*3.3. Indoor Temperature Analysis Results*

In the indoor temperature study, *MAE* and *RMSE* for the hourly criteria were used in the analysis in order to measure the quantitative variation in the temperature curve, and *R*<sup>2</sup> was used to study the deviation in the temperature curve form. The results are shown in Table 4. In this case, the same energy was injected into the model for the two simulations, using the *on-site* and *third-party* weather datasets. The comparison of the thermal zones' temperature provided by the simulations provided the influence of the weather dataset employed in the indoor temperature conditions. In the table, the results are shown using three criteria: (1) for all the thermal zones (*All*), where the statistical metrics are calculated using the indoor temperature of all the thermal zones of the building; and (2, 3) for the maximum (*Max*) and minimum (*Min*) temperature, where the metrics are calculated using only the temperature of the thermal zone that provides the maximum/minimum temperature in each time step in order to compare the effect in the internal healthy conditions when using the two weather data sources. For this study, the statistical metrics were calculated only for the hourly criteria. The rest of the time grains were not considered since, for the temperature analysis, the data were not accumulated when longer periods were analyzed.

This study provided similar results to the previous energy results. Regarding the results for the temperature of all the thermal zones (*All*), the high results in the *<sup>R</sup>*<sup>2</sup> for the four test sites (from ±87 to ±95%) showed that the shape of the indoor temperature curve was very similar when the two weather datasets were used in the simulation. However, the quantitative statistical metrics showed a significant impact on the indoor temperature. The Gedved school was the test site with the highest *MAE* (1.72 ◦C), in line with the energy analysis where this test site reached deviations in the energy demand of ±45%. The main reasons why the Gedved school had a higher impact on the indoor temperature were the influence of the lack of correlation of the wind speed between the *on-site* and *third-party* weather datasets and the way infiltrations were simulated based on the leakage area. The office building in Pamplona was the site with a minor impact on the indoor temperature (*MAE* of 0.55 ◦C) when the weather dataset was changed. When the maximum and minimum temperatures reached in the building were employed in the analysis (Max and *Min* in the table, respectively), it can be seen that the statistical metrics were similar to *All*. This means that the variation in the indoor temperature when using third-party weather data was stable and produced similar variations when the indoor conditions were minimum or maximum.

**Table 4.** The statistical metrics (*MAE*, *RMSE*, and *R*2) used in the indoor temperature analysis for the four test sites. *All*: metrics calculated with the temperature of all thermal zones; *Max/Min*: metrics calculated with the temperature of the thermal zone with the maximum/minimum temperature in each time step.


There was no high differences between the *MAE* and *RMSE* results for the four cases, which indicated that the variations in the indoor temperature were quite homogeneous with no significant outliers. Figure 8 shows, in a visual way, the results with a scatter plot for each test site where the temperature was weighted by a thermal zone volume of air. The office building in Pamplona (above left) had fewer scattered temperature points since they were closer to the black line than the rest of the cases, and most of the points had a difference of 1 ◦C or less. On the other hand, the Gedved school (above right) provided the worst correlation for almost all the temperature points with a difference bigger than 1 ◦C due to the difference between both weather files.

**Figure 8.** Scatter plots for each test site of the indoor temperature simulated using the *on-site* and *third-party* weather files.
