ABSTRACT In order to reduce carbon dioxide emissions, traditional coal‐fired power generation has been strictly limited in China recently, which has led to the problem of reserved capacity shortage. As one of the critical east provinces of China, Zhejiang province also suffers from the shortage of reserved capacity, whereas gas power generation is commonly utilised to mitigate this problem. However, the gas generator dispatch plan is still made manually at the current stage, which is quite time‐consuming and has the space to optimise. Therefore, this paper first proposes a practical day‐ahead gas generator dispatch optimisation (GGDO) model to maximise the daily minimum reserve capacity. Besides, several constraints related to the system operation and natural gas limitations are given. Finally, the data of three typical days in Zhejiang province are employed for demonstration, which shows the strength of the proposed GGDO model.
ABSTRACT Freezing rain and snow weather is likely to cause ice accretion on overhead transmission lines and threaten power grid security and social stability. Currently, ice‐melting technology using DC current is the most effective active de‐icing method which has been widely applied in China. This paper summarises the development history and engineering application status of DC ice‐melting devices (DCIMD) in China, analyses the principles, advantages, disadvantages and adaptability of mainstream topologies and discusses the key problems and corresponding improvement directions. This research indicates that, a sophisticated DC ice‐melting technology system has been established in China, which significantly enhancing the safety resilience of power systems. New type of DCIMD adopting fully‐controlled power electronics, which has been exhibited technical advantages, represents important development direction adaptable to the construction of new power systems. At present, the key technical bottleneck urgently to be solved is the implementation of the one‐key sequential control function, which may significantly improve response speed and operational efficiency of ice‐melting.
ABSTRACT This paper presents a deep reinforcement learning (DRL)‐based control framework for a bidirectional dual‐input single‐output (BDISO) converter applied to low‐power electric vehicles (EVs). The study addresses the dynamic instability introduced by constant power loads (CPLs), which emulate the negative impedance behaviour of EV motor drives. A twin delayed deep deterministic policy gradient (TD3) algorithm is employed to directly generate control signals, thereby eliminating reliance on explicit converter models and intermediate control layers. The proposed model‐free approach enables fast voltage regulation and stable bidirectional power flow between sources and load, particularly under regenerative operating conditions. Simulation results obtained in MATLAB/Simulink demonstrate that the proposed strategy achieves up to 4–7 times faster settling time and 30%–90% lower voltage error compared with conventional proportional‐integral (PI) and super‐twisting algorithm (STA) controllers. These findings establish DRL‐based TD3 control as a scalable, model‐free solution for robust and efficient energy management in next‐generation EV powertrains.
ABSTRACT Accurately quantifying the carbon emission reduction (CER) of renewable energy (RE) is crucial for the coupled electricity‐carbon market. However, traditional methods based on regional average emission factors ignore topological constraints and transmission losses causing systematic biases in evaluating RE's environmental value. To address this, this paper proposes a high‐precision physics‐based CER accounting framework integrating topological sensitivity mapping with bidirectional network loss allocation. First, an improved generation shift distribution factor (GSDF)‐based marginal emission factor model is established to overcome geographical blindness and precisely identify specific high‐carbon marginal units displaced by RE. Second, upstream and downstream power flow tracing algorithms are utilised to bidirectionally allocate network losses, equitably attributing carbon emissions embedded in dissipated energy. Third, a virtual lossless network is reconstructed and RE is treated as a negative virtual load, through which the precise displacement pathways from marginal units to RE nodes are rigorously mapped. Case studies on a regional power grid in Zhejiang, China, demonstrate that the proposed framework successfully corrects systematic underestimations providing a highly credible pricing benchmark for green certificates and carbon assets.
ABSTRACT This paper addresses the problem of joint detection and estimation of cyberattacks and physical faults in smart‐grid cyber‐physical systems. In practical smart‐grid environments, physical faults arising from equipment degradation may coexist with malicious cyber threats, leading to significant degradation in system reliability and operational security. To tackle these issues, a model‐based resilient estimation and control framework is proposed, which integrates observer‐based attack detection and fault estimation mechanisms. First, a set of distributed local estimators based on unknown input observers (UIOs) is developed to achieve robust state and fault estimation under disturbances and sparse sensor attacks. Subsequently, a resilient global fusion strategy is established to enhance tolerance against compromised measurements and to ensure bounded estimation errors even in the presence of sensor corruption. Based on the fused estimation results, a resilient fault‐tolerant control scheme is formulated to guarantee closed‐loop stability and desired performance. The proposed framework is applicable to both centralised and distributed smart‐grid configurations. Simulation results on a nonlinear benchmark system demonstrate its effectiveness in detecting cyberattacks, estimating physical faults, and maintaining reliable system operation under adverse conditions.
ABSTRACT To achieve the Vision 2030 target of raising renewable energy capacity to 50% of total installed generation, large‐scale solar deployment has been identified as a core strategic pathway; however, the pronounced variability of solar generation poses significant challenges to grid stability. This paper systematically examines the current status of the Saudi Arabia's power system and identifies four major challenges: (i) renewable output variability and its temporal mismatch with load demand, (ii) the continuous decline of system inertia, (iii) economic bottlenecks for long‐duration energy storage and (iv) increasing complexity in regional coordination. Building on this analysis, the study focuses on compressed air energy storage (CAES) and demonstrates its distinctive advantages in providing long‐duration energy storage as a system‐level solution to mitigate the variability of renewable generation. Considering Saudi Arabia's high solar irradiance, temperature‐driven load characteristics and grid structure, this paper proposes an integrated New Energy + Energy Storage configuration and operational approach, through which intermittent renewable electricity can be converted into stable, reliable and dispatchable high‐quality power. The findings provide a technically grounded pathway to improve system flexibility, secure electricity supply and facilitate large‐scale renewable integration during Saudi Arabia's energy transition.
ABSTRACT To coordinate local green power absorption and key section power flow stability in high‐penetration renewable grids, this paper proposes a source–grid coordination strategy integrating green power tracing and DC power flow sensitivity. A green power distribution matrix with a path priority coefficient is built to quantify green power supply paths and prioritise absorption, and a DC power flow sensitivity matrix is established for accurate grid regulation. A dual‐objective optimisation model (minimising power flow volatility, maximising green power absorption) is formulated and solved via convex optimisation. Year‐round simulations on the IEEE 11‐bus system and Yancheng's actual grid verify the strategy effectively suppresses power flow volatility, raises green power absorption rate, resolves the dilemma of curtailing renewables for stability and offers a new optimisation method for high‐renewable grids.
ABSTRACT To address the challenges of insufficient system flexibility, fuel shortages and the lack of rotational inertia under high renewable energy penetration, this study develops an 8760‐h chronological optimisation model for an integrated wind‐PV‐CSP system. A full‐factorial search method is proposed to explore the three‐dimensional design space of wind capacity, PV capacity and solar‐field size, and a multiparameter sensitivity analysis is conducted to identify key factors affecting system performance. The results show that, under given grid‐connection limits and CSP turbine capacity constraints, the integrated system exhibits a clear optimal configuration structure. Sensitivity analysis further reveals the marginal contribution patterns of critical parameters, such as power‐to‐heat capacity, thermal storage duration, battery storage capacity and peak‐period weighting windows, indicating that appropriate parameter selection enables a balanced trade‐off among economic efficiency, peak‐period supply capability and energy utilisation. The proposed modelling framework and optimisation approach demonstrate that integrated wind–PV–CSP systems can achieve economic, efficient and stable clean‐energy utilisation in regions with high solar irradiance, limited fuel supply and insufficient grid flexibility, providing valuable technical references for regional generation planning, feasibility studies and energy transition policymaking.
Multi‐step ahead forecasting of power load is crucial for optimising power system scheduling and participating in energy market transactions. However, existing forecasting methods often suffer from issues such as cumulative prediction errors and insufficient modelling of sequence dependencies. To solve the above problems, a multi‐step ahead forecasting method based on multiplexed convolutional neural networks (MCNN) and multi‐gate mixture of long short‐term memory networks (MMoL). First, multi‐branch convolution is adopted to construct independent feature spaces for different levels of power loads, achieving multi‐scale feature fusion to enhance the representation ability of the input samples. Next, the multi‐step prediction task is transformed into a multi‐task joint optimisation problem. Multiple independent LSTMs are used as shared experts, and task‐specific gating units are utilised to dynamically learn the optimal combination of expert models for each future time step, achieving more refined time‐series feature modelling. Finally, comparative experiments are conducted based on two real‐world datasets. The results show that the proposed model exhibits better accuracy and robustness.
ABSTRACT The lack of active power support capability in power systems with large‐scale renewable energy integration has become a significant constraint on stability. This paper proposes a variable speed condenser (VSC) control strategy based on a doubly fed machine. The rotating rotor shaft can provide inertia support for the system and participate in primary frequency regulation of wind farms or PV stations by adjusting the rotor speed. The stator of the VSC is directly connected to the grid. During short‐circuit faults, the principle of flux linkage conservation can provide short‐circuit current support for the system, suppressing transient overvoltage phenomena. Meanwhile, the fully controlled power electronic converter can provide rapid voltage regulation capability for the system. Simulation and field test results indicate that the VSC can provide stable inertia support, short‐circuit current support, primary frequency regulation and rapid voltage regulation. It can be installed in PV stations and wind farms to enhance the grid‐connection performance of renewable energy power plants.
Integrated source–grid–load–storage (SGLS) systems with high renewable energy penetration face dispatch challenges, including multi-factor coordination, renewable uncertainties and multiple policy constraints. Traditional deterministic optimisation struggles to handle the uncertainty of wind and solar power, whereas complex chance-constrained programming often suffers from low computational efficiency and poor practicality. To address these challenges, this paper proposes an adaptive chance-constrained hybrid second-order cone programming (SOCP)-based AC optimal power flow (ACOPF), designed for the coordinated optimisation of SGLS systems. The model incorporates a simplified chance-constrained method that exploits the structural characteristics of SGLS systems, substantially lowering computational complexity. An iterative mechanism combining external power flow verification and dynamic constraint augmentation ensures solution robustness while avoiding the dimensionality curse caused by embedding numerous scenarios. Additionally, the algorithm features self-adaptability. When the original problem is infeasible due to constraint conflicts, it automatically relaxes chance constraints into the objective function to ensure solvability. The main contribution lies in establishing a coordinated dispatch framework that balances computational efficiency, reliability and practicality while incorporating industrial load demand response characteristics and relevant policy constraints. Simulations demonstrate that the model rapidly achieves cost-effective and robust scheduling by converging to exact global optima.
Because of the inherent randomness and variability of renewable energy, their capacity cannot be treated equally to conventional power generation units in power system planning. Experts and scholars in the power industry have proposed the concept of credible capacity for renewable energy. System-friendly renewable energy power stations adopting the ‘renewable energy + energy storage + smart regulation’ mode can enhance their credible capacity. This study proposes a practical method for calculating the credible capacity of system-friendly renewable energy power stations. Firstly, based on manual standards and practical planning, the definition of credible capacity is established. Secondly, by analysing renewable energy characteristics (credible capacity, daily guaranteed electricity) and energy storage characteristics (output characteristics, stored electricity characteristics), calculation methods and formulas for determining the credible capacity of system-friendly renewable energy power stations are developed. Finally, practical case studies reveal that: energy storage capacity constitutes a limiting factor for improving the credible capacity of system-friendly renewable energy stations. When sufficient energy storage capacity is available, the daily electricity generation from renewable energy sources becomes the determining factor for system-friendly renewable energy stations reaching their credible capacity saturation point.
Micro-integrated energy systems, as a small-scale, distributed energy supply system that integrates multiple energy technologies, is one of the most promising application directions for artificial intelligence. Similar to robots, micro-integrated energy systems have strong sensing, decision-making and control capabilities through their large number of installed sensing units, energy management units and control devices, which are sufficient to constitute an atypical broad-sense embodied intelligent system in terms of hardware and software resources. The present paper proposes a systematic framework of embodied intelligence for micro-integrated energy systems, which is discussed in the context of its potential applications in embodied sensing, decision-making and execution within micro-integrated energy systems. The paper then proceeds to analyse the inspiration of embodied intelligence in the design of typical tasks, include monitoring and energy-saving optimisation, emergency fault handling and load shifting, and self-construction and self-maintenance, providing references for the construction of the next-generation intelligent micro-integrated energy system for use in ocean exploration, deep space exploration, polar research and other missions.
Deep reinforcement learning (DRL) has become a promising approach for electric vehicle (EV) charging scheduling. However, its practical deployment poses potential risks to power infrastructure. DRL relies on trial-and-error interactions during training to approximate optimal policies, which may lead to unsafe decisions. To address this, a novel framework called dual-layer safety modules for EV charging scheduling (DuMES) is proposed. This framework introduces a decision-level safety layer into the conventional DRL architecture that adaptively detects and replaces unsafe actions. Furthermore, by integrating dual safety layers with reward shaping, the framework promotes convergence between raw and safe actions. This enhances training efficiency while ensuring power system stability during both training and deployment phases. The method was evaluated through simulation experiments on a charging station equipped with renewable energy and energy storage system (ESS). Comparative analyses with baseline methods demonstrate that DuMES effectively satisfies user charging demands, reduces operational costs and ensures compliance with safety constraints.
To address the retrofitting and upgrading needs of existing combined cooling, heating and power multi-energy systems (CCHP-MES) and meet the modular construction requirements of future CCHP-MES, this paper proposes an innovative expansion configuration optimisation method for CCHP-MES considering tracking ability improvement based on scheduling degree of freedom. Firstly, according to the directed graph theory, a scheduling degree of freedom (SDOF) determination method is defined to verify the structural reasonableness of CCHP-MES and assist in expansion equipment selection. For two types of operational scenarios—load demand expansion and load demand type expansion—the tracking-economy multi-objective expansion configuration optimisation models considering SDOF are constructed, and different expansion schemes are evaluated based on comprehensive evaluation of system tracking, economy and environment. The efficacy of the planning model is verified using typical load demand and renewable energy data in Nanjing as an example. The results show that the proposed building block expansion method can achieve better tracking performance improvement with relatively small economic loss and also enhance the renewable energy share in total energy consumption of CCHP-MES to achieve the ultimate carbon neutrality goal.
To enhance stability and reliability of an electric distribution system, the monitoring and data acquisition through measurement devices, sensors and communication networks is essential to maintain the system observability for proper operation and control. However, insufficient measurement devices, and malfunctions of sensors and communication networks may lead to the monitoring data losses, one solution is to use the synthetic datasets, known as pseudo-measurements, to substitute missing data and improve the system observability. Pseudo-measurements are created through probabilistic, statistical and machine learning techniques using historical measurements of distribution systems. In this paper, potential roles of pseudo-measurements to enhance monitoring of distribution networks have been reviewed. Two categories of pseudo-measurement models are examined in this review: (1) probabilistic and statistical-based models, including parametric, semiparametric and nonparametric approaches; and (2) machine learning-based models, including shallow (conventional machine learning) and deep learning structures. Each model's computational demands and practical applications are analysed, highlighting their advantages and limitations. This review aims to identify the research gaps of pseudo-measurement models and suggest future research directions for robust and adaptive monitoring of distribution networks.
Transcritical CO 2 heat pump systems integrated with renewable energy sources and energy storage are being paid great attention to develop sustainable energy and energy savings in civil and industrial applications so as to achieve net-zero carbon emissions by 2050. This paper presents a comprehensive review on the progress in research and technology development of transcritical CO 2 heat pump technology with clean energy and energy storage systems. Transcritical CO 2 heat pump systems are very competitive with conventional systems for space heating, hot water production, air-conditioning, waste heat recovery and other engineering applications. This review focuses on the recent advances in the key research and technology development of transcritical CO 2 heat pump integrated energy systems including sustainable ‘green’ heating, energy storage, electro-thermal storage, waste heat recovery and disruptive electrification of renewable energy, combined heating and power, and new hybrid systems using transcritical CO 2 heat pump systems. The electrification of heat is a key opportunity for industry and is crucial for tackling the climate emergency. Future research needs of transcritical CO 2 heat pump integrated systems are identified according to this comprehensive review. Furthermore, Perspectives and deployment of transcritical CO 2 heat pump systems integrated with renewable energy sources and energy storage technology are discussed.
The increasing penetration of renewable energy and growth of flexible loads have introduced considerable uncertainty and complexity into integrated energy systems (IES). To address these challenges, this study proposes a multi-timescale scenario-driven hierarchical capacity configuration framework in which multi-source data fusion is employed to improve data quality, extract joint features and generate representative scenarios for system optimisation. First, a variable-order bidirectional Markov interpolation approach based on time-segmented probability models is developed, incorporating anomaly probability information to reconstruct missing data. Then, multi-source data fusion of random variables is achieved through time- and frequency-domain feature extraction, principal component-based dimensionality reduction and mini-batch clustering, thereby generating representative scenarios for IES configuration. On this basis, a hierarchical capacity configuration framework is established. The upper level optimises equipment selection and installed capacity at the daily scale with objectives of investment benefit and system reliability, whereas the lower level refines these decisions at the hourly scale with objectives of environmental performance and operational economy. The multi-objective optimisation problem is solved using nondominated sorting genetic algorithm-III (NSGA-III). Simulation results indicate that the proposed method significantly improves economic efficiency, environmental performance and reliability of IES by leveraging representative scenarios that accurately capture both uncertainties and dynamic characteristics.
In the context of intensified global climate change, the establishment of a new-type power system dominated by renewable energy sources is a core pathway to achieving the ‘dual carbon’ goals; however, its reliance on meteorological services has significantly increased. This paper systematically reviews the meteorological demands of the new-type power system in areas while identifying three major technical challenges in current electric meteorological services. Firstly, there are blind spots in the meteorological monitoring network covering renewable energy power plants and transmission lines, leading to insufficient acquisition of micro-meteorological data such as icing and snow cover under extreme conditions. Secondly, traditional numerical weather prediction has limited accuracy in forecasting power-related meteorological factors at high spatial and temporal resolutions and poor interpretability of extreme weather events. Thirdly, the depth of data fusion between meteorology and power systems is insufficient, making it difficult to support customised services tailored to the actual operational needs of power grids. To address these challenges, this paper proposes future technical development directions, including constructing a ‘satellite-radiosonde-ground’ integrated monitoring system to enhance full-element perception capabilities in special regions such as deserts, Gobi areas and high-altitude zones; the development of a multimodal forecasting approach that combines physical mechanisms with AI large language models and improves the spatial and temporal resolution of renewable energy power forecasting and grid fault warnings under extreme weather conditions; and creating specialised meteorological service products for power systems, deeply integrating meteorological data with grid dispatch and energy storage optimisation algorithms to achieve flexible load response and dynamic market risk management, thereby providing critical support for the large-scale integration of renewable energy and the ‘dual carbon’ goals.
In recent years, AI large models, also known as large pre-trained models or foundational models, have achieved remarkable success in various tasks across multiple domains. These models leverage extensive unlabelled datasets from multiple fields and modalities, enabling them to generalize across tasks with minimal labelled data. Their ability has led to advancements in numerous domains. However, the application of large models in power systems remains in their early stages, and the potential of large models has not been fully explored. This paper aims to help researchers and engineers grasp the latest advances and trends in large models to foster the development and applications in the power industry. It traces the development stages of large models, introduces the concept and architecture of large models, and concludes the verified and remarkable capabilities of the large model. Additionally, by integrating existing research, this paper reviews recent advancements and potential applications of large models in power systems, with a focus on perception, planning, and control. Moreover, it summarizes the key enhancement technologies for optimizing the effectiveness of large models of power systems. Finally, the challenges and risks associated with developing large models including computing power requirements, reliability, and safety considerations for power systems are also discussed. Based on the survey, large models for power systems are proven to be a promising paradigm that can improve the efficiency and effectiveness of intelligent power systems, which contributes to the reference for the intelligent development of the power industry.
