*4.6. CSI Response Scenario*

In the third scenario, we modeled a CSI sensitivity for all agents. After every charging process, the MAS computes the CSI value for the last charging session of each user. If for EV *u*, this values is under 80%, more than *nf* = 3 times, the agent *u* will not use the charging system anymore due to frustration. Obviously, this is a very simple scenario, but visualizes the potential impact of adaptive customer behavior.

## **5. Results**

We simulated the three scenarios defined above for a complete year using weather data for a city in Germany. Now, we want to take a look at the total charged energy in the CSI scenario for different values of *PBase* in Figure 4. We see that PV power on its own was not sufficient since the transferred energy increased substantially when *PBase* was increased. However, the selected value of *PBase* = 30 kW covered 87% of the maximum charging energy amount.

Looking at the impact of different scenarios on the total charged energy in Figure 5, we observed a small increase in the temperature-dependent usage scenario and a strong decline to about 65% of the base demand for the CSI scenario.

The daily charging power for the three different scenarios is shown in Figure 6.

The dependency of the number of daily charging processes is shown in Figure 7 for different scenarios and in Figure 8 for different values of *PBase*. We see that in the CSI scenario, the number of charging processes fell strongly in the first few months. In the final month (December), the mean number of charging processes for the CSI scenario was 47% lower than for the base scenario. An increasing grid energy supply (Figure 8) increased the number of long-term charging processes (customers), but even unlimited charging power did not avoid the effect of disappointed customers. This was due to charging requirements that were sometimes too demanding given the limited charging power of the CS or their EVs.

**Figure 4.** Accumulated charging energy over time for different values of *PBase* for the CSI scenario.

**Figure 5.** Accumulated energy demand over time for different agent scenarios (with *PBase* = 30 kW).

**Figure 6.** EV charging power (for all vehicles) for different scenarios.

**Figure 7.** Number of daily charging processes for different scenarios.

**Figure 8.** Number of charging processes each day in the CSI sensitive scenario for different values of *PBase*.

A boxplot of CSI values for all three scenarios is shown in Figure 9. Note that this plot only shows charging processes with a CSI of <1.0. In fact, 93–98% (depending on scenario) of all charging processes provide full customer satisfaction. Differences for the three scenarios were very small, with the CSI scenario even having the best average values and the Temp scenario the worst values. The changes for the CSI scenario when adapting *PBase* are shown in Figure 10. We note an increase in CSI values with more base charging power (taken from the grid). However, due to limited charging power, customer satisfaction might still be suboptimal even when a very large grid supply is available.

Finally, we study the total number of unsatisfying (CSI < 0.8) charging sessions. Figure 11 shows variations for different scenarios and Figure 12 the impact of *PBase* in the CSI scenario. Assuming that a CSI level below a threshold (0.8 in our simulations) indicates that the corresponding charging session led to customer frustration, these plots show the development in the number of weak charging sessions. These numbers might for example in real usage be related to the number of customer complaints. It is interesting to observe that in the CSI scenario, the number of low CSI charging sessions increased much slower than in the other two scenarios (Figure 11). After about two months, there was only a moderate increase in the total number, while in the Base and Temp scenario, those numbers increased much faster. This was due users with more difficult charging requirements switching to a different charging location, leaving the remaining users with a more suitable solution. A possible implication of CSI-sensitive behavior compared to the other scenarios is that a charging operator might substantially overestimate the required effort for handling customer complaints.

As expected, the temperature-dependent charging demand also increased the number of low CSI sessions relative to the baseline. A quite surprising finding in Figure 12 was the low difference in the number of final low-quality charging sessions at the end of the year for different *PBase* values (101 for 0 kW, 76 for 30 kW, and 68 for Max power (1 MW)). In terms of service efficiency (providing high CSI level charging sessions), even large variations in *PBase* had a rather small impact. In contrast, the response of the customers seemed to add actually some self-regulating feature to the overall system.

**Figure 9.** Average CSI values over a year for different scenarios for the standard (30 kW *PBase*) setting. Note that only charging processes with a CSI value of less than 1.0 are reflected here.

**Figure 10.** Boxplot of yearly CSI statistics for different values of *PBase* in the CSI-sensitive scenario. Note that only charging processes with a CSI value of less than 1.0 are reflected here.

**Figure 11.** Accumulated number of low (<0.8) CSI charging processes for different scenarios (with *PBase* = 30 kW).

**Figure 12.** Accumulated number of low (<0.8) CSI charging processes in the CSI-sensitive scenario for different *PBase* values.
