An Experimental Study on the Performance of Check Valve-Aided Savonius Wind Rotors with Semi-Circular Blade

The Savonius wind turbine is perceived favorable for a small-scale application because it is simple, relatively inexpensive, insensitive to wind directions, and has a good self-starting ability. However, it has a low power coefficient caused by the negative torque from the returning blade. This study aimed to investigate the effect of valve addition on the semi-circular Savonius rotor's performance. The experiments for this study were conducted on an open jet wind tunnel, with the valve located near the rotation axis, blade center, and rotor tip. The valve opening area ratio used was 0.02, 0.04, and 0.06, with Reynolds numbers 73,000, 86,000, and 99,000. The results showed that rotors with a valve placed near the tip performed better than those at the blade center and near the rotation axis. The performance decreased with an increase in the valve opening area ratio. Furthermore, the performance-improvement magnitude decreased with an increase in the Reynolds number. The rotor with a single valve near the tip performed the highest at a Reynolds number of 73,000. It achieved a maximum power coefficient of 0.199 compared to 0.183 obtained by a conventional Savonius rotor.    

Augmentation; Check valve; Renewable energy; Savonius wind-rotor; Wind turbine

The Intergovernmental Panel on Climate Change (IPCC) warned that the world would reach 1.5 ^oC of warming by 2040, and only swift and drastic cuts in carbon emissions would help prevent an environmental disaster (IPCC, 2022). One way to reduce carbon emissions is reduced by using renewable energy sources such as wind energy to decarbonize the global electricity generation systems. In 2019, global wind power capacity reached 651 GW, contributing approximately 5.9% of global electricity generation (REN21, 2019). With wind energy costs expected to drop significantly in the future, this contribution is expected to increase (Wiser et al., 2021).

Savonius rotor is a vertical axis wind turbine (VAWT) comprising two semi-circular buckets arranged asymmetrically to resemble an 'S' shape. It works based on drag force with a slight contribution from the lift force (Alom, Borah, and Saha, 2018). The positive torque generated by an advancing blade moving with the wind is higher than the negative torque generated by the returning blade moving against the wind. The difference between the two torques causes the rotor to spin. The Savonius rotor power coefficient (C_P) is much lower than the horizontal axis wind turbine (HAWT) due to the negative torque produced. However, it has several intrinsic advantages over the  HAWT,  such as simpler design, good starting ability, omnidirectional, lower noise, and easier to maintain (Kumar, Raahemifar, and Fung, 2018). These advantages make the rotor more suitable for harnessing wind power in urban sites, building rooftops, and remote areas (Cho, Jeong, and Sari, 2011; Ishugah et al., 2014; Mao et al., 2020). In addition to energy generation, wind turbines installed in an urban area allow direct use of energy which eliminates energy transmission losses (Krasniqi, Dimitrieska, and Lajqi, 2022). When combined with a suitable and improved generator design, such as a reduced cogging-torque permanent magnet generator (Nur and Siregar, 2020), the Savonius rotor may provide an interesting method for small-scale wind power generation. The basic geometry of a conventional Savonius rotor is illustrated in Figure 1.

Figure 1 Basic geometry of a conventional Savonius rotor

The performance of a Savonius rotor highly depends on its geometric parameters. Saad et al. (2020) found that the rotor with end plates performed better, with the optimum value for end plate diameter (De) at 1.1 of the rotor diameter (D). Mahmoud et al. (2012) investigated the aspect ratio (AR), a non-dimensional parameter resulting from dividing the rotor height (H) by its diameter. The study found that the maximum C_P (C_PMax) was improved as the AR increased. Saad, Ookawara, and Ahmed (2022) observed a similar finding on multi-stage rotors. This C_P improvement for a higher AR reduces loss at the tip, similar to adding end plates (Akwa, Vielmo, and Petry, 2012). Studies on the overlap ratio (?) of the distance between rotor blades (e) and the blade chord length (d) found that the optimum value ranges from 0.15 to 0.2 (Alom and Saha, 2017; Cuevas-Carvajal et al., 2022; Roy and Saha, 2013). Additionally, Kamoji, Kedare, and Prabhu (2009) investigated the effect of the Reynolds number on a conventional and modified Savonius rotor. The study found that the C_PMax improved by 19% as the Reynolds number increased from 80,000 to 150,000.

The modification f the conventional Savonius rotor blade shape can also improve its performance. Abdelaziz et al. (2022) numerically studied the performance of a conventional Savonius rotor with various outer and inner arc angles and gap ratios. They managed to improve the C_PMax by 4.5% and 12.9% with a 160^O outer angle and 20^O inner angle, respectively. A novel Savonius rotor with a modified-bach blade was proposed by Elmekawy, Saeed, and Kassab (2021). They found a 32.2% C_PMax improvement over a conventional Savonius rotor through numerical study. The blade shape was also modified to improve the Savonius performance at a higher TSR range (above 0.8). Savonius rotor with a combination of the semi-circular and elliptical profile was found to have C_PMax at TSR=1.4, double that of a conventional rotor's TSR=0.7 (Le et al., 2022). Based on the result, Dinh Le, Minh, and Trinh (2022) further modified the previous rotor into a combination of multi-curve and auxiliary blades. The new rotor combination not only possessed a higher C_PMax at a higher TSR, but it also exhibited a 6.9% higher C_P than the conventional rotor in the lower TSR range. 

Several studies have suggested different methods to enhance the performance of the Savonius rotor. For instance, Mohamed et al. (2010) improved the CP by 27.3% by placing a deflector upstream to prevent the wind from hitting the returning blade. Nimvari, Fatahian, and Fatahian (2020) used a porous deflector and achieved a 10% increase in C_PMax. El-Askary et al. (2015) proposed a design that incorporated a Savonius rotor inside curved guide plates, which achieved a C_PMax of 0.52 at a TSR of 1.1. However, this design produced large wakes behind the rotor, which could affect the operation of other turbines in a wind farm. Although these methods significantly improve performance, they compromise the simplicity and omnidirectional nature of the Savonius rotor.

Ideal augmentation methods can improve rotor performance while keeping all the Savonius advantages intact. One of the methods is the addition of a check valve to the rotor blades. This method allows a portion of wind freestream to pass through the returning-blade convex side, reducing the drag force and lowering the negative torque. Furthermore, valve addition can be a complementary method with the potential to further enhance an already high-performance Savonius rotor with a modified blade.

Rajkumar and Saha (2006) first investigated the valve-addition method by placing slot-shaped aluminum hinged-type valves on a conventional and twisted Savonius rotor. Saha Thotla, and Maity (2008) later used a thin fabric called Rexine as the valve material instead of aluminum and conducted experiments using an open jet wind tunnel. The findings showed that the two-stage, valve-aided Savonius rotor had a 19% higher C_PMax than a conventional rotor. Both of the aforementioned studies used a fixed-size valve placed at the center of the blade. Amiri and Anbarsooz (2019) investigated the effect of valve location on a conventional Savonius rotor's performance. The study used a rectangular pivot-type valve made from the same material as the rotor. The valve's opening area ratio (OAR), which is the ratio between the area of the valves and the rotor's frontal area (D.H), was fixed at 0.33. This pivotal valve was placed near the rotation axis as well as in the middle and at the tip of the rotor. The results showed that the valve at the tip location yielded a 20.8% performance improvement. Furthermore, Borzuei, Moosavian, and Farajollahi (2021) performed a numerical experiment on the effect of adding a rectangular pivot-type valve to the static torque coefficient of a Savonius rotor with three blades. The results showed a 14.5% improvement in the static torque coefficient. This was indicated by the large pivot valve with an unlimited opening angle and counter-clockwise opening direction placed at the center of the rotor blade.

        Previous studies have shown that using a part of the rotor blade as a pivot-type valve can have a positive effect (Borzuei, Moosavian, and Farajollahi, 2021; Amiri and Anbarsooz, 2019). However, a large OAR can have a negative impact on the structural integrity and manufacturability of the rotor. The slot-shaped hinge-type valve proposed by Saha, Thotla, and Maity (2008) has better structural integrity and manufacturability. However, previous studies on Savonius rotors with hinge-type valves did not investigate the effect of valve locations and OAR on the rotor's performance under different wind speeds. This study aims to fill this gap. Moreover, in contrast to the slot-shaped valve used in by Saha, Thotla, and Maity (2008) this study employs a circular valve to improve the manufacturability of the rotor blade.

2.1. Studied Rotor Design and Manufacturing

        This study investigated a single-stage Savonius with two semi-circular rotor blades. All the studied rotors have the same design and dimension. The valves were placed near the rotor rotation axis (A), at the center of the rotor blade (B), and near the rotor tip (C). Moreover, the study used three increasing OAR for each location to study ten rotors, including the non-valve-aided rotor. Tables 1 and 2 show the geometrical dimension and rotor specifications, while the valve location and arrangements are depicted in Figure 2a.

The rotors were manufactured using 3D printers and PLA+ material with a thickness of 2.5mm. Each rotor was divided into lower and top plates and rotor blades, a design chosen to reduce manufacturing time and material use while allowing for quick reconfiguration. The valve body was made from synthetic leather pasted onto the rotor blade. Figure 2b shows the manufactured rotors.

Table 1 Geometrical dimension of the studied Savonius semi-circular rotor

Table 2 Studied rotors identification and specification

Parameter	Rotor Identification
	SCNV	SCA1	SCA2	SCA3	SCB1	SCB2	SCB3	SCC1	SCC2	SCC3
Valve Location	N/A	Near the rotation axis			At the blade center			Near rotor tip
Valve Qty	0	1	2	3	1	2	3	1	2	3
OAR	0	0.02	0.04	0.06	0.02	0.04	0.06	0.02	0.04	0.06

Figure 2 (a) Valve locations and arrangement on the rotor; (b) 3D printed rotors     

2.2. Experimental Setup

      The experiments were performed using the low-speed open jet wind tunnel at the Institut Teknologi Bandung. The wind velocity at the tunnel exit () could be adjusted up to 10 m/s using a variable speed drive. The averaged wind velocity at the tunnel exit was measured using a calibrated hotwire anemometer. The experiments were conducted at a Reynolds number of 73,000, 86,000, and 99,000. The rotor axis was positioned 300 mm away from the wind tunnel exit. The open test section in which the rotor was situated measured 1,000 mm x 600 mm, resulting in a blockage ratio of 12%. This value is low for an open-type wind tunnel test and does not require a blockage correction factor (Gonçalves, Pereira, and Sousa, 2022; Van Bussel et al., 2004). The overall setup of the experimental study is depicted in Figure 3a.

        The study used a DYN-200 rotary torque meter to measure the torque (T) and the rotor's rotation speed (N). The torque meters measurement was calibrated using known torques, while the rotation speed was checked against a calibrated optical tachometer. The meter accuracy was specified as ±0.1% and ±1 rpm for the torque and rotation speed, respectively. Furthermore, the torque meter output was connected to a laptop via a Labjack U6 data acquisition module. A hysteresis brake and a DC generator were combined to simulate a dynamic load. At each test point, measurements were taken at a sampling rate of  4 Hz for 15 seconds. Figure 3b shows the instrumental setup.

Figure 3 (a) Overall experimental setup; (b) Instrumentation setup

2.3. Data Reduction

    The study measured the freestream wind velocity (), rotor torque (T), and rotor rotational speed (N). The rotor was operated under a steady wind velocity and constant load during a dynamic test. The torque and rotation speeds were measured for a fixed time and then averaged. The load was then varied to obtain the rotor's performance characteristics. The torque loss due to bearing frictions cannot be ignored at higher rotational speeds. Therefore, a separate experiment was conducted to obtain the frictional power loss characteristic. The dimensionless rotor tip speed ratio (TSR) was then calculated using Equation 1:

where D is the rotor diameter (m),  is the wind velocity (m/s), and  is the rotor angular velocity (rad/s) calculated using Equation 2: 

where N is the measured rotor rotation speed (rpm). The dimensionless rotor power coefficient C_P is expressed using Equation 3:

where TSR is the tip speed ratio given by Equation 1, and  is the dimensionless rotor torque coefficient calculated using Equation 4:

where T is the measured torque (N.m), H is the rotor height (m), and  is the air density in (kg/m³).

    The uncertainty related to the measurement is calculated using the root-sum-square method (Wheeler and Ganji, 2009). The total uncertainties  for the measured quantities (T, N, and ) are calculated using Equation 5: 

where is the systematic uncertainty taken from the instrument's accuracy specification, and is the random uncertainty based on the standard deviation of the mean. The uncertainties for derived quantities  are then calculated using the propagation of error method shown in Equation 6:

where is the total uncertainty for the derived quantities, and is the sensitivity coefficient of derived quantities R with respect to variable  The confidence level associated with the uncertainties was chosen at 95%. The total uncertainties for the TSR,  are 2.3%, 4.0%, and 6.3% respectively.

3.1. Conventional (Non-Valve-Aided) Savonius Rotor Performance

Figure 4 The effect of Reynolds number on the Savonius rotor (a) coefficient of power; and (b) coefficient of torque

The results suggest that the power coefficient increased as the Reynolds number increased. The C_PMax value increased by 14.6% as the Reynolds number increased from 73,000 to 99,000, which is consistent with the findings of Kamoji, Kedare, and Prabhu (2009). The increase in C_PMax was attributed to the delayed separations at the rotor blades at higher Reynolds numbers or wind velocities (Kamoji, Kedare, and Prabhu, 2009). Similarly, the C_TMax also increased with the Reynolds number. The C_TMax values for Reynolds numbers 73,000, 86,000, and 99,000 were 0.284, 0.304, and 0.323, respectively. The C_T peaked at a tip speed ratio of 0.4 and decreased almost linearly as the tip speed ratio increased, which was observed for all three Reynolds numbers.

3.2. Effect of Valve Locations on the Rotor Performance

Figure 5 The effect of valve locations on SCNV, SCA1, SCB1, and SCC1 rotor (a) C_P vs TSR at Re=73,000; (b) C_P vs TSR at Re=86,000; (c) C_P vs TSR at Re=99,000; (d) C_T vs TSR at Re=73,000; (e) C_T vs TSR at Re=86,000; and (f) C_{T vs} TSR at Re=99,000

The rotors with a single valve performed better than the non-valve-aided (SCNV) rotors in all three locations and for all Reynolds numbers. However, the magnitude of C_PMax improvement decreased as the Reynolds number increased. For instance, the SCC1 rotor yielded an 8.7%, 7.0%, and 5.2% C_PMax improvement when tested at Reynolds numbers 73,000, 86,000, and 99,000, respectively. This condition may be caused by the valve's presence on the blade surface disrupting the flow-separation delay that normally occurs when the Reynolds number increases. Adding a single valve did not significantly affect the tip speed ratio corresponding with the C_PMax. The tip speed ratio range related to the C_PMax is 0.85 to 0.87 compared to SCNVs 0.84 to 0.86. A similar observation was made by Amiri and Anbarsooz (2019).

The rotor with a valve near the rotor tip (SCC1) consistently showed higher C_PMax than the other two for the three Reynolds numbers. This result is consistent with Amiri and Anbarsooz (2019) that the negative torque reduction by the valve at the tip has a greater impact because the air flows through the valve at a higher rate. In contrast, the rotor with a valve placed near the rotation axis (SCA1) produced the least improvement. A valve at this location is too close to the moment axis, resulting in the lowest negative torque reduction. Placing a valve near the rotation axis also disturbs the overlapping-gap flow that is supposed to energize the returning blade