**3. Results**

#### *3.1. Effect of the PAGV in the Real Savonius*

*AF* of around 2 has been corroborated in the experiments of the wind tunnel. This obviously depends on the angle and the length of the vanes and the exact position of the Pitot tube, but this value of *AF* can be obtained with a suitable disposition of the vanes. However, this measurement is strongly influenced by the exact position and size of the Pitot tube, and future works should study the behavior of *AF* for higher and more turbulent wind speeds.

Although a maximum outlet/inlet width ratio of 1.4 can be obtained due to the lack of space in the tunnel with the Savonius rotor in the center, the instantaneous power measured for different wind speeds and different angles of the vane in Figure 6 gives a coherent result for this augmentation and subsequent power. Figure 8 shows this behavior with the vane at 30◦, 45◦ and 70◦ with respect to the horizontal. The pilot test was developed without the vane and the negative-torque wall (NT wall) was applied with the vane in vertical position, obstructing the negative torque's drag. It should be noted that, removing the negative torque, the power is doubled and the other cases (30◦, 45◦ and 70◦) also remove the negative drag.

The results in the curves of Figure 8 show the best working condition for the vane at 45◦, in which the captured power almost triples the pilot test power at each wind speed. Being the geometrical augmen<sup>t</sup> relation of 1.4, and 1.4<sup>3</sup> ≈ 3, the power also keeps the typical proportionality relation with *U*<sup>3</sup> for *AF*, and, again, a constant *AF* equal to the geometrical relation is deduced. This fact establishes an important particular case for an hypothetical law that should be demonstrated: for the adequate vane angles, the estimation of power production can be performed using *AF* × *U* as the input wind speed for any free wind speed *U*.

**Figure 8.** Power production versus wind speed for different positions of the vane and the pilot test without the vane.

The behavior of *Cp* vs. *TSR* was also studied and is presented in Table 2 for the optima with which *AF* can be estimated. The presence of the vane increments *TSRopt* from 0.5 to around 1 with *AF* ≈ 2. The best case in power augmentation (almost three times) for *AF* = 2.2 is measured for the vane at 45◦. These results are totally coherent with previous works using different augmentation techniques that obtain *TSRopt* ≈ 1 compared to a value of 0.5 for a conventional Savonius. In these

cases, *Cp*,*max* is also two or even three times higher thanks to the vanes, deflectors, curtains or other kinds of concentration configurations [44,51]. Additionally, as mentioned above, Altan et al. also already showed for their curtain type augmentation technique that the inferior vane should be at 45◦ to optimize the energy capture [52].

**Table 2.** Optimum *Cp*, and corresponding *TSR*, *AF* and power increment for the Savonius turbine in the wind tunnel for the pilot text, negative torque vertical wall, and different angles of the vane.


#### *3.2. Augmentation Factor in the Small-Scale Building Model*

In this case, *TSR* for each wind speed is measured instead of the *TSRopt*. Consequently, the ratio of the *TSR*s with (*TSRv*) and without (*TSR*0) the vane should be corrected according to the rotor speed, since the torque is incremented with the speed. Figure 9 shows these results: both TSRs, their ratio, and the corrected ratio that equals *AF*. This correction factor is established by the increment relation between *U* and the cut-in wind speed for the pilot experiment (5 m/s). A logical step in the procedure considering the constant speed–torque relation of the DC generator, since both TSRs (therefore, the rotor speeds) are practically linear with respect to *U* and also fulfill the same increment relation. Thus, *AF* is between 2.5 and 3, a very relevant result given the fact that the inlet–outlet width relation is 3:1. The 4:1 relation of the initial configuration could not be installed yet, due to the sensible construction details of the small model. Because of this delicate structure, the fabrication process of which has been very laborious, the free wind speed range in the tunnel has been kept below 10 m/s.

**Figure 9.** TSRs, the ratio, and the corrected ratio for the rotor speed with and without the vane.

#### *3.3. Wind Rose around the Building*

The ERA5 grid around Eibar city is shown in Figure 10 with the ERA5 points in blue. The building on which the anemometer is located is marked in red. The nearest grid point, at a distance of 2.48 km, was chosen to perform the calibration.

**Figure 10.** Nearest ERA5 grid points (blue) around the study point (red).

With the wind rose representing the nearest ERA5 grid point and the anemometer (see Figure 11), it is easy to realize that the ERA5 data have to be calibrated to make an appropriate estimation. ERA5's wind rose shows a strong predominant direction toward the northwest, as it is well-known that the climate of the Basque Country is highly related to the behavior of geostrophic winds [14]. This predominant direction is perpendicular to the valley in Eibar, and it is clearly diminished by the roughness of the terrain and the obstacle of the mountains in the anemometer's data. In fact, Eibar is an industrial city with a population of 20,000 in a deep valley surrounded by mountains that are around 600 m from where the River Deba opens toward the northeast direction.

Thus, this big difference could be explained by the shape and direction of the valley in which Eibar is located. It shows the need for field measurements and indicates that a good calibration methodology must use atmospheric reanalysis to study wind potential in places such as cities and

deep valleys, which have high surface roughness. Furthermore, the valley direction determines not only the calibration direction for energy estimation purposes, but it also defines which of the facades of the building should be selected for the implementation of the BIWT.

**Figure 11.** Details of the location of the anemometer and the nearest ERA5 grid point (**left**). Representation of the corresponding wind roses of ERA5 and anemometer data (**right**).

#### *3.4. Comparison between ERA5 and the Anemometer*

These data should be established at a referential height using the log law and the roughness of urban environments [9]. The ERA5 grid-point height is 411 m, thus both datasets should be established at the same height, which is the anemometer's height in this case, since it is the observation.

According to usual considerations in the wind energy sector, the roughness (*z*0) of the urban terrain is around 1–10 m. Roughness is used to apply the logarithmic law of vertical wind shear,

$$\frac{\mathcal{U}(z)}{\mathcal{U}(z\_r)} = \frac{\mathcal{L}n(z/z\_0)}{\mathcal{L}n(z\_r/z\_0)'} \tag{5}$$

which results in a correction factor between 0.86 and 0.77 for a wind speed at a height of 178 m in ERA5. In terms of the speed of reference at its original height, the authors calculated the following:

$$\text{LI}(178) = \frac{\text{Ln}(178/1)}{\text{Ln}(411/1)} \text{LI}(411) = 0.86 \times \text{l}\newline (411); \frac{\text{Ln}(178/10)}{\text{Ln}(411/10)} \text{l}\newline (411) = 0.77 \times \text{l}\newline (411) \tag{6}$$

A correction factor of 0.86 was used prior to the calibration method, which is based on quantile mapping. However, first, the correlation between ERA5 and the anemometer had to be directionally studied, mainly in the direction of the valley line. Furthermore, anemometer data had been previously filtered using advanced filters in meteorology, such as temporal checks, persistence tests, and climate-based range tests [53], which were implemented in the R programming language by the authors [54].

Figure 12 shows a time series of a week in June 2018 when the Pearson's correlation between ERA5 and the anemometer was very high (around 0.95); the wind direction vectors are illustrated above each time point that shows a strong westerly component. The parallel patterns shown by the wind speed

series are obvious in the graph. These examples verify the quality of the anemometer's data since their results are comparable to the reliable ERA5 data in the predominant direction line established by the valley (southwest–northeast). In the first days of this week, a cut-off low occurred in the Bay of Biscay, and it caused strong wind and a large amount of precipitation in that area. During the following days, without the influence of the cut-off low, wind moving in the north direction was observed. This is a global-scale synoptic situation that is easier to detect by the ERA5 model than local setups. Therefore, a high correlation between the observed data and ERA5 data was confirmed.

Additionally, the approximation to observation of the corrected ERA5 signal resulting from the application of the log law is clear in the time series using the 0.86 factor, but it is not enough to totally correct the general overestimation presented by ERA5. Although an extreme correction for a high roughness *z*0 = 10 m with a factor of 0.77 would strongly reduce this overestimation (Equation (6)), the usual roughness values for urban environments are kept in this graph.

This example is an extraordinary case, but, if all the cases of wind between the south and east were selected during the study period, a good correlation of 0.70 would finally be obtained. This validation was therefore enough to justify the calibration in this directional range, from which the corresponding building facade will be selected for capturing wind energy.

**Figure 12.** Wind speed of nearest ERA5 grid point (blue) and anemometer (red) during a week of June. In addition, wind direction vectors are represented in the graph.

#### *3.5. Estimation of the Energy Potential*

On the basis of the resource assessment results of the wind potential in buildings and the described wind tunnel experiments, a general methodology is presented that also references previous results from the scientific literature to estimate the annual energy production (*AEP*) of ROSEO-BIWT:


Although the results of the comparison and the calibration of data are very relevant, the objective of this paper is mainly methodological, and a preliminary estimation of AEP should be made using a well-known device. Thus, for the estimation of generated power, a commercial Savonius model (SeaHawk-PACWIND) was used: the rated power is of 1.1 kW, the rated wind speed is 17.9 m/s, the cut-in wind speed is 3.1 m/s, the cut-off without a given limit is a drag device, and the swept area is 0.92 m<sup>2</sup> [55].

When the pure ERA5 wind speed distribution was considered for the best facade, after the quantile-matching calibration using the anemometer data, these are the preliminary results:


$$[1 - F(17.9)] \times 365.25 \times 24 = 160\tag{7}$$


At low annual average wind speed of 3 m/s, the maximum *AF* = 4 can produce 2000 h at rated power. At *U* = 4 m/s, *AF* between 2.5 and 3 is necessary to ensure the 2000 h. At *U* = 5 m/s, the minimum *AF* = 2 obtained with only the inferior vane (Figure 5) is almost sufficient. At high *U*s, an *AF* = 3.5 or 4 implies 75% of the time at rated power.

**Figure 13.** Annual working hours at rated power versus *U* for different *AF* values.

Given the strong directionality presented by the anemometer's wind rose (Figure 11), two ROSEO-BIWTs installed in opposite facades of the building that are perpendicular to the valley would capture almost all of the mentioned hours because winds that are perpendicular to the valley are infrequent.

### **4. Conclusions and Future Outlook**

An integral methodology with preliminary results is presented for a new type of BIWT. The preliminary results include the energy potential estimation, measurements of small-scale building aerodynamic effects, and the influence of PAGVs. In the future, an AEP increase of 20% via PAGVs at the edge of the buildings must be demonstrated using a real prototype of ROSEO-BIWT at the edge of the building in Eibar. For that, the building in the city of Eibar will be used in the Bizia Lab project of the University of Basque Country, together with the previous wind tunnel experiments for the mentioned small-scale building with PAGVs and short Savonius prototype.

The anemometer was installed on the roof; with the new data provided by ERA5 for the nearest grid point, the identification of the best facade and the corresponding wind distribution were obtained following the methodology described in this paper applied to a longer period. This methodology will be relevant when the one-year period has elapsed. These preliminary results and the methodological discussion developed to date encourage us to implement future refinements of ROSEO-BIWT and the related wind energy estimation methodology.

If the building edge effect and the PAGVs produce a wind speed augmentation of *AF*, our general mathematical proof for working hours at the rated power shows that the hours without the augmentation can be considerably incremented. This is a very important general rule for turbines

with vanes, as shown in Figure 13. Furthermore, the values of *AF* between two and four are coherent with the literature and the experimental results, even under-valued, since the augmen<sup>t</sup> of the free wind in the edge of the building is not considered. The edge effect augmen<sup>t</sup> factor of 1.2 documented by the literature and the higher inlet-outlet width relationship other type of PAGVs could increase the overall *AF*.

Additionally, a novel validation method for anemometers developed by the authors in a recent study for wind farms [56] will be very beneficial since it enables the comparison and combination of data from more than one anemometer installed on the roof of the building. This allows us to consider both the zonal and meridional components in a single comparison score.

Future experiments in the wind tunnel with a small-scale building and the PAGVs will be carried out to obtain the optimum value of the angle between the vanes and the augmentation factor for different wind speeds within the operating range of the turbine. The augmentation factors measured with an interval of 0.5 m/s within this range, together with the measurement of the power curve of the longitudinal profile of the Savonius with the same step, will allow us to apply the corresponding histogram distribution of the corrected and calibrated wind to the augmented power curve. However, it is expected that future energy production results will be similar to the values presented here.

Finally, it should be emphasized that lacking the ability to change the viscosity of the air in the tunnel is an important inconvenience for small-scale building experiments. In the future, a more advanced tunnel with the ability to change the pressure and temperature is necessary, together with a parallel validation of the results using CFD simulations of the edge effect of the building with PAGVs.

**Author Contributions:** Conceptualization, A.U., O.G., and M.d.R.; Methodology, A.U., O.G., and M.d.R.; Software, A.U., O.G., M.d.R., S.C.-M., and A.G.-A.; Validation, O.G.; Investigation, A.U., O.G., and A.G.-A.; Writing—Original Draft Preparation, A.U. and O.G.; Writing—Review and Editing, all authors; Supervision, all authors; Project Administration, A.U.; and Funding Acquisition, A.U.

**Funding:** The research leading to these results was carried out in the framework of the Programme Campus Bizia Lab EHU (Campus Living Lab) with a financial gran<sup>t</sup> from the Office of Sustainability of the Vice-Chancellorship for Innovation, Social Outreach and Cultural Activities of the University of the Basque Country (UPV/EHU). This programme is supported by the Basque Government. We acknowledge also the availability given by the School of Engineering of Gipuzkoa-Eibar in the University of Basque Country, the EDP-Renewable awards in which we obtained the main award in September 2017, the Youth Enterprise Grant of UPV/EHU, and the project GIU17/02 of EHU/UPV. All computations and representations of this work were developed using the programming language R [54].

**Conflicts of Interest:** The authors declare no conflict of interest.
