Distinguishing the types of partial discharge (PD) caused by different insulation defects in gas-insulated switchgear (GIS) is a great challenge in the power industry, and improving the recognition accuracy of the relevant models is one of the key problems. In this paper, a convolutional neural network and long short-term memory (CNN-LSTM) model is proposed, which can effectively extract and utilize the spatiotemporal characteristics of PD input signals. First, the spatial characteristics of higher-level PD signals can be obtained through the CNN network, but because CNN is a deep feedforward neural network, it does not have the ability to process time-series data. The PD voltage signal is related to the time dimension, so LSTM saves and analyzes the previous voltage signal information, realizes the modeling of the time dependence of the data, and improves the accuracy of the PD signal pattern recognition. Finally, the pattern recognition results based on CNN-LSTM are given and compared with those based on other traditional analysis methods. The results show that the pattern recognition rate of this method is the highest, with an average of 97.9%, and its overall accuracy is better than that of other traditional analysis methods. The CNN-LSTM model provides a reliable reference for GIS PD diagnosis.

Partial discharge (PD) is one of the major form expressions of gas-insulated switchgear (GIS) insulation defects. Because PD will accelerate equipment aging, online monitoring and fault diagnosis plays a significant role in ensuring safe and reliable operation of the power system. Owing to feature engineering or vanishing gradients, however, existing pattern recognition methods for GIS PD are complex and inefficient. To improve recognition accuracy, a novel GIS PD pattern recognition method based on a light-scale convolutional neural network (LCNN) without artificial feature engineering is proposed. Firstly, GIS PD data are obtained through experiments and finite-difference time-domain simulations. Secondly, data enhancement is reinforced by a conditional variation auto-encoder. Thirdly, the LCNN structure is applied for GIS PD pattern recognition while the deconvolution neural network is used for model visualization. The recognition accuracy of the LCNN was 98.13%. Compared with traditional machine learning and other deep convolutional neural networks, the proposed method can effectively improve recognition accuracy and shorten calculation time, thus making it much more suitable for the ubiquitous-power Internet of Things and big data.

The assembled beam bridge is one of the most commonly used and most developed structural types in rapid industrialized bridge construction technology. The hinge joint is the key to the overall cooperative performance of this type of structure, and it is also the most vulnerable part in the service period. This article proposes a monitoring and evaluation method for the lateral collaborative working capability of this type of bridge based on the dynamic strain correlation coefficient index. The proposed method has the advantages of no need to interrupt traffic, easy implementation, concise physical meanings, being independent of short-term traffic load, and so on. Its potentialities to be used in the monitoring and evaluation of the technical state of the hinge joint and the lateral collaborative working performance of the whole bridge have been expatiated theoretically. Through numerical simulation analysis and real bridge monitoring data analysis, it is proved that the method is effective and accurate, and can be applied to the health monitoring system of such bridges.

A porcelain insulator is an important part to ensure that the insulation requirements of power equipment can be met. Under the influence of their structure, porcelain insulators are prone to mechanical damage and cracks, which will reduce their insulation performance. After a long-term operation, crack expansion will eventually lead to breakdown and safety hazards. Therefore, it is of great significance to detect insulator cracks to ensure the safe and reliable operation of a power grid. However, most traditional methods of insulator crack detection involve offline detection or contact measurement, which is not conducive to the online monitoring of equipment. Hyperspectral imaging technology is a noncontact detection technology containing three-dimensional (3D) spatial spectral information, whereby the data provide more information and the measuring method has a higher safety than electric detection methods. Therefore, a model of positioning and state classification of porcelain insulators based on hyperspectral technology is proposed. In this model, image data were used to extract edges to locate cracks, and spectral information was used to classify the surface states of porcelain insulators with EfficientNet. Lastly, crack extraction was realized, and the recognition accuracy of cracks and normal states was 96.9%. Through an analysis of the results, it is proven that the crack detection method of a porcelain insulator based on hyperspectral technology is an effective non-contact online monitoring approach, which has broad application prospects in the era of the Internet of Things with the rapid development of electric power.

Measuring the electron density of low-density plasma has long been a challenge, as traditional diagnostic techniques often lack sufficient sensitivity or spatial resolution. To address this, this paper presents an electron density diagnostic method for corona plasma generated by electrical wire explosion, based on a Shack–Hartmann wavefront sensor. The diagnostic system integrates an 8 ns, 532 nm nanosecond pulsed laser, a 400 mm focal length lens array, and a high-resolution charge-coupled device camera. By integrating the proposed neural network methods with micro-lens optical simulations, we achieved sub-pixel-level centroid localization of the focal spot, reducing shift errors by 21% compared to conventional convolutional neural network methods and achieving sub-pixel accuracy with an error of only 0.25 pixels. The theoretical sensitivity of the system reaches 2 × 1015 cm−2. In experiments using a silver-wire load under pre-pulse conditions, this diagnostic technique demonstrated strong agreement with laser interferometry results in high density regions. In low-density regions, the electron density measured at 4.4 mm from the wire axis reached a minimum of 2.2 × 1016 cm−2, offering a broader spatial range for plasma electron density diagnostics compared to laser interferometry. This method effectively captured the electron density distribution of the corona plasma formed under pre-pulse conditions. Future work will focus on further optimizing the experimental system to reduce sources of error and extending the application of this technique to large-scale pulsed power facilities for diagnosing the formation and parameter distribution of low-density electrode plasmas.

We present a unified Bayesian inversion framework for multi-diagnostic plasma analysis, demonstrated here on Z-pinch experiments. A shared Bayesian neural network (BNN) core couples with interchangeable physics-based forward models, so that one variational inference engine serves all diagnostics with built-in uncertainty quantification. Two modules are developed. (i) A two-angle Thomson scattering (TS) module that enforces a single self-consistent plasma state across viewing angles by modeling the full dynamic structure factor convolved with pre-characterized instrumental functions—unlike conventional analyses that treat each angle independently. (ii) A joint interferometry and Faraday rotation module that simultaneously constrains phase and magnetic field, replacing the standard sequential workflow in which phase-unwrapping errors propagate uncorrected into field estimates; a hybrid semi-global matching plus BNN approach resolves phase-branch ambiguities where conventional algorithms fail. On synthetic data with realistically modeled noise, the TS module achieves overall coefficient of determination R2>0.98 for ne, Te, Ti, and ion velocity, and satisfactory reconstruction of the electron velocity, which is inherently harder to infer. The interferometry module achieves phase mean absolute error (MAE) <0.14 rad over a ∼5π rad dynamic range and path-averaged magnetic-field MAE <7.2 T against ∼40 T peak values. The framework is validated on a Z-pinch experiment, yielding spatially resolved ne, Te, Ti, and flow velocities with full uncertainty quantification.