Unsupervised industrial anomaly detection requires accurately identifying defects without labeled data. Traditional autoencoder-based methods often struggle with incomplete anomaly suppression and loss of fine details, as their single-pass decoding fails to effectively handle anomalies with varying severity and scale. We propose a recursive architecture for autoencoder (RcAE), which performs reconstruction iteratively to progressively suppress anomalies while refining normal structures. Unlike traditional single-pass models, this recursive design naturally produces a sequence of reconstructions, progressively exposing suppressed abnormal patterns. To leverage this reconstruction dynamics, we introduce a Cross Recursion Detection (CRD) module that tracks inconsistencies across recursion steps, enhancing detection of both subtle and large-scale anomalies. Additionally, we incorporate a Detail Preservation Network (DPN) to recover high-frequency textures typically lost during reconstruction. Extensive experiments demonstrate that our method significantly outperforms existing non-diffusion methods, and achieves performance on par with recent diffusion models with only 10% of their parameters and offering substantially faster inference. These results highlight the practicality and efficiency of our approach for real-world applications.

Time-series forecasting in the energy sector is a labor-intensive process requiring expertise in multiple areas, such as data preprocessing, feature engineering, and neural network optimization. Although large language models offer automation potential, existing large-language-model-based multi-agent systems lack specialized structures for forecasting workflows, balancing computational cost and tool capability remains challenging. To address these issues, we propose an Energy-domain End-to-end Automated Forecasting Multi-Agent System, an on-premises end-to-end forecasting framework built upon a novel Department-Collaborative Multi-agent System Structure, specifically fine-tuned for energy forecasting applications. Within the Energy-domain End-to-end Automated Forecasting Multi-Agent System dominated by Department-Collaborative Multi-agent System Structure, two core and innovative modules are introduced: (i) Structured Behavioral Knowledge Distillation, which enables small-parameter large language models to operate feature engineering tools, facilitating lightweight and efficient on-premises deployment; and (ii) the Alignment-to-Selection Framework, which automates neural network selection by integrating architectural knowledge with historical performance data. Extensive experimental results demonstrate that the novel Energy-domain End-to-end Automated Forecasting Multi-Agent System significantly reduces manual effort, while maintaining high forecasting accuracy across diverse datasets.

A digital twin (DT) should continuously and accurately reflect the dynamic state of its physical twin (PT). In safety-critical and time-sensitive domains such as power systems, anomalies in PT data stemming from measurement or communication malfunctions, can compromise DT updates and the integrity of DT-driven applications. To overcome this risk, it is essential to identify and correct anomalous PT data prior to DT updates. However, conventional DT implementations lack mechanisms for evaluating PT data before executing updates, posing challenges to operational reliability and system safety. This paper proposes a trustworthy digital twin (TwDT) concept with the PT data evaluation function, supported by a denoising physics-informed autoencoder (De-PI-AE) algorithm. The PT data evaluation function of a TwDT comprises three key steps: i ) detection for direct updates using trustworthy data while flagging anomalous PT data, ii ) identification for localisation of the specific anomalous values, and iii ) estimation for correcting previously identified anomalous values. The inclusion of each component effectively forms an entire PT data evaluation function to ensure reliable and continuous updates. A power system digital twin (PSDT) is used to demonstrate the development and effectiveness of the proposed De-PI-AE algorithm to realise the TwDTs, highlighting its strong potential to be embedded into DT update process. The De-PI-AI solution outperforms other artificial intelligence-based (AI) algorithms, achieving detection (88.5%), identification (98.9%), and estimation (mean square error reduced by 98.8%) of the specific PT data anomaly. The proposed solution enables the detection, identification, and estimation under one framework, while achieving positive performance gains. The proposed TwDT with the PT data evaluation function can be adopted and expanded to more DTs within power systems and beyond.

Converter valve-side single-phase-to-ground (SPG) faults are among the most critical issues affecting the secure operation of modular multilevel converters (MMCs). However, the fault characteristics of these faults in symmetrical monopolar MMCs have not been thoroughly explored. This paper aims to fill this gay by theoretically investigating SPG faults in such systems. The analysis considers various influencing factors, including MMC grounding schemes and dc line types. The findings demonstrate that valve-side SPG faults cause dc voltage oscillations and therefore, large discharging currents from the distributed capacitors of dc lines. These scenarios are examined through mathematical analyses and verified through simulations of a generic MMC-HVDC system using PSCAD/EMTDC. The results accurately reveal the fault behaviors under different system conditions, providing valuable insights for the design of protection systems and insulation coordination for internal ac grounding faults in MMC stations.

Valve-side, single-phase-to-ground (SPG) faults in bipolar high-voltage direct current (HVDC) systems based on half-bridge modular multilevel converters (HB-MMCs) result in i) significant upper arm overvoltage, and ii) non-zero-crossing currents in the grid-side ac circuit breaker (ACCB). In this paper, a dc-link thyristor-based protection scheme is proposed to address both issues. The dc-link thyristor branches are installed in the dc terminal of each station. A dc short-circuit loop is formed following triggering the thyristor branch, redirecting currents from the upper arms to the branch, thus preventing continuous capacitor charging in the upper arms. Moreover, the creation of the dc short-circuit loop induces symmetrical components in the grid-side currents, which ensure current zero-crossings in the grid-side ACCB. The peak current flowing into the dc-link thyristor branch is also calculated to facilitate the design of the thyristor branch. The effectiveness of the proposed protection scheme is validated in an HB-MMC HVDC system simulated in PSCAD/EMTDC.

The broader use of modular multilevel converters (MMCs) in medium-voltage direct current (MVDC) systems combined with the multiple possible network topologies of a multiterminal dc (MTDC) network pose major challenges to fault analysis, short-circuit calculations and, eventually, system reliability. In order to evaluate dc fault levels in MTDC systems with MMCs, this paper derives an expanded RLC equivalent model for representing converters and networks with system characteristic matrices, based on which, state-space analysis can be implemented to facilitate fault current calculations. Analytical calculations and simulation results of MMC and MTDC branch currents across multiple scenarios and different configurations of an MTDC network with five converters identify fault influencing factors and reveal unique characteristics of fault analysis in MVDC systems.

The interoperability challenges of emerging modular voltage source converter (VSC) topologies HVDC systems should be addressed along with rapid transformation of power systems driven by the need for more interconnected networks. This paper derives a hybrid VSC-HVDC model and demonstrates the interoperability of emerging alternate arm converter (AAC) with the state-of-the-art modular multilevel converter (MMC) in HVDC systems. The study provides primary steps towards detailed analysis of AAC interoperability in complex dc grid configurations. A detailed set of results based on the 800 MVA hybrid VSC-HVDC system demonstrate the interoperability performance of the AAC under different operating scenarios and verify associated control functions.