Investigation of inflow wind conditions and load performance of small wind turbines in the built environment

One of the fastest growing renewable electricity technologies has been wind energy, in both large-scale wind farms as well as in smaller distributed and community wind applications. Small wind turbines have traditionally been designed for rural areas over open terrain as off-grid system. There has, however, been an increasing trend towards the installation of wind turbines in non-open terrain, such as in urban areas and in mountainous regions as on-grid system, where the turbulence levels are higher and wind speeds are lower. The design, installation and performance of small wind turbines in urban areas are areas lacking in guidelines and procedures, despite evidence of failure of small wind turbines at urban sites and concerns over reliability and safety. This thesis is designed to address this lack of guidelines and procedures and can be considered as an investigation in two distinct but related parts. Firstly, it is an investigation of aspects related to wind resource assessment in urban areas through both wind monitoring programs and computational flow simulation. This research was designed in order to make a contribution to a document planned by the International Energy Agency (IEA Task 27) entitled ‘Recommended Practice for Design of Small Wind Turbines in the Built Environment’. Secondly, it is an investigation of the turbulence models currently used in the design standard for small wind turbines (IEC61400-2), their relevance for urban sites and their prediction of small wind turbine loads. This research was designed in order to make a contribution to the next revision of IEC61400-2, planned for 2019. In the first part of the study, the impact of sampling rate and averaging period on a study of the turbulent wind regime in the built environment was investigated. It was shown that choice of sampling rate did not significantly influence the values of turbulence intensity and power spectral density. Unlike sampling rate, changing the parameter of averaging period significantly affected the results of calculated turbulence intensity, including the value of the characteristic intensity I15. Generally, the results of this study showed that the turbulence intensity and power spectra of the longitudinal and lateral components of wind over a rooftop are sensitive to the choice of averaging period. Due to this sensitivity, an averaging period of at least 10 minutes was suggested for rooftop wind monitoring to avoid underestimating values of turbulence intensity and turbulence power spectra. In summary, a 10 Hz sampling rate and a 10 minute averaging period yield upper estimates for values of turbulence intensity and turbulent power spectral density. Taking this conservative approach may be the best strategy if looking to ensure that the wind resource assessment accurately captures the inflow to the wind turbine so that the turbine can be designed to handle both the loading and resonance due to turbulent gusting. The first part of the study also focused on the assessment of the combination of a ANSYS CFX computational fluid dynamics software with wind atlas software (WAsP) to form a wind energy resource assessment tool to test the feasibility of installing small wind turbines on the roof of a building. The roof of the Bunnings Ltd warehouse at Port Kennedy, Perth Australia was chosen as the test site. The results of the study showed that the combination of ANSYS CFX with WAsP proved to be a promising tool for wind resource assessments for small wind turbines on the roofs of buildings. As the Bunnings Ltd. warehouse building is surrounded by different obstacles in terms of shape and height, the results were strongly wind direction dependent. Also, the output of the wind stochastic simulator TurbSim was applied to predict the wind velocity and Turbulent Kinetic Energy (TKE) profile at the inlet of the CFD domain. The results showed that use of TurbSim can improve ANSYS CFX predictions for wind direction sectors where the terrain is not so complex as to cause blockage, separation and recirculation of wind flow. The second part to the thesis investigated the extent to which the von Karman and Kaimal turbulence spectral models, as presented in IEC61400-2, are appropriate for use in the design of small wind turbines installed on the rooftop of a building. Research presented in this thesis shows that Kaimal spectra predicted the trends of all wind components better than the von Karman model. This was particularly true at high frequencies which, the literature suggests to be the important frequencies when it comes to fatigue loading on wind turbines. The results from this thesis suggested that there are some key parameters, such as hub-height, atmospheric stability and wind direction, which influence the shape and scale of the turbulence power spectra over the rooftop of a building and need to be taken into account when considering the inflow of a small wind turbine installed on a roof. A sensitivity study with respect to length scale showed that prediction of spectra at high frequencies could be improved by using smaller length scales in the current Kaimal model. As an approach to modelling turbulence power spectra for a rooftop site in the built environment, an adapted Kaimal approach was proposed that incorporates typical length scale ratios for that environment. The approach showed good agreement with measured data from the Port Kennedy site for all heights where there was sufficient measured data. The second part of the thesis also compared turbine blade load statistics for inflow turbulence fields based on the open terrain standard Kaimal spectra and measured turbulence spectra from a built environment site. The most frequently occurring winds in the built environment had low mean wind speeds and low turbulence and the predicted loads from measured spectra were less than that predicted by the standard. In this sense the current standard would appear to be a conservative estimate of loads at least for the majority of the time. But for extreme, high turbulent intensity winds, use of the measured spectra predicted isolated loading events around twice the magnitude of loads predicted by use of the standard spectra. The work suggested the need for improvements to the standard in order to model the non-Gaussian wind statistics that occur in extreme events such as sudden strong gusts.