Wind turbines, large and small, usually rely on the wind to accelerate them from rest to speeds where power production becomes possible. This is less of a problem for large wind turbines than for small, since judicious site selection for commercial wind farms ensures that large wind turbines see relatively high winds for considerable portions of the year. Small wind turbines, on the other hand, are usually sited where the power is required, which means that long portions of their operational lives may be spent idling in sub-optimal wind conditions. It is vitally important, then, that small wind turbines accelerate from rest quickly whenever strong winds start to blow so as not to squander this infrequent resource. The project documented in this thesis employed a multiobjective evolutionary algorithm (based on Differential Evolution) to "evolve" a Pareto optimal set of wind turbine blade designs, with each member offering a unique trade-off between the conflicting objectives of good starting performance and good peak power production. A first-order three-dimensional panel method with constant source and doublet distributions was employed to approximate the aerodynamic performance of candidate blade designs at low tip speed ratio (for starting performance) and at a high operational tip speed ratio (for peak power production). A simple helical wake model with iterative wake pitch adjustment was included to model the circulation shed from each blade's trailing edge. To allow geometry evolution to occur, a framework for complactly representing a huge range of possible three- dimensional blade geometries as bi-cubic B-spline surfaces was developed. Blades represented in such a way were subsequently discretised into quadrilateral surface panels for use by the panel solver. Results show that blade designs evolved by this method accelerate more quickly than current blade designs without suffering a degradation in peak power output. The most significant avenue for improved acceleration performance was via a reduction in second moment of inertia of candidate blades, which seems to involve a move towards more slender tape distributions and thinner "aerofoil" sections, resulting in geometries which nevertheless behave well aerodynamically.