| With the global warming and the melting of the Arctic ice,the opening of the Arctic shipping route is expected to be advance.It can not only provide China with a convenient access to the world,but also get rid of the severe dependence on the traditional shipping routes and the strategic stranglehold of the surrounding hostile forces.This provides a reliable guarantee for China’s economic,trade and energy security.However,the opening of the Arctic shipping route is confronted with many difficulties,especially the complicated natural and geographical environment of the Arctic sea has brought threats to the navigation safety of ships.Therefore,it is urgent to develop sea ice state warning and ice "weather" routing technology.Based on the two core problems of the uncertainty of short-term sea ice prediction and the weak applicability of the traditional route planning algorithm in the ice region,this paper combines the knowledge in the field of meteorology and oceanography with the deep learning algorithm to study the short-term prediction technology of sea ice concentration and sea ice thickness.Meanwhile,the ice environment-speed inference model is improved based on random forest algorithm,and a dynamic route planning algorithm suitable for arctic sea area is constructed.The main research results are as follows:(1)By combining the deep learning model with meteorological and oceanographic knowledge,this paper proposes improvement and development ideas from factors,physical constraints and models to improve the short-term forecasting effect of sea ice concentration and sea ice thickness.(2)Combined with the practice of ice navigation,the polar environmental risks are quantified,and the relationship between ship speed and environmental factors is constructed by using random forest algorithm,thus a segmented ice speed inference model is obtained.(3)Based on the short-term prediction results of sea ice,the dynamic route planning for ice region is realized by using D*Lite dynamic route planning algorithm. |
