Integrated multi-objective optimisation of the support structure of heliostats in concentrated solar power plants using a genetic algorithm

The optimisation of the support structure of heliostats in concentrating solar power plants is a fundamental task aimed at attempting to reduce the high levelised cost of energy (LCOE) of current configurations. In this work, an integrated multi-objective optimisation framework is presented, which relies on the combination of a lean and fast structural model with a genetic algorithm to simultaneously minimise both the overall mass of the support structure and the mean angle of rotation of the mirror surface, which directly affects the optical efficiency of the component. A particular feature of the proposed framework is that it represents an integrated solution, i.e., it allows to simultaneously optimise the main components of the heliostat support structure, i.e., the pedestal, the truss and the back support structure, assuming they are off-the-shelf components easily available on the market. The optimisation problem is set up selecting as design variables (i) the number of elements in the back support structure and (ii) the relevant characteristics of all the components considered, i.e., section shape and dimensions, according to the components commercial datasheets. At each iteration of the optimisation process, the structural model is fed with the current design variables values and, according to some computed aerodynamic loads, it allows evaluating the displacement and rotation of the points of interest within the mirror surface. An aerodynamic model present in the literature based on experimental wind tunnel tests is used to estimate the wind forces acting on the heliostat as a function both of the mirror inclination angle with respect to the ground and of the wind direction with respect to the mirror orientation. In this work, the proposed methodology is demonstrated on a realistic case study and the results commented in detail, highlighting possible future developments and the limitations of the framework.