Abstract:

To solve the problems of unnecessary energy loss and equipment failure of the Antarctic high-altitude automatic monitoring system caused by low temperatures in the Antarctic inland area, we propose a method to predict the ambient temperature on the next day and a strategy to select the corresponding heating power according to different temperature difference. The automatic monitoring system is composed of three parts: the surface new energy generation system, the surface observation cabin and the underground energy cabin. Through an analysis of the meteorological data on the surface of Princess Elizabeth Land, we find that temperature is the main factor affecting the normal operation of local equipment. By establishing thermal network equation, we obtain the relationship between heating power and temperature. Moreover, after comparing four kinds of neural network prediction algorithm, we select the most appropriate deep learning algorithm to forecast the future environmental temperature. Thus, we propose a control strategy that guarantees continuous observation of the surface observation cabin with low energy costs. The system was deployed in Princess Elizabeth Field during China’s 36th Antarctic Scientific Expedition. The system has been running stably for 10 months, which proves the practicability of the control strategy and the reliability of the equipment.

Key words: Antarctica, neural network, control strategy, application