No abstract

Combining local harvest patterns and global weather forecasts, e.g., cloud-cover forecasts, makes solar harvest predictions and online duty cycle adaptation more reliable. For this purpose, an energy and bandwidth efficient network-wide distribution of those forecasts is required. To meet this end, we propose compression methods for cloud-cover forecasts, so that they can be piggy-backed on regular network packets. We evaluate compression performance based on data collected from an online weather service for more than 14 months. We find that (i) cloud-cover forecasts can be compressed by up to 76%, (ii) fit into an average of 5 B for a one-day and 21 B for a seven-day forecast horizon, so that (iii) network-wide distribution leveraging, e.g., software acknowledgments used by prominent low-power data collection algorithms is achievable.