**4. Discussions**

In general, if the assessment of safety procedures is successfully achieved for the worst-case scenarios, as a conservative measure, it can be said that the substation is fully protected. However, this could result in an overdesign of the required protection system and additional unnecessary financial costs. In order to achieve a safety level of the substation, the main characteristics of the protection system must overcome the transient overvoltage generated across the substation that is described by frequency, amplitude, and time-to-peak parameter. Designing a suitable safety protection system requires that the computational domain must contain the on-site physical configuration.

It is important to mention that the TGPR waveform illustrated throughout the current paper contained incident, reflected, and refracted components. The XGSLab software package computed the electromagnetic interaction between each pair of two metallic elements contained by the model. Therefore, the outcome of the method contained the combined effects of inductive, capacitive, and conductive couplings, fully taken into account.

Based on the conclusions from the state-of-the-art section (Introduction), a 3D model representing the entire metallic ensemble was required to accurately assess the holistic transient response of a gas-insulated substation and adjacent metallic structures during voltage breakdown fault, considering the very fast transient regime. The main concept of the proposed modeling technique was to represent the metallic enclosure (coaxial pipe) through several parallel aluminum bars. A first step of the proposed analysis method was to quantify how different numbers of parallel aluminum bars representing the solid metallic pipe will impact the transient response of the enclosure during voltage breakdown fault, taking into account very-high-frequency transients. According to the computed results (see Figure 3) there were negligible differences between the primary parameters of the TGPR when different enclosures were considered during the computational process, considering certain operating conditions.

Keeping in mind the hexagonal geometric model proposed, identified as most suitable for large GIS modeling, the transient ground potential rise was computed at several locations across the substations. According to the graphical and numerical results presented throughout the current paper, the spatial distribution of the TGPR across the grounding grid, especially inside GIS building, was not uniform, although the geometric arrangemen<sup>t</sup> of the earthing system was relatively symmetric with respect to the GIS platform, due to the presence of the metallic enclosure. As the distance from the transient source increased, as well as from the aluminum bars' configuration, the impact of electromagnetic couplings developed between particular metallic enclosures did not affect in a similar manner the

maximum TGPR values recorded at various analysis locations. The maximum attenuation coe fficients were recorded when the TGPR was analyzed starting from the upper side of the metallic enclosure up the GIS platform (69% when five-section configuration was considered, see Table 2). When the observation point moved toward the extremities of the grounding grid (namely, on the vertical rods) negligible TGPR attenuation rates were obtained. Therefore, it can be concluded that during very fast transients flowing throughout the substation the grounding grid e ffective area was located near the transient source, more precisely, on the GIS concrete platform.


**Table 2.** Maximum absolute amplitudes describing the TGPR waveforms at di fferent locations.

When the parametric study is employed several discussions can be initiated regarding the quantification of each GIS bus section into the overall transient response of the system. Through this type of analysis, the proper modeling technique is established. Moreover, through the number of GIS buses' variation during the simulations, the stability and modularity of the proposed modeling technique are achieved. The outcome of the method in all simulation conditions is strictly related to physical phenomenon rather than pure numerical dependencies. According to the TGPR comparison presented in a previous section (see Figures 9, 11 and 12), the oscillatory character of the wave is damped within the traveling path due to resistive, inductive, and capacitive behavior of the metallic ensemble. Therefore, the frequency attenuation coe fficient is directly proportional with the length of propagation path and equivalent with the number of elements contained by the model. According to the maximum overvoltage amplitudes presented in Table 2, when three, four, and five GIS sections are contained by the model, the area of the grounding grid clearing the transient energy is located inside the GIS building. Moreover, the copper strip situated beneath GIS metallic enclosure (copper strip ring and the grounding leads) will dominate the damping e ffect provided by the grounding grid during voltage breakdown (see Table 2).

When a single GIS bus configuration is considered during the computational process, the transient response of the system behaves di fferently, if compared with multiple sections' scenario. Similar TGPR maximum amplitudes are computed across the metallic enclosure and the grounding leads when a

single bus section configuration is considered. Having multiple GIS sections in the model is equivalent to considering multiple existing electromagnetic couplings across the substation due to multiple parallelism conditions provided by the aluminum bars' ensemble, which greatly affect the overall transient response of the substation.

Different attenuation profiles of the transient electromagnetic wave phenomena are observed when different GIS configurations are considered during the computational process. According to the previous presented statements, it must be mentioned that, under no circumstance, should the simplification of the model in order to reduce the computational time be applied; whereas significant differences occurring between the topology of the computed time domain electromagnetic waveform in all considered configurations were observed (see Figures 9, 11 and 12). Taking into account the different TGPR waveforms' topologies computed while single and multiple bus arrangements are considered, in order to design a suitable safety protection system, the computational domain must contain the on-site physical configuration.
