findings underscore the substantial potential of demand response strategies for (electrified) container terminal operations.

As the adoption of electric vehicles (EVs) accelerates, addressing the challenges of large-scale, city-wide optimization becomes critical in ensuring efficient use of charging infrastructure and maintaining electrical grid stability. This study introduces EV-GNN, a novel graph-based solution that addresses scalability challenges and captures uncertainties in EV behavior from a Charging Point Operator’s (CPO) perspective. We prove that EV-GNN enhances classic Reinforcement Learning (RL) algorithms’ scalability and sample efficiency by combining an end-to-end Graph Neural Network (GNN) architecture with RL and employing a branch pruning technique. We further demonstrate that the proposed architecture’s flexibility allows it to be combined with most state-of-the-art deep RL algorithms to solve a wide range of problems, including those with continuous, multi-discrete, and discrete action spaces. Extensive experimental evaluations show that EV-GNN significantly outperforms state-of-the-art RL algorithms in scalability and generalization across diverse EV charging scenarios, delivering notable improvements in both small- and large-scale problems.

This study presents an innovative approach to risk-aware decision-making in water resource management. We focus on a case study in the Netherlands, where risk awareness is key to water system design and policy-making. Recognizing the limitations of deterministic methods in the face of weather, energy system, and market uncertainties, we propose a scalable stochastic Model Predictive Control (MPC) framework that integrates probabilistic forecasting, scenario generation, and stochastic optimal control. We utilize Combined Quantile Regression Deep Neural Networks and Non-parametric Bayesian Networks to generate probabilistic scenarios that capture realistic temporal dependencies. The energy distance metric is applied to optimize scenario selection and generate scenario trees, ensuring computational feasibility without compromising decision quality. A key feature of our approach is the introduction of Exceedance Risk (ER) constraints, inspired by Conditional-Value-at-Risk (CVaR), to enable more nuanced and risk-aware decision-making while maintaining computational efficiency. In this work, we enable the Noordzeekanaal–Amsterdam-Rijnkanaal (NZK-ARK) system to participate in Demand Response (DR) services by dynamically scheduling pumps to align with low hourly electricity prices on the Day Ahead and Intraday markets. Through historical simulations using real water system and electricity price data, we demonstrate that incorporating uncertainty can significantly reduce operational costs—by up to 44 percentage points compared to a deterministic approach—while maintaining safe water levels. The modular nature of the framework also makes it adaptable to a wide range of applications, including hydropower and battery storage systems.

Distribution system operators (DSOs) often lack high-quality data on low-voltage distribution networks (LVDNs), including the topology and the phase connection of residential customers. The phase connection is essential for phase balancing assessment and distributed energy resources (DERs) integration. The existing load profiles-based approaches rely on stepwise subtraction of the identified customers in a step-by-step identification procedure, while the accuracy of each step is not guaranteed. This paper introduces a siamese neural network model to identify single-phase connections without requiring stepwise subtraction. It comprises self-taught learning (STT) and a phase-label identification strategy. The introduced self-taught learning enables DSOs to train a recurrent neural network-based Siamese network (RSN) only relying on an unlabelled dataset. Besides, the siamese network (SN) is robust to noise and fluctuations in the data to a certain extent, making the proposed method robust to measurement errors. A Kendall correlation-based phase modification strategy is introduced to modified phase labels with lower confidence, aiming to mitigate the accuracy loss induced by the limited generalization of SN. The proposed approach is tested on the IEEE European low voltage test feeder and a residential network in the Netherlands Simulation results illustrate the feasibility and robustness of the proposed approach on incomplete datasets. The accuracy exceeded 83% and 90%, respectively, when using datasets of less than 20 days with and without measurement errors.

Industrial electrification plays a crucial role in reducing carbon dioxide emissions, and ensuring power reliability is important in this process. Reliability and techno-economic evaluations are fundamental to designing, operating, and managing power systems, ensuring that electricity is delivered continuously and securely under various conditions. In particular, maintaining a reliable power supply to industrial loads is critical, especially when renewable sources are present, as these introduce greater variability and uncertainty into the operation of industrial systems. Therefore, this document aims to use a cost-effective storage approach to ensure the reliable operation of sustainable industrial multi-energy systems. In addition, three storage mitigation strategies against random operation are formulated based on financial, technical, practical, and other aspects. A synthetic industrial model consisting of generic component representations in DIgSILENT PowerFactory 2024 is taken as a case study. The structure and parameters of the synthetic model are inspired by data from the literature and a hypothetical projection of a future evolution of a 500 MW sustainable industrial multi-energy system in Rotterdam by 2035. Numerical results provide insight into the flexible and cost-effective operation of sustainable industrial multi-energy systems within the context of decarbonised future Dutch energy systems.