This work presents a two-stage optimization procedure for the conceptual design and operation of A-frame dry cooling systems for concentrated solar power facilities. First, the optimal geometry of the A-frame including sizing, number of fans and blade geometry, and unit parameters such as pipe length, configuration and number is determined. Finally, the operation of the system over a year for minimum energy consumption is computed. The geometry problem is formulated as a mixed-integer non linear programming (MINLP) problem. A tailor-made branch and bound algorithm is used to solve the complex non-linear programming sub-problems. The second problem consists of a multi-period MINLP. A fixed geometry is used to evaluate the usage of fans over time. The solution suggests an apex angle of 63 degrees, one row of 75 pipes of 13.5 m long with a diameter of 3.3 mm, and 4 fans are used but they only operate at full capacity during summer. This design allows reducing the energy required by 20% by using the appropriate pipe configuration and number. The unit consumes around 4% of the energy produced by the CSP plant that serves. It is a promising result that can be affected by plant layout and ground availability. (C) 2018 Elsevier Ltd. All rights reserved.