ABSTRACT. Increased penetration of wind and solar PV plants and batteries means that power system characteristics are no longer determined by the physics of synchronous generators operating in traditional plants but can be (almost) freely shaped by power electronics, i.e. programmable inverters. This represents the biggest change to power system operation and control since AC power systems were first developed at the end of the 19th century. The presentation will discuss the differences between traditional and power electronics-based power systems, the challenges for power system modelling and analysis, and limits to what extent power electronics can replace synchronous machines. It will conclude with a discussion on frequency control and system stability. Relevance for the Iberian blackout on 28 April 2025 will also be discussed.

ABSTRACT. Accurate probabilistic load forecasting (PLF) is essential for maintaining grid security in the face of increasing volatility driven by renewable energy sources. However, many spatio-temporal models either suffer from high computational demands or produce poorly calibrated uncertainty estimates at the nodal level. This paper proposes a lightweight hybrid framework that models load residuals relative to operational day-ahead forecasts. A frozen, reservoir-style spatio-temporal graph neural network (STGNN) is used to extract topological embeddings, which then serve as contextual inputs for an XGBoost-based quantile regressor. The resulting predictions are further calibrated using Mondrian conformalized quantile regression (M-CQR) to ensure reliable coverage across node–horizon segments. The framework is evaluated on a 12-country ENTSO-E system and demonstrates superior performance compared to competing baselines at the critical 1-hour and 24-hour horizons, while remaining competitive at intermediate horizons. Ablation studies indicate that the frozen encoder matches or exceeds the accuracy of fully trained alternatives, while reducing training time by approximately 82×. In addition, the M-CQR calibration enables the detection of localized risks that may not be visible through aggregated calibration metrics. Overall, the approach produces sharp and well-calibrated prediction intervals, offering a computationally efficient and reliable solution for risk-aware decision support in real-world grid operations.

ABSTRACT. Short-term load forecasting is essential for the reliable operation of small, climate-sensitive power systems such as Spanish insular grids. Traditional approaches rely mainly on temperature; however, these systems exhibit higher stochasticity and stronger meteorological dependence, suggesting that additional weather variables may improve accuracy.

This work evaluates the incorporation of solar radiation, cloud cover, and wind speed into the Spanish operational forecasting framework using alternative seasonal representations. Results based on real operational data show that including these variables significantly improves forecasting accuracy, particularly during daylight hours. The proposed model is currently used by the Spanish Transmission System Operator (\textit{Red Eléctrica de España}, REE) for short-term forecasting in insular systems.

ABSTRACT. The increasing deployment of photovoltaic (PV) systems in modern power grids has intensified the need to understand and quantify their production variability at both individual and aggregate levels. Variability arising from rapid irradiance fluctuations, cloud dynamics, and spatial dispersion significantly impacts grid stability, forecasting accuracy, and operational planning. This study investigates the temporal and spatial characteristics of PV output variability thanks to a high-quality synchronous dataset recorded at two PV plants located within the same distribution grid. Recorded data are analysed with a ramp-based variability approach using the Swinging Door Algorithm (SDA), which detects and quantifies fluctuations in PV data. This paper presents a systematic comparison of ramp frequency, magnitude, and ramp rate across different aggregation configurations. The preliminary results confirm what it was previously studied with simulations, that aggregation helps to soft variabilities.

ABSTRACT. The increasing integration of photovoltaic (PV) power and its inherent variability require reliable aggregated probabilistic forecasts to support grid operation and market participation. However, most existing approaches focus on deterministic or site-level forecasting, with limited attention to coherent system-level uncertainty that captures spatial and temporal dependencies. This paper proposes a multistage framework for day-ahead probabilistic forecasting of aggregated PV power, combining physics-guided modeling and copula-based dependency construction. Power generation of 175 PV systems in Utrecht, the Netherlands, is clustered by physical similarity, and a physical model chain converts numerical weather predictions into power estimates. Cluster-level residuals are modeled probabilistically to generate quantile forecasts, which are aggregated using Gaussian Copulas to represent spatial dependence and enforce temporal coherence in day-ahead trajectories. Results indicate that physics-guided quantile model improves probabilistic accuracy compared to the pure quantile approach. The proposed framework provides reliable and coherent day-ahead scenarios, reduces extreme ramp behavior and physical boundary violations, while providing a scalable and operationally realistic methodology for aggregated PV probabilistic forecasting in hierarchical settings.

ABSTRACT. Extreme weather events, such as heat waves and cold spells, are difficult to predict accurately, and this can amplify errors in numerical weather prediction (NWP). In this context, the systematic biases of the Global Forecast System (GFS) in irradiance and temperature propagate nonlinearly through the thermal derating of photovoltaic modules, compromising solar generation estimates. To address this problem, we propose a machine learning post-processing framework that incorporates explicit extreme-event indicators, evaluating five algorithms (Gradient Boosting, Random Forest, Ridge, XGBoost, and LightGBM) using strict temporal cross-validation with five splits. Random Forest and LightGBM consistently outperformed linear models, reducing irradiance MAE to 51.98 W/m² and temperature MAE to 1.01 °C, while Ridge showed the poorest performance due to its inability to capture nonlinear relationships. The meteorological bias correction reduced the overall AC power MAE from 36.2 W to 20.8 W, increased the coefficient of determination from 0.633 to 0.869, and brought the daily energy error below 1.1% across all weather regimes.

ABSTRACT. Differences in data, expertise, and computational resources lead to heterogeneous forecasting capabilities across energy-system agents. This paper proposes a market-based framework for trading access to forecasting models, in which sellers provide controlled inference access to trained models and buyers acquire access only when it is both technically beneficial and economically feasible. A central coordinator evaluates candidate models using buyer-specific validation data and selects among feasible offers according to criteria that balance predictive performance and cost. The framework is evaluated using synthetic data, Norwegian wind power data, and Portuguese electricity consumption data. Results show that economically incentivized model sharing can reduce forecasting error for selected buyers and that different selection criteria lead to different trade-offs between predictive performance and transaction price.

ABSTRACT. The rapid deployment of Battery Energy Storage Systems (BESS) has led to increasing competition in ancillary service markets, progressively reducing available revenues and challenging the economic viability of short-duration storage assets. This paper investigates how market saturation influences optimal BESS sizing and duration, with a particular focus on the transition toward longer duration energy storages. This work couples a threephase stochastic price engine for the German day-ahead (DA) and FCR markets with a degradation-aware rolling-horizon Mixed-Integer Linear Programming dispatch model, producing ≈2,500 internally consistent (DA, FCR) paths across 2025, 2030, and 2037 scenarios from the German grid development plan. Results show that FCR cannibalization collapses ancillary revenues toward marginal floor levels by 2030, while widening DA spreads reorient the revenue stack toward energy arbitrage. The economically optimal storage duration migrates from 1–2 h to longer duration systems by 2030. Despite a near-term profitability dip, internal rates of return consistently exceed 10 % even in unfavorable scenarios. A closed-form spread–FCR proxy reproduces MILP revenues with R² = 0.978, enabling rapid investment screening without re-optimization.

ABSTRACT. Successful Battery Energy Storage System projects should accurately account for the optimal balance between immediate usage intensity and future revenues opportunities. This study addresses this aspect by introducing a simple yet effective approach that provides as output signal a monthly endogenous degradation cost profile for a Battery arbitraging in the energy Day-Ahead Market while accounting for the associated cyclic degradation. Notably, this methodology can be easily integrated and updated into the battery Energy Management Systems dispatch algorithms, allowing the asset to make informed-based decisions while accounting for the system’s dynamics. The potential techno-economic improvements arising from the proposed method are proved against a standard approach, where a maximum number of daily cycles is imposed. Simulation results indicate an increase in the final revenues and an extension of the asset’s operating lifetime.

ABSTRACT. Lithium Iron Phosphate (LFP) Battery Energy Storage Systems (BESSs) are a key enabler of the energy transition. However, they are known to exhibit significant inaccuracies in the estimation of their State of Charge (SOC). Such estimation errors can directly impact the participation of BESSs in electricity markets. In this work, we demonstrate that neglecting SOC uncertainty in battery bidding can lead to significant delivery failures, including the inability to meet promised frequency reserves. To address this risk, we investigate bidding strategies that account for SOC uncertainty. We propose three constraint-tightening optimization approaches of increasing complexity: (i) a fixed-margin formulation, (ii) an adaptive-margin optimizer, and (iii) an uncertainty-aware optimization model. The latter explicitly accounts for the decision-dependent nature of the uncertainty. Numerical results demonstrate that while all three approaches robustify against SOC uncertainty, the uncertainty-aware formulation outperforms the others in maximizing revenue while ensuring reliable frequency reserve provision. This highlights the significance of treating SOC uncertainty as an endogenous process within the operational strategy.

ABSTRACT. The operation of Battery Energy Storage System (BESS) in photovoltaic-rich distribution grids creates a fundamental conflict between market-driven dispatch and local congestion management. Conventional volumetric energy charge (VEC) and peak power charge (PPC) tariff signals, which are typically designed for load-dominated grids, can worsen transformer loading in injection-constrained feeders. This paper presents a full-year co-simulation study evaluating BESS Energy Management System (EMS) strategies under TURPE7 HTB1-inspired grid tariff formulations. The study uses the SimBench rural Medium Voltage (MV) benchmark grid and a BESS participating in day-ahead and intraday electricity markets. In the investigated topology, conventional tariff formulations increase peak transformer loading by up to 16 percentage points above the no-BESS baseline. We introduce a tariff gaming index and propose an injection-aligned Path B design that combines charging credits with discharging penalties. The calibrated Path B design with a charging credit of c = −1.0 ct/kWh eliminates all time steps above 80 % transformer loading while achieving a positive net revenue of 856 k € and a gaming index of 0.097, the lowest non-zero value across all scenarios. The results provide quantitative guidance for Distribution System Operators (DSOs) designing dynamic tariffs in injection-dominated grid zones.

ABSTRACT. The high penetration of intermittent renewable generation in low-inertia island power systems imposes stringent operational security constraints, leading to renewable energy curtailment and greater reliance on ancillary services. While battery energy storage systems (BESSs) can facilitate energy shifting and reserve provision, most existing PV–BESS scheduling frameworks are tailored to the regulatory and operational structures of large interconnected power systems and do not capture regulation-driven constraints encountered in island grids. This paper proposes a regulation-aware framework for PV–BESS market participation, explicitly aligned with the Cyprus electricity market and ancillary services regulation. A mixed-integer co-optimization jointly schedules day-ahead energy dispatch and per-period reserve commitments, enforcing mandatory minimum reserve requirements (reserve floors) and state-of-charge (SOC)–based reserve sustainability under PV forecast uncertainty. Case studies demonstrate that explicitly reflecting ancillary services regulation improves reserve reliability while mitigating renewable energy curtailment in low-inertia island power grids.

ABSTRACT. Multi-terminal DC (MTDC) transmission systems based on modular multilevel converters (MMCs) are a key component of the envisioned future energy sector, where sustainability and efficiency are increasingly prioritized. To ensure their reliable operation, MMC currents must be regulated safely and rapidly under a wide range of uncertain operating conditions. Consequently, the design of current controllers faces a fundamental challenge: achieving fast transient response while maintaining robustness against uncertainties. This paper addresses this challenge by proposing a linear matrix inequality (LMI)–based design framework that leverages Lyapunov stability conditions to synthesize a less conservative static state-feedback controller. The proposed design method explicitly accounts for system constraints, including input saturation and overcurrent limits. The proposed method's effectiveness is assessed on the CIGRE MT-HVDC benchmark, simulated in RTDS, and compared with existing methods.

ABSTRACT. In this article, we propose a model predictive control (MPC) scheme with an exponential cost function, along with an observer for the Modular Multilevel Converter (MMC), to enhance converter dynamic performance. In particular, as the prediction horizon Np increases, the numerical conditioning deteriorates rapidly, especially when a large Np is employed. This research work uses an appropriate cost function weighted to overcome the limitations of a large Np. We further analyse the effects of constraints, observing that the designed MPC strictly adheres to them and that the control variable influences the MMC plant's response. The presence of the observer improves the prediction of the output, particularly for setpoint changes in the reference signal. We also analyze the prescribed performance, which provides a priori guarantees of closed-loop stability for the proposed controller.

ABSTRACT. Hybrid AC/DC power systems with grid-forming converters can provide frequency support between asynchronous AC areas through HVDC links. However, inertia-like active-power injection through a converter is coupled to the DC-side energy balance and may therefore compromise DC-voltage stability. This paper addresses this trade-off through a decentralized fixed-structure control-design framework for hybrid AC/DC systems. First, a reduced small-signal state-space model is developed that captures the coupling between AC frequency dynamics, converter active-power exchange, and DC-voltage dynamics. The full hybrid AC/DC model is then used offline to jointly tune block-diagonal converter controllers, while online operation relies only on local converter measurements. The control problem is formulated as a fixed-structure multiobjective H-infinity tuning problem: Frequency-support performance is treated as a soft objective, whereas DC-voltage and band-limited rate-of-change-of-frequency (filtered-RoCoF) channels are imposed as hard normalized induced-gain requirements in the linearized design model. The framework is evaluated on a two-area hybrid AC/DC test system interconnected by a bipolar HVDC link and compared with an enhanced VSM baseline including DC-voltage feedback. For the studied generator-outage disturbances and the selected baseline, the proposed controller achieves a more favorable trade-off between AC-side frequency support and DC-side voltage regulation without requiring online inter-area communication.

ABSTRACT. This paper proposes an adaptive model predictive control (AMPC) framework for inner current control of modular multilevel converters (MMCs) in high-voltage direct current (HVDC) systems. The unconstrained MPC feedback gain matrix Kmpc is precomputed offline over a range of control effort weights R and approximated element-wise via polynomial regression, enabling real-time gain evaluation without repeated matrix operations. Since the optimal control law depends only on the ratio Q/R, the state weighting matrix Q is held fixed while R is adapted online to balance transient response speed against control effort. A precomputed lookup table, derived from closed-loop step-response simulations across a grid of initial states and reference transitions, maps operating conditions to the smallest R satisfying the converter voltage headroom constraint at each sampling instant. Simulation results confirm that the proposed framework replicates the closed-loop performance of conventional MPC while significantly reducing the gain computation time and making it well-suited for real-time platforms.

ABSTRACT. The integration of massive offshore wind into hybrid AC-HVDC grids demands robust DC voltage regulation, yet conventional fixed-gain droop controllers struggle under severe stochastic volatility. This paper bridges the gap between system-level economic dispatch and converter-level control by proposing a novel Stochastic Optimal Power Flow (SOPF)-based adaptive droop framework. Rather than relying on heuristic or reactive tuning, wind forecast uncertainty is modeled using a zone-wise Beta distribution that accurately captures the heteroscedastic nature of wind errors across low, mid, and high power regimes. By leveraging Polynomial Chaos Expansion (PCE) within a chance-constrained SOPF, the system's stochastic states are formulated analytically. Crucially, the optimal adaptive droop gain is extracted directly from the first-order PCE coefficients via a Jacobian-free sensitivity analysis, embedding statistical voltage-security guarantees directly into the local converter control. Validation on a 4-terminal AC-HVDC system demonstrates that scenario-adaptive gains significantly outperform standard fixed-coefficient approaches, effectively minimizing active-power tracking errors during extreme wind disturbances.

ABSTRACT. VSC-HVDC links offer controllable active and reactive power output, making them a promising asset for emergency voltage support. This paper presents an analytical method for adjusting the active power setpoint of a VSC-HVDC station to maximise loadability during voltage-stressed conditions. By exploiting the geometric structure of converter capability limits, a closed-form expression for the optimal setpoint is derived under combined current and voltage constraints. The method requires local voltage measurements and an estimate of a wide-area voltage angle difference, making it suitable for real-time emergency control without global optimisation. Validation on the Nordic Test System confirms that the analytically predicted optimum is consistent with the setpoint yielding the highest loadability, and that adjustments of active power setpoint can yield a more-than-proportional increase in loadability. The results further indicate robustness to angle estimate uncertainty.

ABSTRACT. Accurate wind resource assessment relies on an appropriate parameterization of surface roughness, which influences the vertical extrapolation of wind speeds. In current practice, industry-standard modelling predominantly employs static roughness maps (e.g., CORINE Land Cover), which are updated only at multi-year intervals and therefore fail to capture rapid environmental changes. This study presents an Earth Observation (EO)-based workflow to derive temporally explicit roughness length information, providing a representation consistent with conventional land cover-based parameterizations while accounting for both seasonal phenology and inter-annual structural changes. By integrating Sentinel-2 NDVI time series with Global Forest Change data within Google Earth Engine (GEE), the approach captures temporal variations in forest-related roughness length conditions through a bounded, semi-empirical modulation of established roughness classes. A case study in Germany demonstrates pronounced seasonal and inter-annual changes, including a reduction of the highest roughness class from 28.1% in summer 2018 to 15.8% in summer 2022, and from 24.0% in winter 2018 to 8.9% in winter 2022—changes that remain undocumented in static land cover inventories. The proposed workflow does not aim to derive physically validated absolute roughness lengths but provides a framework that supports the interpretation of wind measurements in rapidly changing forested environments, particularly for shortened measurement campaigns.

ABSTRACT. The increasing integration of renewable energy generation units has a significant impact on power grids. The expansion of generation units such as wind turbines is usually uncoordinated and grid impacts are neglected. This paper presents a method for assessing connection nodes for generation units based on their impact on hosting capacity. To this end, the socalled grid utilization factor is introduced. In addition to the optimal placement of units, the hosting capacity also takes into account reactive power management and topology optimization. For an exemplary 110 kV grid in southern Germany, an initial hosting capacity for wind turbines of 1,069 MW is calculated. The results of the grid node assessment show, on the one hand, that a positive evaluation of grid connection nodes is generally possible until a local limit is reached. Furthermore, the analyses show that the optimization of the switching state has a significant influence on the evaluation of grid nodes.

ABSTRACT. This work proposes a multi-temporal aiming optimization framework for concentrating solar power tower plants based on neural network-driven constraint learning. The approach extends a steady-state methodology by exploiting temporal correlations between consecutive time instants through a fine-tuning process, allowing the reuse of previously computed solutions and trained models. This reduces the need to solve the full optimization problem at every instant while promoting smoother transitions between consecutive aiming configurations. The methodology is evaluated on the Dunhuang 10~MWe plant. Results show that the proposed strategy achieves comparable energy collection to the steady-state approach, while consistently reducing peak flux levels and heliostat angular displacements. In addition, a significant reduction in computation time is observed, improving the practical applicability of the method under changing operating conditions.

ABSTRACT. This work investigates the value of hybridizing wind, solar PV, and hybrid energy storage systems (HESS) in isolated, gas‑dominated power systems. Such systems face strong operational constraints, including limited grid capacity and significant variability in renewable generation, which increases the burden on gas turbines during unit commitment (UC) optimization. By colocating wind, PV, high‑energy (HE) batteries, and high‑power (HP) batteries behind a single grid connection point, Hybrid Power Plants (HPPs) can exploit overplanting, provide operational flexibility, and contribute to reserve requirements. A rolling UC framework is developed to compare HPPs with standalone wind/PV assets. The HPP forecaster computes available power based on wind/PV forecasts and HE battery state of charge, while the UC allocates reserves and dispatches thermal units and HP batteries. The HE battery state of charge is updated iteratively using forecast and realised power. Results show that hybridization alters gas turbine behaviour substantially. HPPs reduce residual load through coordinated storage use, shifting gas turbine operation away from inefficient mid‑load regimes and concentrating their usage at low‑load or high‑load states. They also decrease the need for gas turbines to stay online solely to provide reserves, thanks to HP‑battery‑based reserve provision. The results will be complemented by sizing sensitivities on RES and HESS capacities, identifying optimal hybrid configurations that minimize gas consumption and improve overall system flexibility.

ABSTRACT. Wave Energy Converters (WECs) are typically controlled using hydrodynamic strategies aimed at maximising absorbed mechanical power, while grid-side converters are designed independently to ensure compliant power injection. This decoupled approach may lead to suboptimal wave-to-grid (W2G) performance. This paper investigates whether control parameters that maximise absorbed mechanical energy also maximise the net electrical energy delivered to the grid. An integrated W2G modelling framework is developed, coupling a hydrodynamic model of interacting WEC arrays based on the Cummins' formulation with a detailed representation of the electrical conversion chain, including generator, rectifier, DC-link, and grid-connected inverter. A first-order impedance matching (IM) controller is adopted, and a systematic exploration of the controller parameter space is performed. The system behaviour is analysed under different sea states, and performance maps are constructed for both absorbed mechanical and grid-injected energy. The results reveal a consistent mismatch between the controller settings that maximise mechanical energy absorption and those that maximise grid-injected energy. Mechanically suboptimal configurations can yield significantly higher net electrical energy, as highly oscillatory power profiles associated with IM increase mechanical reactive power exchange and reduce the fraction of usable energy delivered to the grid. These findings demonstrate that maximising mechanical energy alone is not sufficient to achieve optimal W2G performance and highlight the need for integrated control strategies that explicitly account for electrical conversion dynamics and grid-level objectives.

ABSTRACT. This study introduces a mixed-integer linear programming model for the hourly scheduling of a closed-loop, price-taker pumped-storage hydropower plant (PSHP) operating under hydraulic short-circuit (SC) mode. The plant is assumed to participate in the day-ahead and secondary regulation reserve markets. The turbine input-output relation is represented through a two-segment linear curve that is updated iteratively as a function of the net head. Although the formulation is general, it has been tested with real prices from the Spanish electricity market, taking into account the current market design, which involves different auctions for upward and downward regulation reserves. The proposed formulation explicitly models the SC water flow. The model has been used to compare the plant's revenue with and without the hydraulic SC, and the results show that its implementation increases the plant's market revenues.

ABSTRACT. This paper presents a stochastic day-ahead scheduling model for energy communities (ECs) with small-scale pumped hydro storage (PHS) units. The communities do not manage the distribution network, which is owned by an independent operator with a third-party connection obligation. The model allows direct energy transactions among the members of the same community and includes different storage capabilities. The model uses a two-stage stochastic mixed-integer linear programming (MILP) methodology that represents the uncertainty in load and photovoltaic (PV) production profiles through copula-based scenario generation. The PHS unit model uses a piecewise linear representation based on special ordered sets of type 2 (SOS2) to limit the number of binary variables and ensure computational performance. The case study involves multiple ECs in a real medium-voltage distribution network. Load and production data are available for each user on three different days with 15 minutes resolution. The results show that internal transactions and optimized PHS operation increase flexibility and decrease electricity procurement costs.

ABSTRACT. This paper presents a comprehensive framework for solving a revenue-maximization problem for a battery storage integrated behind the meter of a PV plant. It computes the optimal schedule of the system accurately accounting for relevant capacity fade due to cyclic operation and for state-dependent operational efficiencies. The multiple non-linearities of the problem are handled relying on a Dynamic Programming framework, which enables the combined optimization of short-term PV–battery operation and long-term capacity revamping strategies due to degradation. The effectiveness of the proposed methodology is demonstrated through a case study considering a battery arbitraging in the Italian Wholesale Energy Market, while complying with Capacity Market contractual obligations. Results show that the proposed formulation yields an increment in the net profits compared to simplified modelling setups, e.g. based on constant efficiency assumptions.

ABSTRACT. The paper studies the coordinated operation of a Li-ion battery energy storage system (BESS) and a hydropower plant (HP) for the provision of primary frequency control (PFC) in the Spanish electric power system. For this purpose, dynamic simulation models of the HP and the BESS have been developed and integrated into a hybrid power plant (HPP) simulation model. The HPP simulation model considers the hydraulic transients in the HP penstock by means of an elastic water column approach, as well as the electrical and thermal dynamics of the BESS and its aging. The HPP simulation model comprises a control module aimed to distribute the HPP power set-point in such a way that the wear in the HP is minimized, and to keep as long as possible the state-of-charge (SOC) of the BESS within a prespecified band. The control module was proposed and studied in previous works of the authors. The paper compares with each other two different settings for the control module. First, the control module is assumed to be integrated in the speed governor of one of the HP units and to coordinate the operation of the BESS and this unit, as done in the few existing HPP comprising hydro generation and BESS. Second, the control module is assumed to be at a higher hierarchical level and to coordinate the operation of the BESS and all the HP units. The performance of the HPP is simulated under historic frequency profiles using the two control module settings. The comparison is performed based on the results of simulations as a function of the wear of different HP components, the BESS aging and the quality of the PFC provided by the HPP. Preliminary results show that the coordinated operation of the BESS and all HP units reduce the wear in the HP components at the expense of a higher BESS aging and a lower quality of the HPP PFC.

ABSTRACT. This work develops a methodology for maximum efficiency tracking for variable-speed pump as turbines systems. The system is connected to the grid through a voltage source converter which allows the speed control of the induction generator that is coupled to the turbine. The results demonstrate that the algorithm successfully tracks the maximum efficiency point, significantly improving the energy conversion efficiency compared to fixed-frequency operation and extracting more power from the flow.

ABSTRACT. In recent years, the growing adoption of electric vehicles has positioned lithium-ion batteries as a prominent research area, attracting extensive scientific attention. Lithium-ion batteries play a crucial role in ensuring that electric vehicles operate efficiently and safely. However, to ensure the reliability of electric vehicles and the accuracy of range estimation, the State of Charge (SOC) value of lithium-ion batteries must be calculated correctly. In this study, machine learning and deep learning-based data-driven methods were tested under different temperature and C-rate conditions; the results obtained were analysed using the RMSE, MAE, maximum error (MAX) and training time metrics. Using the data obtained under the different conditions, the performance variations of the selected data-driven models will be systematically analysed, enabling the determination of the operational conditions under which each model achieves the highest prediction accuracy.

ABSTRACT. The increasing share of converter-based resources fundamentally alters the relationship between grid strength and system stability. Conventional short-circuit–based indices, such as the Short-Circuit Ratio (SCR) and its extensions, characterize static current-injection capability but fail to capture critical dynamic aspects, including damping, frequency resilience, inertia, and control headroom. In contrast, impedance-based grid-strength methods are more suitable for small-signal and converter-driven stability assessments but not provide information about transient. To clarify these limitations, this work investigates the intrinsic dynamic behavior of grid-forming (GFM), grid-following (GFL), and synchronous generator (SG) systems, establishing reference characteristics for voltage, frequency, and power-control interactions. Based on these insights, multiple network scenarios are developed and verified using electromagnetic transient (EMT) simulations to assess the impact of topology, control mode, and grid capacity on system response [1]. The results demonstrate that networks with identical grid-strength values can exhibit markedly different transient and harmonic behaviors, highlighting the insufficiency of classical metrics for converter-dominated systems. In such networks, maintaining stability and power quality becomes increasingly challenging due to gridweakness–induced phenomena such as voltage dips, harmonic distortion, oscillations, and synchronization loss. Conventional GFL inverters relying on phase-locked loops are particularly sensitive to weak-grid conditions, where increased impedance and reduced short-circuit capacity may introduce negative damping and instability [2]. GFM inverters have emerged as promising alternatives by providing voltage-forming capability and synthetic inertia [3]; however, their behavior is governed by control loops rather than physical inertia, making them vulnerable under large-signal disturbances such as faults. As a result, the stability assessment of converter-dominated devices and microgrids remains an open challenge. Grid strength is commonly defined as the ability of a power system to maintain stable voltage under disturbances [4]. This work develops a comparative EMT-based framework to evaluate existing grid-strength concepts by analyzing the dynamic behavior of system components and identifying the factors critical for robust stability assessment.

ABSTRACT. The growing integration of converter-based and black-box assets in modern power systems poses significant challenges for comprehensive stability assessment. Traditional methods address steady-state, small-signal, and large-signal stability separately, leading to fragmented analyses and high computational demands, particularly for large-scale systems with operating-point variability. Moreover, many existing approaches require detailed dynamic models, which are often unavailable for black-box components. This paper presents a sequential stability assessment methodology that unifies steady-state, small-signal, and large-signal analyses within a coherent and scalable framework. By combining frequency-domain impedance-based techniques with scenario-driven simulations, the proposed approach enables efficient and robust stability evaluation while remaining compatible with black-box converter-based assets. The effectiveness of the methodology is demonstrated through its application to an electrical energy hub.

ABSTRACT. Power systems are undergoing a significant transformation due to the increasing penetration of power converters, which do not inherently provide inertia and damping. This reduction of synchronous generators in power systems is leading to new stability challenges, prompting system operators to update grid codes. This paper evaluates the impact of electromechanical oscillations on the benchmark proposed by the Spanish Transmission System Operator under grid-forming converter operation. A methodology for damping selection presented by the authors in a previous work on virtual synchronous generators is updated for grid-forming converters. The study is carried out using Control Hardware-in-the-Loop simulations, where the system is subjected to the disturbance defined in the Spanish grid code for evaluating the damping of electromechanical oscillations.

ABSTRACT. In converter-dominated power systems, frequency support is provided by control rather than by synchronous-machine physics. Dynamic Virtual Power Plants (DVPPs) coordinate heterogeneous converter-based resources through prescribed aggregate transfer behavior, but existing studies mainly validate single aggregates and therefore leave multi-DVPP interaction insufficiently understood. This paper investigates coordinated DVPP control in an islanded IEEE 9-bus-derived benchmark with heterogeneous grid-forming (GFM) and grid-following (GFL) plant clusters replacing all synchronous generators. Under identical physical conditions, decentralized droop-based and coordinated DVPP control are compared at aggregate level, implementation-level channel interaction is examined, and coupled-system admissibility is assessed under gain variation. The results show that DVPP control preserves a similar stationary frequency response to droop-based control while redistributing transient dynamics more flexibly. They further show that converter-side stabilization can weaken the nominal P-f / Q-V decoupling and that plant-wise admissible DVPP parametrizations need not remain admissible in multi-DVPP operation.

ABSTRACT. In the maritime industry, environmental regulations are driving a rapid shift toward electric and green propulsion systems. In vessels adopting the DC technology, an insufficient design integration (particularly between control systems and filtering stages) can cause underdamped behavior of the DC bus voltage. The latter becomes critical when main bus protections are triggered or loads malfunction (specifically in the case of vital loads). To mitigate these risks and preserve power system stability, one solution is the integration of the Active Damping (AD) function. The latter transiently increases the system damping factor, thus avoiding uncontrolled voltage oscillations after perturbations. The paper aims to validate Active Damping in a Power-Hardware-In-the-Loop (P-HIL) environment, specifically modeling a DC-powered passenger vessel. P-HIL tests demonstrate the critical role of AD in managing vital loads, like the electric steering gear. The results also show how the P-HIL implementation can close the gap between design and testing of electric DC vessels.

ABSTRACT. The increasing penetration of distributed energy resources (DERs) has significantly altered the dynamic behavior of emerging power systems, particularly during severe voltage disturbances. Unlike synchronous generators, inverter-based DERs are inherently current-limited and rely on phase-locked loop (PLL) strategy for grid synchronization, which makes them prone to loss of synchronism during system faults. This paper develops an analytical framework to characterize the synchronism and stability limits of DERs under under-voltage conditions. A DER nonlinear state–space representation is used to mathematically derive conditions that ensure DER synchronism around a feasible equilibrium point. These conditions lead to the formulation of a voltage-dependent DER current envelope that defines the permissible DER current range required for stable fault ride-through (FRT) operation. The analysis shows that, beyond the inverter over-current protection constraint, more restrictive current limits may arise from the interactions between DER control dynamics and grid impedance characteristics. The impacts of grid impedance, controller parameters, and DER power factor on the stability boundary are analytically studied. It is studied how and to what extent DER power factor adjustment can enhance its stability during faults. Time-domain simulations validate the efficacy of the mathematically derived current envelope and demonstrate the effectiveness of the power factor adjustment in preventing DER tripping events.

ABSTRACT. The ongoing energy transition demands that distribution system operators perform advanced calculations and analyses in near-real time. This is challenging in low-voltage distribution networks, where exact network topology is often unknown or insufficiently documented. Machine learning can bring significant advancement in addressing this problem, as it enables the use of smart meter measurement data to train various algorithms, such as deep neural networks, for effective model-free voltage calculations. Existing model-free approaches, that are reliant on machine learning algorithms training for input-output mapping, suffer from scalability issues when observed networks have high dimensionality and a lot of end-users. This paper assesses the robustness of various dimensionality reduction techniques, such as Principal Component Analysis, autoencoders and Isomap, in the compact learning model-free pipeline. The results demonstrate that all three of the observed dimensionality reduction methods produce fast, scalable and accurate voltage estimation in the compact learning model-free voltage calculation context, that is independent of network topology at any point. Final validation of the results produces an average estimation error below 1 V, which can be considered negligible in real-world application.

ABSTRACT. Distribution grids are often sparsely monitored, which limits observability and accurate voltage estimation. While traditional state estimation methods rely on pseudo-measurements that can reduce accuracy, most data-driven estimators depend on a specific monitoring configuration. This paper proposes an iterative optimal meter placement strategy for a Correlated Gaussian Process (Corr-GP) voltage estimator. Using spatial voltage correlations to infer unmeasured states via Corr-GP, meter locations are selected based on incremental information gain. As such, near-optimal performance compared to an exhaustive search is achieved while significantly reducing the required computational effort. Tests across multiple operating scenarios and comparisons with a traditional exhaustive approach confirm the robustness and superiority of the proposed method, demonstrating its potential as a practical solution for improving observability in future distribution grids.

ABSTRACT. The increasing integration of photovoltaic (PV) systems necessitates accurate voltage monitoring in distribution feeders, yet traditional model-based assessments are often constrained by the lack of detailed grid parameters. While data-driven models offer a scalable alternative, purely black-box approaches frequently fail to generalize to operating conditions unseen during training, such as high PV penetration. This paper proposes a physics-informed neural network (PINN) framework that achieves robust, model-free voltage estimation by integrating an empirical sensitivity loss—derived solely from historical measurement data—into the objective function. By embedding the inherent physical relationship between changes in active power and voltage, the model internalizes grid behavior without requiring explicit knowledge of line impedances or topology. Validation on a typical Japanese distribution model demonstrates that the PINN significantly improves extrapolation performance, successfully reproducing midday voltage rises even when trained exclusively on no-PV data and reducing root-mean-squared error (RMSE) by 23.2% at 75% PV penetration compared to a simple neural network. Furthermore, the model exhibits resilience to data scarcity, maintaining high precision with only 50% of customer nodes available for training, and achieves inference speeds orders of magnitude faster than conventional power-flow solvers. This framework is theoretically adaptable to dynamic network changes, such as on-load tap changer operations, through a sliding window approach that periodically updates sensitivities, providing a practical and scalable monitoring tool for distribution system operators and aggregators.

ABSTRACT. Accurate load modeling is key to power system analysis at the planning and operational time scales. Traditionally, loads are represented using static models such as the ZIP and exponential models, which express power as a polynomial function of voltage. The aim of load modeling is to estimate the model parameters from measurements. This paper introduces an alternative perspective on the problem: rather than refining the estimation algorithm itself, the measured power must first be decomposed into its constituent components: nominal and voltage-dependent powers. Two distinct approaches to perform this decomposition are developed and benchmarked. The first is an analytical method grounded in the Frisch-Waugh-Lovell (FWL) theorem from econometric theory. The second approach is data-driven, employing a Patch Time Series Transformer (PatchTST) model trained on simulated load data with known ground-truth decompositions. The results show that both methods can successfully decompose power measurements; however, the FWL method performs better than PatchTST.

ABSTRACT. The integration of detailed Smart Power Cell (SPC) simulations into higher-level coordination frameworks, such as transmission system controllers, is limited by their high computational cost. This paper presents a data-driven approach to model the dynamic behavior of an SPC at the TSO–SPC interconnection. A high-fidelity MATLAB/Simulink model is used to generate data for learning the system response to active and reactive power reference signals. A Long Short-Term Memory (LSTM) neural network is employed to approximate the aggregated SPC behavior. The results show that the main power flow dynamics are captured with low prediction errors while requiring considerably less computational effort than the high-fidelity simulation. This makes the approach suitable for future studies of SPC behavior in higher-level coordination applications.

ABSTRACT. Accurate residential photovoltaic (PV) self-consumption estimation is crucial for grid management, yet traditionally requires rare, dedicated sub-metering. Most residential setups provide only aggregate net load data, masking the underlying bidirectional energy flows. While existing load disaggregation methods attempt to solve this problem, their reliance on supervised training, appliance-level labels, and site-specific metadata severely limits scalability. To overcome these limitations, this work proposes a novel non-intrusive framework that estimates PV self-consumption using exclusively aggregate net load and clear-sky irradiance data. The core contribution is a Physics-Anchored Domain-Adversarial (PADA) mechanism that embeds physical constraints to reconstruct gross household demand and PV generation in a strictly unsupervised manner. By eliminating the need for target-domain labels or model retraining, PADA exhibits robust generalization to unseen residential environments. Validation on diverse real-world datasets confirms that our framework successfully disentangles unmetered energy flows, outperforming standard supervised baselines (e.g., Convolutional Recurrent Neural Networks and Transformer architectures) by achieving up to a ~33% relative reduction in Self-Consumption Rate (SCR) error, for scalable, physics-aware prosumer monitoring.

ABSTRACT. Electric distribution networks are expected to experience significant challenges in the coming years due to the continuous growth in demand, driven by the connection of new residential and industrial loads, data centres, and the electrification of mobility [1]. This increase could lead to voltage issues and overloads, highlighting the need for network investments to accommodate and support these demand rises, in order to ensure an efficient power supply for all consumers.

To address these issues and reinforce the network, traditional solutions [2] have been the most widely used. These are based on investments in network components, such as new lines or the enhancement of existing lines and substations. In some cases, these solutions present drawbacks, including high costs, permitting challenges, and even the impossibility of implementing further improvements.

In contrast to traditional solutions, there is an alternative approach to address these issues, namely Non-Wire Alternatives (NWAs) [3], which involve the use and deployment of distributed resources or management techniques. As indicated in [4], these solutions can require lower investment and have the potential to significantly increase the capacity of the existing network without the need for conventional infrastructure investments.

In the state of the art, there are various studies that size and locate distributed resources such as NWAs from different approaches. These studies focus on the more technical aspects, maximizing integration, minimizing solution costs, or selecting the most appropriate technologies, including hybridization with storage systems. The main and most important drawback for real or practical application is that they do not analyse the interactions among the actors involved, such as the operator or the private owners of generation assets (in many countries, legislation does not allow distributors to own generation assets).

Some methods [5] address the problem using the locational marginal value, which is the remuneration provided to NWAs for their services, but without modeling a direct interaction within the planning process. Other methods [6], in their approach, do consider the desired profitability of the private owner by integrating it into the choice of solutions; however, they do not represent the distribution utilities’ planning sequence nor the remuneration mechanisms. modeled as a cooperative game, for the siting and sizing of distributed energy resources (DERS) acting as non-wires alternatives (NWA) in power networks. The DSO and private investors are modeled as cooperative agents that coordinate their decisions to achieve mutually beneficial outcomes. The model captures the strategic interaction between network planning and private investment, integrating technical criteria (congestion reduction, loss minimization, and reliability improvement) and economic criteria (investor return on investment). Through cooperation and information sharing, the participants align their objectives to enhance overall system efficiency and social welfare. The approach is flexible and scalable, capable of incorporating multiple DERS technologies and investment scenarios, providing a robust framework to evaluate the potential of NWAs as a substitute for traditional infrastructure investments, aligned with current trends in energy transition and smart grids.

[1] European Parliament, “Report on electricity grids: the backbone of the EU energy system,” A10 0091/2025, European Parliament, 2025. [2] Distribution network expansion planning: An updated review of current methods and new challenges, A. Rastgou, Renewable and Sustainable Energy Reviews, vol. 189, p. 114062, 2024, doi: 10.1016/j.rser.2023.114062 [3] Eastern Interconnection States’ Planning Council, “Getting the Signals Straight : Modeling, Planning, and Implementing Non-Transmission Alternatives Study,” Tech. Rep. February, 2015. [4] ACER and CEER, Challenges of the Future Electricity System – Recommendations and Commitments, July 11, 2024. [5] P. Andrianesis, M. Caramanis, R. D. Masiello, R. D. Tabors and S. Bahramirad, "Locational Marginal Value of Distributed Energy Resources as Non-Wires Alternatives", in IEEE Transactions on Smart Grid, vol. 11, no. 1, pp. 270-280, Jan. 2020, doi: 10.1109/TSG.2019.2921205. [6] N. D. Laws, M. E. Webber, and D. Chen, "Valuing distributed energy resources for non-wires alternatives," Electric Power Systems Research, vol. 234, p. 110521, 2024. doi: 10.1016/j.epsr.2024.110521. [7] R. F. S. Budi, S. Sarjiya, and S. P. Hadi, “Multi-level game theory model for partially deregulated generation expansion planning,” Energy, vol. 237, p. 121565, 2021, doi: 10.1016/j.energy.2021.121565

ABSTRACT. The secure operation of power distribution systems is challenged by the growing integration of distributed energy resources. Leveraging the flexibility of battery storage offers a cost-effective alternative to measures like generation curtailment, which results in energy losses. However, developing an effective operational model for battery storage is hindered by inaccurate grid models, unavailability of load data, nonlinear relationship between power injections and network states, intertemporal constraints, and complex electrochemical and thermal dynamics. To address these challenges, this paper proposes a data-driven operational control scheme for battery storage in distribution systems. Linear and convex quadratic operational constraints are constructed based on real-time distribution system and battery storage measurements. Lyapunov optimization decouples multi-period battery operation, enabling a real-time, forecast-free control strategy with low computational complexity. Numerical studies using nonlinear distribution system and battery storage simulators validate the effectiveness of the approach in ensuring secure distribution system operation and satisfaction of voltage and thermal constraints of battery storage.

ABSTRACT. This paper proposes a bilevel optimization model for Demand Response Programs (DRPs) in medium-voltage distribution systems, utilizing Battery Energy Storage Systems (BESS) in low-voltage circuits. The model aims to minimize system loading, particularly on critical lines, to defer network reinforcement investments. It also seeks to reduce system losses and promote tariff arbitrage, enhancing BESS economic viability. The model uses locational and hourly sensitivity signals to guide DRP implementation and BESS placement. The effectiveness of the model is demonstrated through simulations on the IEEE 123-bus system, showcasing its ability to reduce peak loading, minimize losses, and optimize BESS operation for DRP. The system voltage is indirectly controlled, preventing potential overvoltage and undervoltage. Furthermore, the model is effective in systems with a high penetration of distributed generation, particularly photovoltaic systems, taking advantage of periods with higher generation for battery charging. The model's accuracy is validated by comparing its results with OpenDSS simulations, showing good agreement, with an average error of less than 0.4%. The proposed framework offers a flexible solution for integrating BESS into distribution systems, facilitating DRPs, and improving overall system efficiency.

ABSTRACT. The increasing deployment of Battery Energy Storage Systems (BESSs) provides a source of decentralized flexibility, but purely market-driven operation may aggravate loading constraints in distribution grids. This motivates a need for grid-neutral integration concepts that limit adverse grid impacts while preserving sufficient economic incentives for BESS operators. This paper presents an Operating Envelope (OE) framework for grid-neutral BESS operations that aligns with current Distribution System Operator (DSO) practices. First, we introduce a definition of the term grid-neutrality. Second, we derive static and dynamic OEs from measured transformer power flows and admissible transformer capacity limits using either historical measurements or day-ahead forecasts. Third, we integrate resulting limits into a sequential BESS dispatch optimization to evaluate techno-economic impacts. The results indicate a clear trade-off between grid protection and storage profitability. Conservative static OEs effectively reduce transformer loading but significantly decrease annual profits, while dynamic OEs achieve a better balance. However, the performance of dynamic OEs heavily depends on forecast quality, and exponential smoothing does not capture critical loading peaks well enough. Overall, OEs seem to be a practical short-term approach for enabling grid-neutral BESS integration.

ABSTRACT. Battery energy storage systems (BESSs) are increasingly deployed in distribution systems (DSs) to enhance stability and improve the utilization of energy from renewable energy sources (RESs). However, their optimal sizing, siting, and operation remain open challenges. This paper proposes a novel optimization strategy for joint BESS sizing and daily charging/discharging scheduling in unbalanced distribution systems with a given level of RES penetration, while fulfilling given voltage and state-of-charge (SOC) limits. Voltage variability is estimated using affine arithmetic (AA) applied to historical data and refined through a sensitivity-based linearized model. The resulting formulation relies on a relaxed continuous approach (i.e., avoiding binary variables for siting) with a differentiable cost function, ensuring convexity, scalability, and computational efficiency, with only minor suboptimality. Simulations on a modified IEEE 34-bus system with different PV penetration levels demonstrate the effectiveness of the approach, achieving overall cost minimization for Distribution System Operators (DSOs) with just sporadic voltage violations and with 90% of the total BESS capacity concentrated in no more than 7 bus-phase pairs.

ABSTRACT. Rapid adoption of electric vehicles (EVs) introduces new challenges to modern power systems, as charging infrastructure is still inadequate. This paper examines the impact of on-board EV charging on power quality of an electric grid, analyzing 121 real-world AC charging sessions at power levels ranging from 3–22 kW. The study evaluates current distortion during three key charging stages: ramp-up, constant, and ramp-down. The difference between current total distortion (TDi) and total harmonic distortion (THDi) is found to be often small (~0.2 A). The results also show that the relative share of distortion increases when the total charging current decreases. Charging multiple EVs with low currents, such as <10 A, may lead to excess current distortion.

ABSTRACT. A novel topology of single-phase multi-objective active rectifier (MOAR) for applications in smart grids is proposed. The MOAR can operate as an active rectifier, while it is also capable of offering a set of additional features for enhancing power quality in smart grids, including the possibility of: (i) operating as a grid-tied converter consuming only AC current with selective harmonic content; (ii) operating as a grid-tied converter consuming only AC current with controlled phase angle; (iii) and a combination of the previous modes, i.e., controlling active and reactive power and producing selective harmonics for the electrical grid. In addition, the proposed MOAR presents as main advantages the possibility of supplying the DC side load with non-switched power and, in case of failure of the control system or of the controlled switching devices, it can operate as a passive rectifier, continuing to guarantee the supply of the DC side load. The proposed MOAR, including the operating principle and control strategies, is presented throughout the paper. The validation is performed with a laboratory prototype, and a set of experimental results are reported for validating the proposal.

ABSTRACT. The adoption of Ship-to-Shore (S2S) power systems is accelerating as ports transition toward low-emission and smart energy infrastructures, increasing the need for reliable shipboard transformers capable of supporting continuous shore side operation. However, existing studies rarely evaluate how real port operating conditions and berth load uncertainty affect transformer level harmonic stress in bulk carriers. This paper addresses this gap by introducing a week-long, high-resolution berth-load dataset from operational activities at two North American ports, and by developing a dynamic state-space model of the ship side transformer driven by measured nonlinear and uncertain load behavior. Harmonic analysis up to the 50th order, including magnitude and phase angle characterization, is performed to assess power quality impacts and reliability risks under realistic loading scenarios. Results show a significant divergence between total harmonic distortion (THD) and total demand distortion (TDD): voltage and current THD reach 16.84% and 19.84%, well above IEEE 519 limits, but the corresponding TDD remains compliant at 4.47%. These findings highlight that S2S systems may satisfy demand-based distortion criteria while still experiencing excessive harmonic stress capable of accelerating insulation ageing, reducing transformer lifetime, and compromising power quality requirements. The study demonstrates the necessity of reliability oriented transformer assessment for robust S2S integration and supports future development of harmonic mitigation strategies tailored to port activity–driven uncertainty.

ABSTRACT. Power distribution networks have undergone major transformations globally, both on the generation and demand sides. The most recent shift has occurred on the demand side, driven by the massive proliferation of data centers as part of the significant investment in artificial intelligence. In fact, one of the most significant challenges currently faced by energy systems worldwide is the uncertainty in estimating the load growth for data centers. Likewise, data center loads can also become a significant source of electromagnetic interference in terms of power quality (PQ) issues due to the wide use of power electronics-rich devices and their own power consumption dynamics, as they often alternate between periods of relatively low energy consumption and sudden demand peaks. Concerning harmonic distortion, several strategies have been proposed for mitigation at the load level, mainly in the field of power electronics converters, although more complex strategies regarding the optimal location and sizing of compensation devices within the distribution network have not been extensively studied in the literature with these new loads. This paper addresses the problem of optimal placement and sizing of active harmonic filters in three-phase four-wire power distribution systems. To do so, a second-order cone programming problem is proposed to minimize the average voltage harmonic distortion throughout the buses, unlike most problems in the literature, which are formulated as nonlinear and solved using heuristic and metaheuristic strategies.

ABSTRACT. This paper studies how the choice of synthetic training data affects the performance of data-driven power quality disturbance classifiers. Three datasets with systematically increasing parameter variability are constructed to represent low-, medium-, and high-variability training scenarios. Four literature-based classifiers, ranging from explicit feature-engineering approachesto end-to-end deep-learning models, are trained on each dataset and evaluated both on their respective hold-out test splits and under cross-dataset transfer to the most variable dataset. The results show that low-variability synthetic data can lead to overly optimistic in-distribution performance, while robustness under increased signal diversity depends strongly on the model architecture. In particular, feature-engineered approaches degrade markedly under variability shift, whereas models operating directly on the waveform remain substantially more robust, with the transformer-based model showing the strongest overall generalization.

ABSTRACT. Harmonic monitoring in radial distribution feeders is essential for maintaining power quality, yet it typically requires costly measurement infrastructure. This paper proposes an approach based on a single power quality meter located at the point of common coupling of a microgrid. From the acquired three-phase voltage and current signals, a metaheuristic was employed to determine both the localization and identification of harmonic sources located in the utility's side. The method aims to assess the minimum hardware requirements for enabling assist power quality monitoring and, for this purpose, it was validated on the IEEE 34-bus test feeder. The localization stage achieved mean absolute errors ranging from 6.8 to 24.2 km, considering noisy measurement scenarios. Despite the imprecision noticed for localization stage, the identification stage (i.e., the most important process) yielded mean absolute errors ranging from 0.010 to 0.081 p.u. Thus, as the main objective is to understand the harmonic contribution from the utility's side, the results demonstrate the practical feasibility of low-cost harmonic source monitoring, even under noisy measurement conditions.

ABSTRACT. The increasing integration of renewable energy sources and demand electrification is increasing the operational complexity of power grids. Reinforcement learning (RL) agents have shown promising performance in real-time congestion management with topology optimization, but quantifying their reliability under unseen operating conditions remains difficult. This paper proposes a framework to quantify and predict the reliability of pre-trained RL agents for real-time congestion management. The approach integrates epistemic and aleatoric uncertainties into a machine learning–based failure prediction model that estimates the probability of RL agent failure 1 hour ahead, accounting for potential contingency scenarios. Results obtained in the Grid2Op AI-friendly environment demonstrate the framework’s ability to anticipate with an accuracy between 70% and 90% low-confidence AI recommendations and provide early warnings to human operators.

ABSTRACT. Home Energy Management Systems (HEMS) involve classical scheduling optimization problems that are often NP-hard, particularly when nonlinear components are present. Despite this complexity, HEMS require daily computation, demanding solutions within strict time constraints. To address this, surrogate models are introduced to approximate the original problem by retaining only a subset of its constraints. Accurate prediction of active constraints is therefore essential. By representing the optimization model as a bipartite graph, Graph Neural Networks (GNNs) can be adopted to capture structural dependencies and improve the detection of active constraints. Several message-passing architectures are investigated and compared against traditional Random Forest (RF) and Neural Network (NN) classifiers. The proposed methodology is tested on two HEMS scenarios: one based on Linear Programming (LP) and the other on Mixed-Integer Linear Programming (MILP). Results show that GNN-based classifiers consistently outperform RF and NN models, particularly in the more complex MILP case, thanks to its topological awareness.

ABSTRACT. This paper presents the results of a study on the application potential of artificial intelligence in the energy sector. The analysis uses a fictional municipal utility as a case study. A comprehensive metric is introduced to systematically assess and evaluate activities regarding their potential for AI applications. The results outline domains, processes, and activities with potential for AI, serving as a guide for future investigation.

ABSTRACT. The rapid integration of distributed energy resources and flexible loads is shifting distribution grids from unidirectional to bidirectional power flows. This transition increases the risk of violating thermal, voltage, and transformer limits, often requiring active power curtailment. While Reinforcement Learning (RL) offers a model-free approach to managing this congestion, its efficacy is threatened by grid uncertainties. Measurement inaccuracies and synchronization errors, combined with uncertain grid parameters like line impedances, create a significant model mismatch between the RL agent’s environment and the physical system. This paper investigates the robustness of end-to-end RL for congestion management on a real low-voltage grid, specifically analyzing how these uncertainties impact operational reliability and decision-making performance. Moreover, this work provides a set of metrics that assess the impact on power curtailment and distribution grid constraint violations to distinguish between corrective and unnecessary control actions.

ABSTRACT. In this work, we compare the performance of three reinforcement learning (RL) algorithms, namely Proximal Policy Optimization (PPO), Rainbow and Implicit Quantile Networks (IQN), for learning power system topology control. Agents are trained and evaluated on a version of the IEEE 14-bus test system, as well as a 36-bus subset of the IEEE 118-bus system. Evaluation on the 14-bus system shows high performance across algorithms, with PPO narrowly outperforming IQN and Rain- bow. However, while IQN achieves full survival across the test scenarios for the 36-bus system, PPO only narrowly outperforms a passive baseline. Additionally, we introduce a novel action space factorization which enables training on the 36-bus system without any action pruning, and a novel problem formulation and reward function which align the problem structure with real time operations.

ABSTRACT. Hyperscale AI data centers induce spatially and temporally correlated load fluctuations that violate classical independence assumptions and are not captured by time-averaged spectral methods. These correlations are episodic and non-stationary, requiring analysis that resolves transient structure. This paper applies Dynamic Mode Decomposition (DMD) to the temporal evolution of pairwise inter-bus correlation coefficients to form a low-dimensional state representation that enables modal analysis without a stationarity assumption. DMD eigenvalues encode the correlation regime: their location in the complex plane distinguishes sustained coherence, decaying transients, and intensifying events, while oscillation frequency maps to underlying physical coupling mechanisms. Using an IEEE 39-bus Real-Time Digital Simulator (RTDS) testbed with three converter-interfaced AI data center loads driven by synthetic workload profiles, global DMD provides a time-averaged modal baseline in a slow thermal band ($f \approx 0.005$\,Hz, $|\mu| = 0.91$) captures 93.6\% of total correlation energy. A sliding-window DMD formulation identifies transient intensification events: 51 of 775 windows (6.6\%) satisfy the $|\mu_k^{(n)}| > 1$ criterion, which aligns with stochastic workload coincidences. Cross-validation with RTDS voltage coherence confirms elevated coupling during these intervals. The proposed modal growth indicator provides an early-warning signal of correlation intensification prior to peak pairwise coherence.

ABSTRACT. The increasing penetration of renewable generation is introducing new operational challenges in power systems, particularly in terms of frequency stability. Battery Energy Storage Systems (BESS) have demonstrated a strong capability to provide fast frequency support. However, there is a need for dynamic models of BESS and their controls in phasor-oriented power system simulation models to analyse their effect on large transmission networks after severe disturbances.

In power system simulation tools, battery behaviour is typically represented using converter models adapted to emulate storage operation. These models involve a large number of parameters, making parametric analysis difficult. A notable exception is CBEST, a dynamic battery model developed by EPRI. While CBEST is specifically designed to represent BESS, it only provides voltage related control capabilities and doesn't include the capacity to measure the bus frequency and act accordingly without the support of additional modelling developments.

This work presents an open-source auxiliary user-model compatible with the CBEST model that enables frequency response of battery systems. The proposed model measures the frequency at the connection bus and generates an active power control signal based on a droop control, which is transmitted to the main battery model. The proposed structure allows isolating the frequency control contribution of BESS, facilitating replicable studies and the evaluation of parameter settings, battery locations and control configurations. The proposed model has been implemented and tested in a representative benchmark power system, demonstrating its applicability and scalability to large -scale transmission networks.

ABSTRACT. The proliferation of distributed renewable generation inherently reduces system inertia, resulting in weaker MGs where rapid and robust frequency support is vital. To provide this essential stability, these modern systems increasingly rely on grid-forming converters and synchronous units. However, inherent phase unbalances in low-voltage distribution networks and microgrids create asymmetrical voltage profiles that severely restrict the active power capability of these balanced three-phase sources. During frequency events, balanced power injections can trigger voltage protections on depressed or elevated phases, creating a critical operational bottleneck. To address this, this paper proposes a centralized co-optimization framework based on a Genetic Algorithm and Optimal Power Flow. The proposed procedure optimizes the dispatch of existing distributed single-phase resources alongside the strategic placement of newly installed three-phase generators, both balanced and unbalanced. Furthermore, the study incorporates thermal calculations demonstrating that, for the short durations typical of inertial responses, nominal current limits can be safely exceeded.

ABSTRACT. Microgrid formation has been widely recognized as a promising approach to enhance the resilience of distribution systems under extreme events. Resilience is typically assessed through the amount of load-shedding required in the post-contingency state, which is commonly evaluated using different optimal power flow (OPF) formulations, ranging from simplified DC OPF to AC OPF and frequency-constrained AC OPF models. However, the impact of the modeling level on the accuracy of resilience assessment remains insufficiently explored. This paper investigates how the level of modeling detail influences the estimation of load-shedding in islanded microgrids, while explicitly accounting for the pre-contingency operating state. The analysis is conducted on the IEEE 33-bus system under multiple case studies representing different generation-demand conditions. The results show that the DC OPF formulation systematically underestimates load-shedding, while AC OPF provides more realistic results by considering voltage and reactive power constraints. Furthermore, the inclusion of frequency constraints leads to additional load-shedding, particularly in scenarios characterized by generation deficits, where frequency stability becomes a limiting factor.

ABSTRACT. Load–frequency control is a fundamental function in the operation of power systems, responsible for maintaining within acceptable values the frequency of the system and the scheduled power exchange. In 2024, the Spanish Transmission System Operator, Red El´ectrica, implemented a new Secondary Regulation Service (SRS) as part of the European PICASSO initiative for the coordination of automatic Frequency Restoration Reserves. This transition represents a fundamental change with respect to the former shared peninsular regulation scheme and requires a profound adaptation of the national Automatic Generation Control system, particularly of the master regulator. This paper presents a comprehensive framework for the modeling, tuning, and validation of the Spanish master regulator under the new SRS. The proposed approach integrates a detailed simulation of the new SRS. The methodology is validated through a realistic case study that reproduces actual SRS operating conditions and evaluates both master regulator performance and BSP dynamic response. The results demonstrate the effectiveness of the proposed modeling approach and its applicability to other European aFRR implementations.

ABSTRACT. With the use of converter-interfaced renewable generation, the system inertia reduces, and the need for fast primary frequency response (PFR) increases. This paper proposes a decentralized frequency-temperature droop control framework for grid-interactive electric water heaters (GIEWHs) that enables autonomous support using only local measurements. We develop a physics-based thermal model to represent the dynamic behavior of individual GIEWHs and evaluate the proposed control strategy on the IEEE European Low-Voltage Test Feeder with 55 residential customers under low-inertia disturbance conditions. Our results show that the proposed method perfoms better than static droop control by improving the frequency nadir and enabling faster disturbance recovery, providing a better balance between grid requirements and user convenience. As such, this framework provides a practical, scalable, and communication-free approach to decentralized grid support.

ABSTRACT. Electric water heaters (EWHs) represent a promising aggregated flexibility resource for participation in balancing energy markets due to their inherent thermal storage capability. This paper proposes a stochastic Model Predictive Control (MPC) framework for optimal bidding by aggregators of domestic EWHs in the mFRR energy activation market (EAM). The framework jointly optimizes bid submission decisions, up- and down-regulation volumes, and device-level dispatch within a rolling-horizon mixed-integer quadratic programming (MIQP) formulation. Activation uncertainty is captured through quasi-Monte Carlo scenario generation with temporally correlated activation patterns, while market timing requirements, bid commitment obligations, and end-user comfort constraints are directly embedded in the optimization. Evaluated on a 100-EWH, 300 kW portfolio, using real Nordic market data across three representative seasonal weeks, the framework yields weekly net profits ranging from 158 to 976 EUR, depending on market conditions, with average iteration times below 40 seconds, confirming its viability for real-time market participation.

ABSTRACT. This paper addresses the problem of maintaining measurement integrity in distribution power systems under false data injection (FDI) attacks in the presence of distributed energy resources and high stochastic variability. We propose a real-time detect-and-repair framework based on a parallel generative adversarial network (ParaGAN) that learns normal multivariate time-series behavior in an unsupervised manner and uses reconstruction-based scores to identify stealthy anomalies. Unlike reconstruction-based baselines such as long short-term memory (LSTM) autoencoder, LSTM variational autoencoder, and TadGAN, ParaGAN employs parallel convolutional and recurrent branches with dual critics and a shape-aware reconstruction loss to jointly enhance detection sensitivity and recovery fidelity. The framework is evaluated on an IEEE 33-bus distribution test system with two photovoltaic generators using datasets generated from alternating-current power-flow simulations with randomized load and photovoltaic conditions. In the most challenging combined-attack scenario, ParaGAN achieves an F1-score of 0.9964 and an accuracy of 0.9871, outperforming all baselines, while yielding the lowest mean squared error and highest cosine similarity for reconstruction. An economic case study further shows that the proposed recovery mechanism can reduce attack-induced operating-cost losses by approximately 75%, indicating that ParaGAN can serve as an effective cyber-resilience layer for measurement streams in distribution grids.

ABSTRACT. Time alignment governs the correctness of phasor and sampled value processing across modern substations: protection, measurement, and control functions rely on accurate time stamps to fuse data streams. When timing fails, systems may suffer loss of service or, more critically, misoperation. Although GNSS-based time synchronization is a well-known attack surface, empirical end-to-end evidence linking GNSS-layer timing manipulation to utility-grade protection misoperation remains scarce.

This paper presents, to the best of our knowledge, the first fully IEC 61850-compliant multi-vendor laboratory demonstration in which an RF-only GNSS-layer timing attack without substation-network access triggers false trips of transformer differential protection (ANSI 87T). We quantify consistency-gap durations (relay-independent exposure) and trip onset/success rates (relay-dependent realization) under commanded time steps up to ±10 ms across three MU pairings. We observe strong pairing- and sign-dependent behavior, including non-monotonic behavior for large backward steps, consistent with defensive blanking/blocking. Finally, we outline pragmatic countermeasures, including GNSS source hardening and interoperability improvements.

ABSTRACT. The ongoing digitization of energy networks is transforming traditional grid infrastructure and increasingly relies on standards such as IEC 61850 to enable interoperable and automated system operation. This opens new opportunities for the application of virtualization in operational technology (OT) networks which are used to manage and control primary and secondary equipment in electrical energy supply systems. By decoupling software functions from dedicated hardware, virtualization enhances the flexibility, scalability, and cost-efficiency of critical infrastructure management. In this paper, we propose a framework for the flexible virtual replication of OT networks to leverage these advantages. We introduce a use-case-independent process model that defines a standardized workflow for setting up and configuring the virtualized system. The proposed approach utilizes a Substation Configuration Description (SCD) file and container images with station-function code as primary inputs. To demonstrate the model's versatility, it is validated through three use cases that benefit from virtualization during the planning, implementation, and testing phases. Finally, specific test cases are evaluated for each use case, and a state-machine-based validation approach to ensure functional integrity is proposed.

ABSTRACT. Smart homes and smart devices are becoming more prevalent across millions of homes around the world. With the rise of AI, the smart home industry is quickly increasing its integration to manage common smart home tasks. However, existing work in large language models (LLMs) as agents within smart homes have shown minimal resilience due to limited environment scenarios or poor performance in complex tasks. We explore several strategies for LLMs as agents within the popular open-source software (OSS) smart home automation framework HomeAssistant to increase overall smart home resilience. Our approach combines traditional supervised learning techniques and optimized prompting as the core learning process. We use a diverse set of LLMs covering different levels of reasoning and costs in our optimization pipeline and evaluate their performance on a subset of the hardest tasks in a smart home benchmark. We include ReAct and Reflexion agentic paradigms and reveal how both provide marginal return on investment compared to fine-tuned LLMs for multi-device control within HomeAssistant but show promise in resilience tasks such as failure response.

ABSTRACT. Power grids are increasingly threatened by deliberate attacks, yet defense strategies are often evaluated under unrealistic deterministic assumptions that overlook key sources of uncertainty. This paper bridges such a critical gap by applying global sensitivity analysis (GSA), for the first time, to defender-attack-defender vulnerability models. The proposed GSA-augmented framework quantifies how uncertainties in operating conditions (e.g., renewable generation and load patterns) and attacker resources affect the worst load shedding that a defended system might experience. The framework is tested on IEEE 14-, 118-, and 200-bus systems. The results obtained show that the attacker budget consistently dominates the variability in the worst load shedding, with renewable generation contributing the second-largest share. By contrast, conventional one-at-a time sensitivity analysis and single-scenario worst-case analysis fail to capture these effects, hence leading to an incomplete picture of the defense strategy robustness. The framework can be easily replicated to different power system vulnerability studies, is scalable to large-scale applications across high-dimensional power grids (e.g., at EU level), delivers actionable guidance for prioritizing defense investments, and advances the quantification and communication of power grid vulnerability under realistic uncertainty. Finally, the paper advocates that GSA should become a standard component of power system vulnerability assessments.

ABSTRACT. Cold ironing is increasingly adopted in ports to reduce local emissions by powering docked vessels through shore-based electrical systems. However, its deployment introduces protection and monitoring challenges associated with the medium-voltage equipment, grounding constraints, renewable integration, and the limited availability of validated fault datasets. High-impedance faults (HIFs) are particularly difficult to detect because their low and irregular currents can appear very similar to normal operating conditions, allowing them to escape conventional protection devices. This paper proposes an unsupervised HIF detection framework for a cold-ironing microgrid using a Temporal Convolutional Network (TCN) autoencoder trained only on non-fault operating data. To enhance detection capability, a physics-informed regularization term is incorporated into the loss function, which encourages physically consistent reconstructions by penalizing zero-sequence imbalance. A synthetic dataset was generated in MATLAB/Simscape using a modified IEEE 9-bus microgrid configured for shore-to-ship operation, featuring renewable distributed energy resources and a stochastic arcing fault model. The proposed method was evaluated against a 59N residual-voltage indicator, Isolation Forest (IF), One-Class Support Vector Machine (OCSVM), and a Long Short-Term Memory (LSTM) autoencoder. The results show that the physics-informed TCN achieves an average segment-level Receiver Operating Characteristic Area Under the Curve (AUC) of 0.9989. This represents an improvement over both the standard TCN (0.9275) and traditional 59N indicator (0.5826), confirming the effectiveness of combining unsupervised sequence reconstruction with zero-sequence regularization for protection-focused monitoring in cold-ironing microgrids.