6 November 2025 in IEEE Transactions on Smart Grid
The digitalization of distribution networks intensifies the coupling between cyber and physical networks. Edge computing devices offer significant flexibility for organizing distributed energy resources and enabling multi-area coordinated control, demonstrating considerable potential to enhance the power restoration capability of active distribution networks. This paper proposes a power restoration method based on edge computing technology, which accounts for the interdependencies between cyber and physical systems during the restoration process. For the cyber system, a self-organization method for edge computing devices is proposed to establish a dual-layer emergency communication network, supporting strategy generation for power restoration. For the physical system, an ADMM-based solving method is developed to achieve rapid solution of the restoration problem in a distributed manner by pre-fixing binary variables. Case studies are conducted on the IEEE 123-node test system and the Taiwan 94-node test system, highlighting the advantages of the proposed method in terms of restoration speed and effectiveness.
21 May 2025 in IEEE Transactions on Instrumentation and Measurement
With the high penetration of distributed generators (DGs), state estimation is essential for fast and accurate operational status tracking of active distribution networks (ADNs). However, there are generally unobservable areas in ADNs in ADNs due to the paucity of measurements. Traditional state estimation methods may encounter difficulty coping with limited measurements. Thus, aiming at the accurate and fast state perception of unobservable ADNs, this paper proposes a randomly switched subsystem-based state estimation method. First, the unobservable ADN is modeled as the randomly switched ADN (RSADN). Stochastic observability is defined to distinguish the unobservable conditions, which ensures the feasibility of state estimation. Further, considering different operational conditions, the observability enhancement mechanism is designed under limited measurement conditions to make the network observable and reduce the impact of topology switching on the system. Then, the improved unscented Kalman filter algorithm is utilized to solve the state estimation problem based on state transformation. The results of case studies confirm that the proposed method can achieve more accurate estimation results compared to other methods under unobservable conditions and exhibit satisfactory computational speed in large-scale systems.
9 February 2026 in IEEE Transactions on Transportation Electrification
Developing advanced fast-charging methodologies for lithium-ion batteries (LIBs) is critically important for electric vehicles (EVs), yet remains challenging due to accelerated degradation modes including lithium deposition and fracture of electrode particles. This article presents a health-aware fast-charging framework based on reinforcement learning that simultaneously mitigates multiple degradation mechanisms while maintaining the computational efficiency. A coupled electrochemical–thermal–mechanical model is developed within a phase-field framework to simulate the initiation and evolution of lithium plating and particle cracking. Multiphysics constraints are formulated to restrict lithium plating current and diffusion-induced stress (DIS), preventing both short-term capacity drop and long-term aging. A reinforcement learning algorithm is subsequently employed to optimize the charging policy, effectively decoupling the complex model from online operation and significantly reducing the computational burden. Experimental findings confirm that the introduced strategy enables rapid charging while effectively suppressing degradation, outperforming conventional methods in both aging suppression and computational efficiency. The framework proposes a viable approach to attain the high-performance fast charging in practical implementation.
7 July 2025 in IEEE Transactions on Sustainable Energy
The high penetration of distributed generators (DGs) deteriorates the voltage violations in active distribution networks (ADNs). Owing to the flexible adjustment capacity, the local power regulation provided by soft open point (SOP) presents a promising solution for eliminating voltage violations in ADNs. A data-driven local control method can fully excavate the potential logic from operational data without requiring precious network parameters. However, the training data may be insufficient in practical applications. In this paper, a self-optimizing local voltage control method for SOP is proposed to achieve adaptive control in label-poor conditions. First, a SOP local control model is constructed based on lift-dimension mapping linearization (LDML), which portrays the complex relationship between ADN states and SOP control strategies. Subsequently, a self-optimizing guidance mechanism is established to obtain the label data of SOP control strategy, which provides a large number of training samples for the local control model. Finally, the effectiveness of the proposed method is validated using a practical distribution network with a four-terminal SOP. Results demonstrate that efficient control strategies can be determined based on local state measurements. A rapid response to DG fluctuations can be achieved while enhancing the adaptability to variations in practical operations.
6 January 2026 in IEEE Transactions on Industrial Informatics
With the increasing integration of low-voltage distributed photovoltaics (PVs), effective nonintrusive detection methods of user-side PV have become critical for the operators of active distribution networks. However, existing methods are mostly offline, requiring long-term data accumulation and lacking rapid response capability. This study proposes a data-driven method for fast detection and capacity estimation of user-side PVs. A cloud-edge collaborative framework is constructed to maximize the utilization of data features from smart meters. Within this framework, a lightweight rolling detection and capacity estimation method is proposed, reducing the required data volume and thus enhancing efficiency. A validation model based on historical data is developed in addition to ensuring the reliability of estimation results. Case studies verify that, with only 4 h of data, the proposed method can attain accuracies similar to existing methods that require multiple days of data. In addition, it can capture the uncertain variations in PV capacity within a short time, enabling operational applications such as PV monitoring and control in active distribution networks.
1 November 2026 in IEEE Transactions on Knowledge and Data Engineering