ABSTRACT Conventional power generation expansion planning (PGEP) models generally treat macroeconomic and environmental factors as exogenous inputs and are primarily used to optimize technology‐specific generation expansion schemes. Although effective for power‐system planning, their exogenous treatment limits their ability to support national energy strategies involving interactions between economic growth and environmental protection. To address this limitation, a multidisciplinary framework is proposed by coupling a top‐layer economic growth model with a bottom‐layer PGEP model. Unlike commonly used recursive‐dynamic, the top‐layer economic growth model is based on forward‐looking intertemporal optimization. It represents the aggregate equilibrium relationships among households, firms, government, and the environmental sector, thereby endogenizing economic growth, environmental expenditure, and the CO 2 emission long‐term pathway. A decomposition‐based intertemporal dynamic soft‐linking method is developed to connect this economic growth model with the PGEP model operating over a different time horizon. Environmental expenditure and electricity environmental efficiency serve as the coupling variables, enabling policy decisions to influence generation planning and power‐sector outcomes to feed back into the economic system. The proposed framework therefore allows policymakers to evaluate long‐term trade‐offs between economic growth and environmental protection while maintaining the technical feasibility of the generation plan. To demonstrate the effectiveness of the proposed methodology, a case study of China from 1990 to 2040 is implemented.
ABSTRACT Voltage instability imposes hard limits on transmission capacity in many power systems, constraining both real‐time operation and ex‐ante capacity allocations. This paper proposes a system integrity protection scheme (SIPS) based on VSC‐HVDC links in hybrid AC/DC power systems, designed to maximise loadability under stressed conditions and thereby increase secure transmission capacity. The core contribution is a novel emergency power control (EPC) strategy, derived in closed form, that considers the active–reactive power trade‐off when converter limits are reached. Two implementations are presented: one based on local measurements alone, and one augmented with synchrophasor data, both targeting the same objective of maximum loadability in the weakened grid. The proposed strategy is validated through dynamic simulations on a two‐node system and a larger benchmark transmission system. Beyond dynamic performance, the scheme is incorporated as a remedial action in the flow‐based (FB) capacity calculation framework, and an improvement to the FB methodology is proposed. Results show that the scheme can increase secure transmission capacity by up to 13%, depending on the initial operating point and proximity to other operational security limits.
ABSTRACT Smart protection methods are essential for modern AC microgrids, as they must be capable of detecting faults quickly, accurately, and reliably under different operating scenarios and the non‐linearity of the system. The proposed protection scheme is novel by combining the One‐Dimensional Sliding Window Median Filter (ODSWMF) with the Discrete Unscented Kalman Filter (DUKF) to achieve improved noise immunity, accurate nonlinear state estimation, and faster and more reliable fault detection and localisation than the existing AC microgrid protection systems. The ODSWMF estimates voltage and current signals from faulty buses with an estimation error of less than 5%, even under high noise conditions. The ODSWMF‐estimated current is processed by the DUKF to calculate the Total Harmonic Distortion (THD), enabling fault detection and classification with 99.9% detection accuracy and 99.8% classification accuracy across multiple fault types. Furthermore, DUKF computes reactive energy from voltage and current signals, and the direction trends of DUKF‐based reactive energy is utilised for fault localisation with very low error. The proposed strategy achieves a processing time of less than 5.2 ms, ensuring ultra‐fast protection. Validation on the CIGRE benchmark test network using MATLAB/Simulink 2023b confirms robust operation in both Grid‐Connected Mode (GCM) and Islanded Mode (IM) under noise measurement, load variations and network changes.
ABSTRACT Koopman operator theory offers a principled alternative to classical Lyapunov methods for power system transient stability analysis by lifting nonlinear dynamics into a linear framework. Existing finite‐dimensional Koopman approximations, including standard DMD with state observables and EDMD with user‐defined dictionaries, can suffer from truncation errors, dictionary sensitivity, and limited region of attraction (ROA) accuracy. To address these issues, this paper proposes a physics‐constrained deep Koopman (PCDK) framework that embeds power system differential algebraic equations into eigenfunction learning. Its loss penalizes the Koopman eigenfunction PDE residual through the Lie derivative, guiding the neural network towards stability‐relevant eigenfunctions consistent with physical constraints. A residual‐adaptive sampling strategy is also used to improve boundary resolution. On IEEE 9‐bus and 39‐bus benchmarks, case studies show that PCDK captures Lyapunov‐decaying eigenfunctions and estimates ROA boundaries with improved accuracy. A DFIG‐integrated case is further reported as a stress test, showing that single‐threshold calibration may be insufficient for conservative inner approximation under renewable‐generation dynamics. These results verify the algorithmic feasibility of PCDK and identify its limitations for future inverter‐dominated applications.
ABSTRACT The increasing integration of distributed energy resources (DERs) into distribution systems requires hosting‐capacity estimates that reflect both planning constraints and implementation‐level operating limits. This paper presents a two‐stage framework that combines optimisation‐based hosting‐capacity estimation with a centralised Hardware‐in‐the‐loop (HIL) setup. In Phase 1, an optimisation model estimates nodal hosting capacity and evaluates selected network‐reconfiguration options. In Phase 2, the candidate capacities are tested using an OPAL‐RT real‐time simulator, ABB's Smart Substation Control and Protection device (SSC600) and SMU615 merging units. Candidate capacities are accepted or rejected according to configured overvoltage, time‐dependent thermal overload, voltage total harmonic distortion (THD), and current total demand distortion (TDD) criteria. For Bus 22 under the original topology, the optimisation model estimated 3.07 MW, whereas the highest HIL‐accepted capacity under the tested conditions was 2.90 MW, corresponding to a 5.5% downward correction. Under the selected reconfigured topology, the optimization and HIL‐accepted values were 3.74 MW and 3.30 MW, respectively, corresponding to an 11.8% correction. Reconfiguration increased the optimisation‐derived capacity by 21.8% and the HIL‐accepted capacity by 13.8%. These results show that optimisation alone can overestimate the acceptable interconnection capacity when implementation‐level device responses are not represented. To the best of the authors’ knowledge, this study provides one of the first experimentally validated frameworks to integrate optimisation‐derived hosting‐capacity estimation with centralised HIL validation using commercial protection and control hardware for device‐level evaluation of voltage, thermal, and power‐quality limits.
ABSTRACT The rapid evolution of new power systems increasingly triggers rare and multi‐modal emergent operating states. While fine‐grained state identification is critical for grid resilience, it faces severe observability challenges under communication‐constrained conditions, forcing reliance on spatiotemporal voltage matrices. However, the rarity of these extreme events creates an acute small‐sample bottleneck, in which the true physical vulnerability boundaries are severely under‐sampled. Although generative techniques have been used to augment these scenarios, existing studies often emphasize generation fidelity rather than downstream physical utility. This paper investigates the mechanisms of real–synthetic data fusion by extracting a structurally consistent empirical saturation boundary. By introducing the spatiotemporal critical exposure density (SCED), we show that this performance ceiling is associated with physical manifold closure, where generative models progressively stabilize the boundary information embedded in real samples. Cross‐topology evaluations further reveal a state dilution phenomenon, indicating that the required synthetic budget is negatively correlated with the grid's critical‐state density. These findings provide a practical heuristic for sizing synthetic datasets and balancing computational costs in emergent state identification.
ABSTRACT The large‐scale grid integration of renewable energy sources, such as wind and photovoltaic generation, relies on power electronic converters, whose nonlinear and time‐varying characteristics introduce severe harmonic pollution and threaten power quality and grid stability. Accurate harmonic source modelling is essential for harmonic assessment, localisation, and mitigation. However, traditional static mechanism‐based models are unable to accurately capture dynamic coupling between harmonic voltages and currents, while purely data‐driven models suffer from limited interpretability and high data requirements. To address these issues, this paper proposes a mechanism–data hybrid‐driven dynamic coupling model for harmonic sources. First, a harmonic coupling admittance matrix is established from a mechanism‐based perspective, and a variable forgetting factor recursive least squares (VFF‐RLS) algorithm is applied for online dynamic identification to adapt to operating condition variations. Subsequently, a data‐driven model integrating LSTM and an iTransformer enhanced with differential multi‐head attention is proposed to capture nonlinear dynamics and stochastic disturbances that are inadequately represented by the mechanism model. Experimental validation using measured wind and photovoltaic data demonstrates that the proposed model achieves high accuracy and stability across different harmonic orders, effectively characterising dynamic cross‐frequency coupling between harmonic currents and harmonic voltages.
ABSTRACT High renewable energy integration sharpens the power system net load curve and triggers frequent ramp events, making sufficient ramping capability critical for ultra‐short‐term power supply–demand balance. As vital flexible resources, electric vehicles (EVs) can effectively assist grid regulation and renewable accommodation, yet conventional approaches fail to assess the adjustable charging capacity of large‐scale EV fleets accurately. This paper proposes a novel adjustable capacity evaluation method for massive EVs participating in the system ramping service market. First, it quantifies system ramping demand considering net load variations and renewable prediction errors, and calculates the inherent ramping capability of conventional units via traditional power market clearing results. A Gaussian mixture distribution is adopted to characterise multiple EV uncertainties, including user behaviours, battery parameters and charge–discharge rates, while the Logit model predicts EV users’ scheduling willingness. The Minkowski metric is utilised to evaluate the aggregated ramping potential of large‐scale EV clusters. Furthermore, a feasible region optimisation model with power–energy coupling constraints is established to forecast the charging potential of EV charging stations, and robust optimisation is applied to optimise key parameters. Case studies based on practical provincial power data verify the accuracy and effectiveness of the proposed method.
ABSTRACT Accurate fault location (FL) in hybrid transmission lines, consisting of non‐homogeneous sections, is challenging due to the differing electrical characteristics of overhead lines and underground cables, which lead to signal distortion, attenuation, and reflections. Such lines complicate the application of traveling wave (TW)‐based methods due to varying wave propagation speeds and reflections at junctions. To address this issue, this paper introduces a novel TW‐based technique that leverages frequency‐domain analysis rather than traditional time‐domain methods. Given that TWs are inherently tied to both distance and frequency, the proposed approach analyzes the frequency content (FC) of the TW's wavefront to determine the FL. Specifically, the method captures the initial TW at both ends of the line, applies Clarke's transform for phase decoupling, performs a fast Fourier transform to analyze the FC, and processes the data through neural networks to accurately determine the faulty section of a hybrid line and the estimated FL. This technique eliminates the need for time synchronization or propagation velocity assumptions. Extensive PSCAD/EMTDC simulations confirm the method's high accuracy across various fault conditions, representing a significant advancement in FL for hybrid lines in modern power grids.
ABSTRACT High penetration of residential photovoltaic (PV) units has caused severe voltage issues in the distribution feeders. During seasons with low demand and high PV generation, distribution feeders face increasing power quality issues. Continuous operation in system conditions with these voltage issues reduces the lifetime of distribution equipment and can cause severe monetary loss. Legacy VAr support devices are too sparse and slow to mitigate these issues. With the introduction of the IEEE 1547–2018 standard, distributed energy resources (DERs) are allowed to participate in providing VAr support to mitigate these issues. This paper proposes a novel unified mode selection (UMS) framework within a real‐time VAr optimization tool for distribution networks. This tool is developed based on an effective unbalanced distribution AC optimal power flow (ACOPF) formulation. The proposed model has the ability to select between the Volt‐VAr and Watt‐VAr operational modes for each VAr enabled smart inverters, and to simultaneously optimize the controller set‐points to optimally dispatch each inverter, minimizing operational costs and mitigating the voltage issues. The tool is tested on a snapshot of a distribution network in Arizona and compared against using only Volt‐VAr or Watt‐VAr operational mode with default settings. The results confirm the effectiveness of the model.
ABSTRACT With the rapid expansion of the global hydrogen industry, large‐scale integration of green electricity‐hydrogen systems into distribution networks has enhanced renewable energy consumption while introducing operational challenges such as increased line loading and congestion. Nevertheless, studies on the integrated hosting capacity of hydrogen systems in distribution networks remain scarce. Existing hosting capacity definitions typically only consider physical operational limits of the distribution network, which yields excessively idealised results. To address this problem, this paper develops an external characteristic model of the green electricity‐hydrogen system based on the energy hub concept, which is incorporated as a special node in distribution network optimisation. Subsequently, the maximum hydrogen load hosting capacity is proposed as an indicator for evaluating the integrated hosting capacity of green power‐hydrogen systems. Furthermore, an integrated evaluation framework is established, jointly considering economic efficiency, operational flexibility and security. In addition, the impacts of renewable energy integration and electrolyser deployment on the integrated hosting capacity of hydrogen systems in distribution networks are analysed. The proposed method is validated using an improved IEEE 33 distribution system and recommendations are provided to enhance the integrated hosting capacity of hydrogen systems in distribution networks.
ABSTRACT Accurate forecasting of power generation is essential for ensuring the reliable and cost‐efficient operation of modern electrical grids. This paper introduces an optimized wavelet temporal convolutional network (TCN) designed for multi‐step‐ahead prediction of thermal power generation in Brazil. The proposed model, called optimized‐wavelet‐TCN, integrates a discrete wavelet transform denoising stage to effectively filter high‐frequency noise from operational data while preserving underlying signal trends, followed by a TCN optimized via a tree‐structured Parzen estimator for hyperparameter tuning. Evaluated on real‐world hourly data spanning multiple years, the optimized‐wavelet‐TCN achieves superior forecasting performance with a mean absolute percentage error of 10.2%, 7.24%, and 8.54% for 15, 30, and 60‐step ahead predictions, respectively. The model significantly reduces computational time to only 1.54 s for a 30‐step forecast, markedly outperforming state‐of‐the‐art transformer‐based and recurrent neural network baselines in both accuracy and efficiency.
ABSTRACT Hybrid photovoltaic and wind microgrids are increasingly deployed for sustainable electrification but face challenges due to resource intermittency and storage limitations. Integrating batteries with hydrogen storage offers complementary short‐ and long‐term energy buffering to enhance system reliability. This paper proposes a hierarchical hybrid power‐management (HHPM) strategy based on fuzzy logic control (FLC) for maximum power point tracking (MPPT), a power‐sharing formulation for coordinating the battery and hydrogen subsystems, and a fuzzy supervisory controller for selecting the system operating modes according to renewable availability and storage conditions. To strengthen the validation of the proposed approach, an additional comparison with an adaptive power‐sharing strategy based on the battery state of charge is presented. In this strategy, the allocation coefficient α adap is continuously updated according to the ratio between the instantaneous battery SOC and its maximum value. Renewable generation, load demand and hydrogen‐storage level are monitored separately by the supervisory controller to determine the instantaneous power balance and enforce the operating constraints of the battery, electrolyser and fuel cell. Simulations using measured Mediterranean climatic data demonstrate improved renewable‐power extraction, with gains of up to 443 W, balanced utilisation of the storage subsystems and stable DC‐bus operation with a maximum voltage deviation of 1.25%. The results indicate that the proposed control strategy enhances the operational flexibility and reliability of the standalone microgrid under variable climatic conditions.
ABSTRACT Cable joints are among the most failure‐prone components in distribution cable systems, making their condition monitoring essential for reliable operation and maintenance in distribution networks. Existing discharge‐based monitoring methods typically provide binary defect identification, leaving degradation‐stage information in discharge observations underexploited and limiting their support for fine‐grained risk‐informed maintenance under system‐level impacts. To address this gap, this paper proposes a stage‐aware degradation modelling paradigm for cable joints based on ordered degradation discharge event sequences. A forward‐labelling strategy is introduced to formulate degradation evolution as a supervised learning problem, and a conditional degradation‐to‐failure propensity model (CDPM) is developed to map ordered degradation discharge event sequences to a probability‐related failure‐tendency score for the next degradation event. The failure‐tendency score is further integrated with post‐fault consequence severity to quantify system‐level risk and support maintenance prioritisation in distribution networks. The proposed CDPM and risk quantification framework are evaluated using full‐scale cable‐joint degradation experiments and an improved 62‐bus distribution network.
ABSTRACT With the increasing penetration of renewable energy, the demand for flexibility in power systems has grown significantly. However, existing studies typically model individual types of distributed energy resources (DERs) in isolation, lacking a unified representation of their heterogeneous power–energy characteristics and response mechanisms. This limitation hinders the effective aggregation of flexibility and restricts the full realisation of DERs’ value in electricity markets. To address this issue, this paper starts from the intrinsic formation mechanisms of flexibility and develops a unified aggregation modelling framework for distributed energy storage, flexible loads and quasi‐energy storage (e.g., a crusher with warehousing in ceramic production) resources. The flexibility of DERs is characterised along three key dimensions, namely maximum capacity, response speed and response duration. Based on this representation, an aggregation model is proposed to achieve a unified and tractable description of heterogeneous DERs’ flexibility. Furthermore, considering the coupling relationships among different electricity markets, an optimal scheduling model for DERs participation in multi‐market trading is established. Case studies demonstrate that the proposed approach achieves high utilisation levels for distributed energy storage (average 49.17%–56.80% SOC) and quasi energy storage (average 46.73%–60.79% warehousing state), while delivering annual total revenues of 71.90 million CNY across coupled multi‐electricity markets.
ABSTRACT For variable‐speed pumped‐storage units with full power converters (FPC‐VSPSUs), converter‐based decoupling control fundamentally reshapes the machine‐terminal measured impedance trajectory during a loss‐of‐excitation (LOE) fault in the excitation system (ES). Consequently, conventional machine‐terminal measured‐impedance‐based protection schemes may suffer from large blind zones and a high risk of failure to trip. To address this issue, this paper presents a comprehensive analysis of the post‐LOE impedance characteristics under two field‐orientated control (FOC) strategies: stator‐flux orientation (SFO) and rotor‐flux orientation (RFO). Under stator‐side unity‐power‐factor operation, it is revealed that, for both FOC strategies, the machine‐terminal measured impedance trajectory moves along or near the real axis during the early LOE stage. Based on this characteristic, an adaptive rectangular LOE protection scheme is proposed, in which the setting boundaries are dynamically adjusted by integrating multidimensional operating‐state information, including rotational speed, operating condition, and FOC strategy. PSCAD simulations and RTDS hardware‐in‐the‐loop (HIL) experiments with an industrial protection relay demonstrate rapid and reliable fault identification over a wide operating range, including low‐active‐power generation conditions, while avoiding maloperation during grid disturbances.
ABSTRACT In power systems with high penetration of renewable energy, fault currents exhibit limited amplitude and controlled phase, which degrades the performance of conventional power‐frequency protection for outgoing lines. To address this issue, this paper proposes a two‐stage transient current differential protection method that accounts for the influence of the fault inception angle. First, an analytical model of the fault transient current is established based on the distributed parameter line model and the compositional differences between the high‐ and low‐frequency components of the differential current under internal and external faults are theoretically analysed. Then, the impact of the fault inception angle on the transient high‐ and low‐frequency currents is examined, revealing the physical mechanism responsible for the distortion of low‐frequency feature extraction under specific time windows and certain inception angles. On this basis, the complete ensemble empirical mode decomposition (CEEMD) algorithm is employed to decompose the differential current into high‐ and low‐frequency modes and the fault inception angle is identified in real time using the steady‐state voltage zero‐crossing information. A two‐stage time‐window protection scheme utilizing fault inception angle identification is thus developed. PSCAD/EMTDC simulation results demonstrate that the proposed scheme effectively avoids the adverse effect of varying fault inception angles on the frequency‐band feature extraction of transient currents, thereby significantly improving the reliability of transient current differential protection. The scheme remains sensitive and reliable under high‐resistance grounding faults and strong electromagnetic noise interference. RTDS simulation results verify the effectiveness and practical applicability of the proposed protection method.
Accurate fault classification and location in transmission lines plays a critical role in ensuring the reliability, stability, and efficiency of the power system. Accurate fault detection and classification is key to safety, enabling brisk resolution of problems and mitigating electrical power supply disturbances. Traditional methods for Fault Type Classification (FTC) and Location Prediction (LP) struggle to handle real-world fault situations, facing challenges like delays, limited capacity, and vulnerability to single-point failures. This study develops and compares various Machine Learning (ML) and Deep Learning (DL) algorithms along with Artificial Neural Network (ANN) to classify transmission line faults — LG, LLG, LL, and Three-Phase faults — and predict fault locations. Fault data was generated using a MATLAB Simulink model incorporating phase voltages, currents, phase angles, zero-sequence, and negative-sequence components from both sending and receiving ends. The method achieved 99.69% and 99.78% accuracy using Random Forest and CatBoost for FTC, and R² of 1.00, MAE of 0.0080, and RMSE of 0.0729 for LP. Noise robustness was validated across SNR values (0–50 dB), achieving 90.53% FTC accuracy at 50 dB by CatBoost and R² of 0.9999 by Random Forest in LP. The study demonstrates exceptional performance of ML and DL models over traditional methods.
The difference in time constants between electric power systems and natural gas networks poses a fundamental challenge to the reliability assessment of integrated electricity and gas systems (IEGS). While steady-state approximations fail to capture critical buffering effects, dynamic solvers often suffer from computational intractability and numerical instability under discontinuous fault conditions, hindering further reliability assessment. To resolve this, this paper presents a reliability assessment framework that integrates high-accuracy dynamic modelling and efficient simulation. We first introduce an approximate analytical method (AAM) that overcomes the limitations of traditional spatial discretisation by solving nonlinear partial differential-algebraic equations of the IEGS. By coupling this solver with subset simulation (SS), the proposed approach effectively overcomes the curse of dimensionality, allowing for the rapid and statistically accurate estimation of failure events in the IEGS. Case studies verified the superiority of the AAM over discretisation-based numerical methods and highlighted the efficiency advantages of the SS algorithm for reliability assessment.
The intermittent nature and stochastic variability inherent in renewable energy generation present significant challenges to frequency regulation in active distribution network and microgrid (ADN-MG) systems. The frequency regulation capability of an ADN-MG system can be enhanced by exploring the potential of complementary intermittent power outputs of heterogeneous distributed generators (DGs). Accordingly, an ensemble equivalent model is established to analyse the frequency domain of the ADN-MG system. The frequency response characteristics are systematically investigated through three operational scenarios, and the parameter related stability condition is established. Compared with conventional frequency regulation results, the ADN-MG system can be controlled with less state of charge for the considered energy storage units. Furthermore, the frequency domain of the ADN-MG system is estimated from the perspective of parametric analysis, which can be implemented by equivalent transfer function-based parameter tuning in actual engineering applications. Then a theoretical guideline is provided for parameter selection, ensuring the frequency deviation of the ADN-MG system remains within an acceptable range. Finally, several indexes of evaluative indicators-frequency nadir, rate of change of frequency and regulation time are introduced for quantitative analysis to further demonstrate the superiority of the proposed results. The effectiveness of frequency ensemble analysis results are validated by three examples of ADN-MG systems with heterogeneous DGs.
