**4. Discussion**

The confidence margins depend on the strength of the cross-couplings and on the accuracy by which they can be determined. For the transition points the contribution of the model uncertainties is about 2%, whereas for the substation it reaches 8% [10]. These uncertainties should be considered when comparing measured and simulated overvoltage magnitudes. Figure 8 shows the statistical distribution of relative differences between simulated and observed maximum overvoltages. The data include six switching actions at all measurement locations, each providing three overvoltage values for the three phases. The width of the fitted normal distributions found for the substation measurement (6.8%, Figure 8a) is only slightly larger as compared to the width for the two transition points (6.0%, Figure 8b). On one hand, this shows that the modelling method is adequate and predicts observations well. On the other hand, it does not represent the difference in measurement accuracy expected from both locations. Apparently, the simulations also contribute to deviations. This is not surprising as, although the complete Dutch EHV grid was simulated in large detail, a number of simplifications were made. Connections to downstream networks (110, 150, 220 kV) were replaced by equivalent impedances based on active and reactive power consumption by the loads according to powerflow calculations [7]. Moreover, connections abroad were not modelled in detail.

**Figure 8.** Comparison of measurement and simulation for transition overvoltages upon energizing from nearby substation: (**a**) Results for the substation; (**b**) Results for the two transition points.

The comparison with divider measurements revealed serious differences and apparently EMC is an issue. High frequency components are overrepresented in the divider waveforms, which points to interference related to the signal time derivative. Common mode currents along measurement cables, as discussed in Section 2, can be induced by the time varying magnetic flux caused by ground currents. Poor measurement cable quality and cable layout are suspected to translate the common mode currents into disturbance adding to the recorded waveforms. Also, slight disturbance seems to be picked up by the D/I system. For the measurements presented in Section 3.2 the ground plates of each sensor were connected to the metal support of the termination near the sensor. This causes ground loops formed by the earth screens of the measurement cables, which can pick up magnetic flux from a switching event. For the measurements presented in Section 3.1 all sensor ground plates were connected to a central point, avoiding the occurrence of such loops to a large extent. It shows, that even with the D/I methodology, with its intrinsic good EMC characteristics, every connection detail is important in order to perform disturbance free measurements.
