This work proposes a semi-empirical framework to predict the noise of wind turbines with serrated trailing edge blades. The framework is employed for studying the reduction of the noise of the SWT 2.3-93 benchmark wind turbine. The framework is verified against field acoustic measurements of the real wind-turbine model and of noise reduction measured for airfoil geometries with serrated trailing edges. Two different serration design strategies are proposed, respectively one with the same serration geometry along the blade and one with serrations scaled with the local boundary-layer properties along the radius. Results show the predicted noise reduction obtained with each of the add-ons and explore the benefits of tailoring the design of the serrations according to the varying flow conditions along the blade span.