The receiver at the optical-to-thermo-hydraulic interface of a solar tower plant needs careful optimization in the design stage to guarantee maximum yield during operation. In this study, a methodology for the optimization of solar tower receivers is developed, which integrates optical, thermal and hydraulic aspects on one hand and considers dynamic operation coupled to the power plant on the other hand. The optimization objective function comprises a detailed receiver model and a transient system simulation approach. Operational limits, as well as aiming and defocusing strategies are integrated. An accurate representation of solar radiation transients is obtained with a sky discretization and a flux level interpolation approach. For annual yield assessment, an artificial neural network is trained with data from the detailed model, such that it accurately replicates the behavior of the latter. Evolution strategies are applied to find the global optimum for a receiver configuration with nine free parameters. The methodology is successfully demonstrated for a small-scale molten salt cavity receiver. Application to other systems like large-scale external receivers is dicussed.