Development of a vertical axis wind turbine farm by using wake energy

The use of renewable energies such as wind energy, solar energy, wave energy has lately gained more importance in order to reduce and ultimately terminate the worldwide energy problem and the negative impacts of conventional energy sources on the environment. At the beginning of this study, the increase in the use of renewable energy sources in recent years, and more particularly the increase and the capacity of wind energy usage are investigated. The horizontal axis turbines and the vertical axis turbines used in the production of wind energy are compared. The advantages of vertical axis wind turbines compared to their horizontal axis counterparts are emphasized and their working principle is explained. In the subclass of vertical axis wind turbines, the Darrieus and Savonius wind turbines are compared in different aspects such as production, maintenance, repair, efficiency, installation, cost and ability to start rotating with wind's effect (static torque). In the end of the initial study, Savonius wind turbine is thought to be the most convenient and considered suitable for the idea of the wind turbine farm which can be built in Istanbul, on the roof of the buildings in the city, which forms the basis of this thesis. Since the minimum wind speed for an economic investment for wind turbine farm is 7 m/s, firstly; wind source data of the city of Istanbul is considered. By using the Wind Energy Potential Atlas shared by the Ministry of Energy and Natural Resources of the Republic of Turkey, areas where wind turbine farms can be established, or in other words, areas with a wind speed of 7 m/s are identified. It is assumed that the wind turbine farm is built on the roof of a building located in these areas and that the roof has an area between 100 and 200 m2. Following the initial study, a Savonius wind turbine model is created in Ansys Fluent software. This turbine is modeled by using the same geometric features of the turbine that was used in the experiment of which the results were discussed in article Sheldahl et al. [4]. There are various studies in the literature about each of the parameters that make up the turbine geometry, and these parameters affect the performance of the turbine. However, in the model validation phase of this study, the number of blades, blade diameter, blade length, aspect ratio, blade shape, endplate geometry, and overlap ratio of the Savonius wind turbine used in the experiment are taken as the same. The aim of this phase of the study is to see that the analysis of the computational fluid dynamics made with the model created in the Fluent software yields results that are close to the results of the experiment conducted with Savonius turbine. This way it is assured that the subjects that could dramatically change the result of the analysis are selected correctly such as: the size of the domain created for the dynamic analysis of the model created in Fluent software is sufficient, the domain and the turbine (cell zone) are divided into a sufficient number of elements, the elements of very small size around the turbine blades spread into the airflow area with a correct ratio, the y + value representing the distance of the closest node to the turbine blade is in the correct range, the model selected among many turbulence models is appropriate, and the boundary conditions. It's been a long-term study to find out which solution method gives the most accurate result at this point. The closest results to experimental results are found by trying different methods such as solving turbulent kinetic energy and turbulent dissipation with second order equations and etc. Performance evaluations of wind turbines are done with power coefficient Cp. It is also defined as sum of the power that is generated by the turbine at nominal speed. Cp equals CT (torque coefficent) times TSR (tip speed ratio). TSR's rough definition would be the ratio of blade tip speed to wind speed. Savonius wind turbines operate in the range of 0.2 to 1.6 TSR, as indicated in different articles mentioned in the this study's content in following pages. The CFD analysis studies that took place in the literature has not gone beyond this range. It should not be forgotten; it is assumed that there is no internal friction of the gas in ideal gas dynamics solved by Euler equations. But in reality, fluids have internal friction due to viscosity. By the influence of the viscosity and the movement of the flow, some of the kinetic energy is converted into heat energy. Since this will change the result of the problem, the viscosity should not be ignored while the boundary layer motion or turbulent flow problem is solved. Since the turbulent flow is studied in this thesis, Reynolds-Averaged Navier Stokes equations are used in CFD analysis studies. The Navier Stokes equations can be seen as Newton's second law of motion for fluids. The Navier-Stokes equations represent conservation of momentum and are solved together with the continuity equation that represents conservation of mass. In engineering problems where the Reynolds number is high, inertial forces are much larger than viscous forces. These problems are actually time dependent, and it is necessary to create very fine mesh in the software for finite volume method so that even the smallest eddies in the flow can be solved. However, it is not possible with today's computers to operate such a simulation with Navier-Stokes equations. Therefore, the Reynolds-Averaged Navier Stokes (RANS) equations are used to obtain time averaged velocity and pressure fields. And this way problems can be solved with relatively coarse mesh. The turbulent viscosity parameter in the RANS equations, which represents the effect of time-dependent fluctuations of velocity, is solved using some different turbulence models. Frequently used models are k - epsilon, k - omega, Transition SST and Reynolds Stress models. In order for the CFD analysis results to converge, some requirements must be met for each of these models. The mathematical structure and conditions of use of these turbulence models are examined. The validation study continues with three turbulence models determined to be appropriate as a result of the study. CFD analysis of single rotor model, which was prepared for the validation of boundary conditions, number of elements and solution method, is done with TSR values in the range of 0.4 – 1.4 as in the article Sheldahl et al. [4]. Different turbulence models are studied in detail which are: k - epsilon / realizable / scalable wall functions, k - epsilon / RNG / scalable wall functions and k - epsilon / realizable / standard wall functions. For each of these turbulence models, individual CFD analyses are performed for TSR values between 0.4 and 1.4. And CT (torque coefficient) values, which are the output of the analyses, are multiplied by the relevant TSR and in the end Cps are found. Finally, the change of power coefficients (that are obtained out of 3 turbulence models) with respect to tip speed ratios is examined on the graph and compared with experiment results. As a result of this comparison, the turbulence model to be used in the CFD analysis solutions of the models that will form the wind turbine cluster is selected. In addition to comparing the power coefficients with the ones from experiment which gives us a chance to evaluate the turbine's performance, flow patterns around the blades while the rotor is at different rotation angles are compared with the pictures taken at experiment. The comparison is made with vectors that show flow directions and velocities. This way, in addition to the correctness of the analysis, the dynamics and the behavior of the flow are also examined. In parallel to these studies, definitions of some phenomena specific to the computational fluid dynamics are made in order to better understand the flow dynamics and how this problem is solved mathematically. Law of the wall, logarithmic law zone and buffer zone that are used in boundary layer problems are examined. Also the phenomena such as; wake effect, magnus effect, coupling effect, viscous forces, azimuth angle of the rotor, flow separation, turbulent kinetic energy, turbulent dissipation, turbulent intensity and turbulent viscosity ratio are defined and their effects on the flow are discussed. As a result of comprehensive literature review, when it is considered from the perspective of performance, manufacturability and economy; the turbine model to be used in the thesis was chosen as modified Bach type bladed rotor geometry. Features of this rotor type that has two blades such as aspect ratio of the blades, shape of the blades, geometry of the end plates, overlap ratio of the blades were taken from literature. In the first step, CFD is performed with modified Bach type rotor and 2% performance development is observed compared to semi-cylindrical model. The second step is running CFD analyses of two parallel and oblique Savonius rotors rotating in the same direction with same phase that have several different gap distances and orientation angles between them. Then observing how power coefficients are changing compared to single Savonius rotor. When all positionings considered, the optimum is when the second rotor is located at -75° and 0.1D distance to the first rotor. Average Cp=0.268 for two rotors whereas Cp=0.231 for single rotor. Therefore there is 16% development in performance. In the third step, number of rotors is increased to three and the change in Cp by the increase in number of rotor is observed. The result is: compared to turbine cluster with two rotor, first and second rotors' performances increase by 11% and 5% respectively whereas Cp=0.227 for third rotor which has 2% lower performance than single rotor. However the average Cp=0.269 for three rotor cluster which means that there is 16.4% development in performance compared to single rotor. In the last step, turbine clusters with five, six, eight and nine rotors in different positions are analyzed and it is observed that the lay-out which provides optimum performance with minimum area occupation is provided by a nine rotor cluster which contains three sub-clusters with there rotors in each. For this newly developed wind turbine farm that occupies approximately 160 m2, Cp=0.31 which shows as a result of this study that; the average Cp is increased by 34% in comparison to a single rotor.