Multivariable Estimation of Wind Shear Profiles and Optimal Wind Turbine Energy Prediction

点击翻译按钮获取中文摘要

Wind energy is one of the major contributors to the renewable energy mix. This global industry has experienced significant, sustained growth in installed capacity over the past 15 years and although it experienced a slight decline in 2013, some experts predict that installed capacity will double by 2020. Competition between turbine manufacturers has created a downward trend in turbine costs, however increasing interest in distributed generation and sub-optimal sites (those with lower wind speeds, higher turbulence and/or shear), demand an accurate and precise prediction of wind speed and turbine energy output at hub height, a factor that requires a true estimation of wind shear. An estimate of wind shear can be produced using power or logarithmic law models based on anemometer data and vertical extrapolation then used to predict wind speed at the proposed hub height. It is generally recognized however that such methods suffer from high levels of uncertainty which impacts the overall project uncertainty. Financial institutions use P90, P75 or P50 (the probability of reaching predicted annual energy output 90%, 75% and 50% of the time respectively) to evaluate the risk associated with a potential project, a factor that impacts the ability to (a) obtain funding, (b) establish debt ratio and (c) leverage interest rates.***Current research has shown that wind shear at the turbine location must be accurately characterized in order to provide these probabilities however advancement in this area is limited with the use of meteorological towers. The introduction of remote sensing wind profilers, in this case, Sonic Detection and Ranging (SODAR) provide wind speed measurements at a range of heights, including heights that far exceed current meteorological towers, permitting wind shear calculations across the entire rotor disk. The potential of this technology has not yet been fully investigated or utilized for wind shear models. This research proposes the development of a wind shear model based upon SODAR data, observable site characteristics and advanced multi-variable modeling utilizing an artificial neural network (ANN). The rationale is that wind shear varies as a function of a number of complexities, many of which are non-linear, with a vertical distribution that needs to be characterized and modeled. Development and validation of this model is one of the expected practical outcomes of this project, contributing to the advancement of wind resource assessment which in turn will reduce uncertainty, the financial risk of wind energy projects.  This is expected to result in less expensive wind farms, enabling greater site diversification and enhancing distributed renewable energy, creating not only environmental benefits but social and economic benefits too.********