As the altitude increases, air density decreases to cause a decrease in the air insulation strength. Test results in low-altitude areas indicate the significant impact of front time $t$ (μs) on the characteristics of long-airgap discharge. Nevertheless, the impact of $t$ on the breakdown voltage at high altitudes has yet to be studied. In this paper, impulse discharge characteristic tests for rod-plane gap were conducted at altitudes of 2500 m and 4300 m, with $t \in [2,1000]$ . The rod-plane gap discharge characteristic curves are obtained, and the critical front time and 50% breakdown voltage in the two areas are determined. Subsequently, the influence of altitude on the discharge characteristics under critical wave and standard switching impulse wave is compared and analysed. Based on the experiments considering low altitude, 2500 m and 4300 m high-altitude areas, the applicability of the altitude correction method recommended by IEC 60 071 is analysed. Moreover, an improved altitude correction method is presented to adapt to wide waveform variations in high-altitude areas. These results provide fundamental data and support for the external insulation design approaches in power grids at high altitudes.

With increasing altitude, the insulation strength of air gap decreases. Currently, research on gap discharge is primarily concentrated in low-altitude regions, lacking experimental and theoretical support for external insulation design of electrical equipment at high altitudes. To investigate the long-gap discharge characteristics at high altitudes, this study conducted experiments to obtain the switching impulse discharge characteristic curves of rod-plane gap at altitudes of 55 m, 2500 m, and 4300 m. In response to the distribution characteristics of the experimental data, we propose an invariant risk minimization neural network ensemble algorithm based on optimal transport. Based on experimental data, a breakdown voltage prediction model applicable to different altitudes was established. The model achieved an average error of 2.3% on the test set, validating its high accuracy and generalization. Additionally, the computational results of the proposed model were compared with existing altitude correction methods and other machine learning models, further validating its effectiveness. Finally, the model was utilized to obtain 50% breakdown voltage under typical meteorological conditions at different altitudes. The altitude correction method proposed in this paper can accommodate a wide range of climatic variations, thus providing valuable reference for the construction of high-altitude power grids.

The dimension and surface state of shielding fittings in ultra high voltage direct current (UHVDC) converter station valve halls have a great influence on their surface electric field and switching impulse characteristics, which are important parameters confirming the air gap distance in the valve hall. The characteristics of impulse discharge under different lengths, dent degrees and burrs around the scratches of Φ1.3 m shield balls with a 2 m sphere-plane gap length were tested, in the UHVDC testing base of the Hebei Electric Power Research Institute. The discharge characteristics under the influence of the surface scratches of the shield ball were obtained. The results demonstrate that the discharge voltage of sphere-plane gap decreases obviously when there are unpolished scratches on the surface of the shield ball. However, when the scratches are polished, the discharge voltage has no significant impact. At the same time, a 1:1 full-scale impulse test model was established based on the finite element method. The electric field intensity and the space electric field distribution of the shield ball were obtained under the influence of scratches with or without burrs. The results of the simulation show that when the surface of the shield ball is smooth, the electric field distribution around it is even. The electric field intensity on the surface of the shield ball increases obviously when there are burrs around the scratches. When there is no burr around the scratches, the length and depth of the scratches have no obvious effect on its electric field distribution. Meanwhile, calculation results are consistent with test results. The results can provide an important basis for the design and optimization of shielding fittings, and technical support for its localization.