基于弹性周期时空预测的北极通航环境风险预测预报研究

近年来随着全球气候不断变暖，北极航道船舶通航逐渐成为现实，但北极海区环境恶劣商船航行面临诸多风险，目前仍缺少对北极航道风险预测的有效方式。本课题首先分析诸多北极航行自然环境风险因素及风险评估模型，搜集数据计算北极长时间序列风险量化和分级数据集。其次，采用多种合适的时空分析方法，从多尺度周期角度评估并分析北极区域航道环境风险时空格局演化规律和趋势。再次，结合深度学习时空预测的理论方法，研究循环卷积网络记忆机制和时空注意力组合方式，捕捉空间区域内不同趋势和周期特征的各元素强度的相关性及其随时间的演变规律，建立基于深度学习的北极航道风险预测预报模型。模型可以自动学习高维时空数据中隐含的趋势和周期依赖关系，兼顾气候变化长期趋势、年度周期季节变化、以及短期变化趋势，自适应地从历史序列中抽取弹性周期和趋势特征以提高时空预测的准确性，最后结合历史数据和时空格局变化趋势分析模型有效性。

In recent years,with the global warming, the navigation of ships in polar channels has become a reality. However, the Arctic sea environment is poor, and there is still a lack of effective methods for the risk early warning and prediction of Arctic channels. This project first analyzes many natural environmental risk factors and risk assessment models of Arctic navigation, collects data and calculates the Arctic long-time series risk classification data set. Secondly, a variety of appropriate spatiotemporal analysis methods are used to evaluate and analyze the evolution law and trend of spatiotemporal pattern of environmental risk classification of Arctic waterway from the perspective of multi-scale cycle. Thirdly, combined with the theory and method of deep learning spatiotemporal prediction, this paper studies the memory mechanism of cyclic convolution network and spatiotemporal attention combination model, captures the correlation of each element strength of different trends and periodic characteristics in the spatial region and its evolution with time, and establishes the Arctic waterway risk prediction model based on deep learning. The model can automatically learn the implicit trend and cycle dependence in high-dimensional spatio-temporal data, take into account the long-term trend of climate change, annual cycle, seasonal change, and short-term change trend, and adaptively extract elastic cycle and trend features from historical series to improve the accuracy of spatio-temporal prediction. Finally, the historical data and spatio-temporal pattern change trend analysis model are effective.