The Response and Optimisation of Hybrid Wind Turbine Towers

Abstract

This thesis investigates the response and optimisation of wind turbine tower structures with a particular emphasis on hybrid steel-concrete towers. The wind turbine blades with tower interaction are represented through equations of motion where mass, stiffness and damping properties have a time varying component. This thesis investigates the response and optimisation of these towers in a global and local sense through review of global tower top behaviour as well as the response at selected local locations from around the tower shell. Structural models of the tower are presented and used in analysing the response. An exact, closed form analytical model was developed using classical beam bending theory, with boundary and compatibility conditions imposed to generate a system of homogeneous linear equations with non-trivial solutions. Approximate, Finite Element models of the tower were constructed using both modified Euler-Bernoulli beam elements and also Reissner-Mindlin shell finite elements of varying numbers of degrees of freedom.

Two reduced order dynamic multi-degree of freedom (MDOF) models for an overall wind turbine assembly are then presented using a mixed formulation approach including Finite Element models incorporated into Euler- Lagrangian based systems. Discrete, global interpolation functions are used to reduce the total number of degrees of freedom (DOF) of the tower models to a selected reduced number of DOF. Continuous mode shapes are used to reduce the blade elements to selected DOF. Rotating blades are exposed to time varying load application through aerodynamic load and periodicity introduced by gravity. Axial effects through gravity and centrifugal stiffening act on the blades to vary their stiffness. Aerodynamic loading has been simulated using the modified blade element momentum (BEM) algorithm which accounts for the angle of attack, blade pre-twist, pitch angle and wind shear. Turbulence was generated from a Kaimal spectrum.

The closed form analytical model was used to assess tower free vibration response and MDOF dynamical models were used to investigate forced vibration response of towers of varying properties. The nacelle mass and hybrid interface height had the most significant impact on the first natural frequency of the tower. This was observed through free vibration response but also through frequency domain review of the forced response. Hybrid interface height was strongly correlated with the mean displacement but to a lesser extent on the velocity and acceleration response. Concrete compressive strength and structural damping properties had an influence on the tower response. The first and second natural frequency of the tower was slightly reduced when introducing and increasing a prestress into the models. The global forced vibration response of the tower showed insignificant change as a result of the introduction of prestress. The effect of prestress was more significant at a local finite element level on review of strain response in three principal directions. Separately, the frequency content of local finite element strain response was significantly different to the frequency content of the global tower top response. This was deemed to be due to the effects of combined deformation through all tower global DOF.

A methodology has been proposed for the optimisation of hybrid concretesteel wind turbine towers. This methodology incorporates the generalisation of free and forced vibration results of such towers using a configuration of Artificial Neural Networks, which are embedded within an optimisation algorithm which itself is a hybrid of a Genetic Algorithm and a Pattern Search Algorithm. Objective functions are defined in terms of both structural and non-structural criteria. Fundamental fore-aft frequency was maximised, peak tower displacement was minimised, as was a weighted sum of concrete and steel stress utilisation ratios. Levelised Cost of Energy (LCoE) was set as an objective and was minimised for a series of load cases and hub heights. Concrete and prestressed reinforcement contributed most significantly to the breakdown of LCoE. The Climate Change Potential (CPP) was also set as an objective to be minimised and followed similar patterns to the LCoE in terms of sensitivity to change in wind speed and height. Contributions to the overall CCP are much more equally spread than was the case in LCoE, with each contributing similar amounts. Multi-objective optimisation was carried out using the epsilon constraint method.

A method was proposed to utilise and process spatial strain and acceleration signals as a means of damage detection around the shell of the finite element model of the wind turbine tower. Processing involved passing the signals through the Discrete Wavelet Transform (DWT) signal processing technique. The spatial signals were all transformed and co-efficients were found for low and high frequency components. GIS spatial images were presented to represent aerodynamic loading and tower responses generated using BEM and the 11 DOF structural models described earlier in the thesis. By generalising the loading and response quantities as a function of spatially distributed environmental exposure conditions, it is possible to plot these loading and response quantities spatially.