Drift is ubiquitous in heliostat fields, and may be caused by diverse geometrical inaccuracies during heliostat installation and operation. This phenomenon is studied for three important primary errors in the present paper: Angular offset in the drive mechanism, pedestal tilt, and canting error. Each error produces characteristic signatures, but there is a diversity of behavior depending on the error parameters and location of the heliostat. The variation of the extent of drift curves is studied as a function of distance, for fixed error parameters. It is found that, in general, this extent is not proportional to distance, except for far heliostats, and depends on a complicated manner on the different parameters involved. Moreover, even though the extent of drift curves becomes proportional to distance for far heliostats, the convergence is very slow, and very variable with the error parameters.

Several error sources can contribute to the global tracking error of heliostats. These sources can be, for instance, angular offset in the reference position of the tracking mechanisms, imperfect leveling of the heliostat pedestal, lack of perpendicularity between the tracking axes, lack of precise clock synchronization. All these possible errors are characterized by angles that have very specific numerical values for each heliostat in a central receiver installation. However, they are intrinsically random in nature, and the errors in different heliostats are independent from each other. In principle, the overall drift behavior of the heliostats can be characterized by a statistical distribution of tracking errors. This global distribution characterizes the angular deviation of the heliostat normal and is used in ray tracing simulations of heliostat fields. It is usually assumed to be Gaussian, although some authors argue in favor of other types of distributions. In the present work, the dependence of the global tracking error distribution on the above mentioned primary error sources is investigated by means of Monte Carlo simulations. Random values are assumed for the different error parameters, and the resulting global tracking error distributions are evaluated for different times of the year for a heliostat field