SKYDOME CLOUD IMAGING

The aim of this project was to create a low-cost, reliable tool to capture images of the cloud cover in the sky at regular intervals. These images would then be used to predict expected power output from PV solar panels across a large area, in this case Northern Ireland. The devices would be distributed around the province near existing solar PV installations to create a large mesh that would ensure near total coverage.

The device needed to be easily manufactured using relatively simple tools, without too much manual effort an in a relatively short time frame to enable many of the devices to be manufactured. The devices would be connected to both mains power and to the internet, in this case combined using POE to simplify installation and cost. The internet connectivity enabled the device to share the images of cloud cover near instantaneously, therefore ensuring the ability to accurately predict power output in near realtime.

A number of prototypes were produced before the final design was chosen. Initially I opted to 3D print the enclosure using ASA filament, a material chosen due to its resistance to both water and UV rays, which was crucial as the devices would spend their lifetimes in highly exposed areas at the mercy of the elements.

To take the images of the sky, a low cost fisheye camera was chosen, connected to a low cost Raspberry Pi Zero. The Pi Zero had a number of desirable characteristics that made it ideal for the application, including its low cost, small package size, readily available accessories (namely cameras) with pre-made libraries to support them, and it's ability to be powered by a power over ethernet (POE) cable, which combined data transfer and power supply into a single easy to install cable.

The first iteration consisted of a 3D printed lower enclosure to house the camera and electronics, with a low cost transparent acrylic dome (made for CCTV cameras) used to protect the camera whilst allowing an unobstructed view of the sky. The dome was bolted to the lower enclosure with four M4 bolts and a flat rubber gasket used to seal the dome against the water and dust ingress.

The design was field tested and worked well, however the time needed to 3D print the enclosure was over 5 hours which was unacceptable as the design needed to be quickly and easily produced in large quantities.

Therefore version 2.0 was created, using low cost off the shelf PVC plumbing components to enclose the electronics, which allowed the 3D printed section of the enclosure to be reduced in size. This resulted in significantly reduced 3D print times of around 2 hours, which was an improvement on the previous design.

However, the print times were still too long to enable the parts to be made in the required quantities in the timeframe given, so the design was further refined.

After some research and further development, version 3.0 was created. A 3D printed mould master was created, from which a flexible silicone mould was created. Using this mould, the enclosure could now be cast in high strength urethane resin, which had exceptional resistance to both water and UV rays. The cycle time required to produce the enclosure was further reduced from around 2 hours to just 30mins, which enabled the enclosures to be produced in sufficient quantity in the required time frame.

The design was altered to suit this new manufacturing method by increasing the wall thickness to aid resin flow in the mould, and by incorporating draft angles into the design to aid in mould release once the enclosure had been cast. An improvised mould agitation method to remove bubbles from the resin mixture was created using an orbital sander.