Electrical equipment is highly vulnerable to seismic hazards. Acquiring accurate seismic response data is crucial for post-earthquake damage assessment and emergency decision-making. Although conventional contact sensors are effective for capturing such data, they cannot be widely deployed on the equipment body due to monitoring constraints. Therefore, seismic response prediction methods based on time-series neural networks must be developed. Most existing studies have emphasized the optimization of neural network architectures, and dataset construction strategies have not been systematically and sufficiently investigated. Dataset construction directly influences the fitting accuracy and generalization ability of predictive models, ultimately determining the overall predictive performance. In this study, the effects of three key elements of dataset construction-ground-motion selection, amplitude scaling of records, and sample-size configuration-on the performance of time-series models were evaluated, and a scientifically grounded dataset construction workflow was proposed.

A 500-kV transformer bushing was selected as the case study. A refined finite element model was developed and validated by shaking-table tests to determine its accuracy in terms of dynamic characteristics and response behavior, and the acceleration at the top oil reservoir was chosen as the prediction target. Two ground-motion selection strategies were adopted for dataset construction: spectrum-matched records and random selection constrained only by site type. Four amplitude scaling strategies were examined: conventional random, conventional fixed, extended-range random, and extended-range fixed scaling. Five sample-size levels of 80, 100, 120, 140, and 160 records were also configured to form multiple strategy combinations. A recursive long short-term memory neural network was used as the representative prediction model. Its performance was assessed based on mean squared error and peak response error, and repeated sampling and multiple independent training runs were performed to mitigate stochastic variability.

Spectrum-matched selection outperformed random selection based solely on the site type, yielding lower overall prediction errors in seismic response time series. Fixed scaling was superior to random scaling, and the introduction of extended-range scaling further enhanced the peak prediction accuracy. Although commonly used, random scaling considerably reduced the overall and peak prediction performance of the model and was not recommended for seismic response time-series prediction. Increasing the number of training samples improved the model accuracy, but marginal gains were observed at a sample size of 120-140 records. The combination of spectrum-matched selection and extended-range fixed scaling was the most effective strategy. Comparative tests with representative ground-motion records further confirmed that this strategy surpassed commonly used empirical approaches in terms of fitting accuracy, peak prediction capability, and training stability; it also enabled more accurate capture of abrupt response transitions and reduced phase errors.

A recommended dataset construction workflow is proposed for the time-series prediction of electrical equipment modeled as linear elastic systems. The proposed process integrates finite element modeling and validation, spectrum-matched ground-motion selection, extended-range fixed scaling, and balanced sample-size configuration. The findings confirm that this workflow considerably improves both the prediction accuracy and overall stability of the model, offering systematic methodological support and practical engineering guidance for post-earthquake emergency assessment and response monitoring of electrical equipment. This approach can also be extended to other structural systems where dataset construction critically affects the model performance.

As a crucial component of power systems, the valve hall plays a vital role in ensuring the safe and stable operation of converter stations, as well as the reliable transmission of electricity. However, earthquakes pose a significant threat to the integrity and functionality of these systems. Under seismic loading, due to the uniformity of the ground motion input and the mechanical and functional coupling between components, the dynamic responses of the equipment are correlated. This interdependence challenges the conventional assumption of independent component failures, highlighting the need for more advanced reliability modeling.

To demonstrate the importance of considering response correlation between components, this study first analyzes the reliability of series and parallel systems under two extreme conditions: complete independence and complete correlation. The findings show that the correlation between equipment responses has a significant effect on system reliability, especially when the number of components increases or the failure probability varies widely. To quantify the correlations within the valve hall system, finite element simulations were conducted on both high-voltage and low-voltage valve halls of a ±800 kV converter station. The seismic responses of individual equipment were obtained under various ground motion inputs. Based on these data, a Gaussian copula-based method was employed to model the joint behavior of equipment responses. This method captures their statistical dependencies without assuming a predefined joint distribution. The analysis process mainly consists of marginal distribution modeling of the responses, transformation to uniform and normal distributions, determination of the correlation matrix, generation of independent normal samples, singular value decomposition of the correlation matrix, standard normal transformation, inverse mapping to the uniform domain, and generation of correlated samples. This approach enables the construction of a realistic joint distribution of equipment responses while preserving their marginal characteristics. Using the correlated samples generated through the Gaussian copula method, the system-level reliability of the valve hall was evaluated while accounting for dependent failure behavior.

The Gaussian copula method effectively modeled the correlation structure of variables using the correlation matrix, enabling accurate modeling of the correlation structure of the response states of the whole system. The assumption of complete independence of component states underestimated the actual system reliability. The degree of underestimation varied significantly with the peak ground acceleration (PGA), with the most pronounced effect observed at PGA=0.400g. Furthermore, neglecting the correlation between component failures resulted in an overestimation of the economic losses of the converter station, and the magnitude of this overestimation increased with the system's design life.

The proposed reliability assessment framework incorporates equipment response correlations, yielding more accurate and realistic assessments of valve hall system performance under seismic conditions. This method also overcomes the limitations of traditional approaches based on independent assumptions, which are commonly adopted in large-scale system reliability analysis due to their computational simplicity. The proposed method is flexible and extensible, with broad application prospects in the reliability and risk assessment of complex infrastructure systems subjected to extreme events.

In line with the development and integration of new power systems, many large-scale renewable energy bases-particularly wind and photovoltaic-are being rapidly established in desert, Gobi, and other arid (DGA) regions across China and beyond. These regions are characterized by harsh climatic and geological conditions, making the reliable operation and rapid recovery of electrical infrastructure under extreme weather events increasingly critical. The increasing frequency and intensity of extreme weather events under global climate change further amplify this challenge. This study investigates the safe operation and resilience of large-scale renewable energy bases in DGA regions under extreme environmental conditions. This study aims to systematically review the impact of various extreme hazards on electrical equipment across the power system chain, assess the current state of disaster prevention and mitigation technologies, and identify critical technical needs for future development. This study provides a solid theoretical foundation and practical technological support for enhancing the resilience and intelligent transformation of modern power systems.

A comprehensive review was conducted on recent domestic and international cases of power system failure and associated economic losses triggered by extreme weather events, including extreme low and high ambient temperatures, atmospheric icing, strong winds, sand and dust storms, earthquakes, lightning strikes, wildfires, floods, and secondary compound disasters. The analysis covers the full lifecycle of electrical infrastructure, including the power generation, transmission, and transformation stages. For each stage, critical threats to the operational security and structural integrity of key electrical equipment are identified. The results indicate that the unique environmental characteristics of DGA regions-high solar radiation, strong convective winds, large diurnal temperature variations, and frequent sandstorms-exacerbate the vulnerability of electrical equipment, particularly outdoor components such as transformers, insulators, switchgears, and towers. The primary types and impact mechanisms of extreme environmental factors on equipment in DGA regions are categorized. Their associated degradation modes, including material embrittlement due to low temperatures, overheating and insulation aging under extreme heat, salt fog and corrosion effects, mechanical fatigue from wind-induced vibration, and flashover risks due to pollution and icing, are discussed in detail. This study delineates the specific vulnerabilities of various types of electrical equipment and the main failure modes associated with each hazard. The current status of monitoring, early warning, emergency response, and disaster mitigation technologies is also critically analyzed. Although solutions such as online monitoring systems, structural reinforcement methods, seismic isolation devices, de-icing systems, and vibration-damping technologies have been proposed and partially implemented, many challenges remain. Despite promising results from pilot-scale deployments and demonstration projects, large-scale practical applications are hindered by technical bottlenecks. These include insufficient monitoring precision in complex environments, limited capacity for real-time online condition assessment, and reduced effectiveness in multihazard detection and degradation tracking. Furthermore, challenges in data integration, system interoperability, and long-term stability in harsh environments significantly undermine the reliability of disaster response systems in real-world engineering applications.

As the risks posed by extreme climate events continue to grow, transitioning from passive disaster response to active, intelligent risk management across the entire lifecycle of power systems is urgently needed. Future efforts should focus on creating a standardized, modular framework for disaster prevention and mitigation that can be rapidly adapted to a wide range of hazards. Intelligent decision-making platforms, supported by digital twin models, big data analytics, and AI-driven prediction algorithms, should be developed to provide real-time operational guidance under extreme conditions. Moreover, cross-disciplinary collaboration among meteorology, materials science, structural engineering, and electrical engineering is essential for designing equipment and systems inherently resistant to compound disasters.