Estimation of Uncertainty in Automated Heliostat Alignment

Engineers seek to reduce the cost of solar-thermal central tower systems by reducing the cost of the heliostat field. Lower-cost heliostats can be achieved by accepting lower precision in manufacture and installation. Automated alignment methods can then be used to correct pointing errors due to misalignments after the heliostat is installed, to obtain and maintain the required pointing accuracy. The method under consideration here is the beam-target method, in which the location of the reflected spot on a target is measured repeatedly, and linear regression is applied to estimate various geometrically-based physical misalignment parameters to facilitate pointing error correction. In this study, a new statistical analysis of the automated alignment process including calculation of alignment uncertainty after the 'training' process is presented. In addition to correcting errors, this new analysis now allows evaluations to when there is enough training data gathered to be confident that, if the calculated corrections are applied, the beam and target will coincide within an allowable uncertainty. The relative value of training data gathered at different times of day and different times of year was also investigated. A large number of randomlygenerated misalignments are studied numerically. Worst-case misalignment combinations are found for an upper bound on the training data requirements for particular heliostats to be subsequently determined.