Use of MEMs and optical sensors for closed loop heliostat control

The Helio 100 project at STERG (Stellenbosch Solar Thermal Research Group) aims to help reduce the cost of Concentrated Solar Thermal plants by deploying large numbers of small (1x2 m) low cost heliostats. One of the methods employed to reduce the cost of the heliostat field is to have a field that requires no site preparation (grading, leveling, vegetation clearance) and no expensive foundations or concrete pouring for each individual heliostat base. This implies that the heliostat pod frames and vertical mounts might be slightly out of vertical, and the normal method of dead reckoning using accurately surveyed and aligned heliostat bases cannot be used. This paper describes a combination of MEMs and optical sensors on the back of the heliostat, that together with a simple machine learning approach, give accurate and reproducible azimuth and elevation information for the heliostat plane. Initial experiments were done with an android phone mounted on the back of a heliostat as it was a readily available platform combining accelerometers’ and camera into one programmable package. It was found quite easy to determine the pointing angle of the heliostat to within 1 milliradian using the rear facing camera and correlating known heliostat angles with target image features on the ground. We also tested the accuracy at various image resolutions by halving the image size successively till the feature detection failed. This showed that even a VGA (640x480) resolution image could give mean errors of 1.5 milliradian. The optical technique is exceedingly simple and does not use any camera calibration, angular reconstruction or knowledge of heliostat drive geometry. We also tested the ability of the 3d accelerometers to determine angle, but this was coarser than the camera and only accurate to around 10 milliradians.