基于元胞自动机和数据同化的溢油模拟

The rapid development of the economy makes the growing national demand for oil. Maritime activities, such as, ocean oil transportation and offshore oil exploitation, are prompted more and more frequent. Therefore, the probability of oil spill also will increase. If oil spill cannot be controlled in time, it will seriously harm the national offshore marine ecological environment. Therefore, it is necessary to make a series of policies for oil spill. Before making policies, the government should predict the accurate dynamic changes of oil spill. However, the traditional methods are often used empirical methods, simplify the oil spill process and use static model parameters, it is difficult to obtain high temporal and spatial resolution of the oil spill. So the project intends to combine cellular automata model (CA), data mining and data assimilation to obtain the continuous and accurate oil spill change process in time and space. The content includes: 1) oil spill CA model based on data mining, 2) oil spill CA model based on data assimilation, 3) oil spill CA model based on ensemble Kalman filter and multi-parameter estimates. The project will help to improve oil spill emergency response capacity and technical level and provide a scientific means to protect the ecological environment.

频繁的海上石油开采和石油运输使得发生溢油事故的概率增大。海上溢油事故一旦发生，若得不到及时控制，必将严重损害海洋生态环境。因此，在溢油发生后，应准确地模拟并预测溢油的时空动态变化以控制溢油范围、保护海洋生态环境。元胞自动机（CA）是模拟溢油时空动态变化过程的有力工具，但目前这方面的研究、理论和技术方法尚处于初始阶段。本项目拟建立基于元胞自动机和数据同化的溢油模拟平台，结合元胞自动机模型、数据挖掘和数据同化等技术获取在时间和空间上均连续的溢油变化过程，为政府制定有关的应急方案提供科学的决策支持。具体内容包括：1)构建基于数据挖掘的溢油CA模型；2)集成数据同化的溢油CA模拟；3)耦合集合卡尔曼滤波和多参数估计的CA模型。

溢油事故频繁若得不到及时控制将严重损害海洋生态环境，如何在溢油发生后，准确地模拟并预测溢油的时空动态变化以控制溢油范围、保护海洋生态环境是项目的研究核心。本项目拟建立基于元胞自动机和数据同化的溢油模拟平台，结合元胞自动机模型、数据挖掘和数据同化等技术获取在时间和空间上均连续的溢油变化过程，为政府制定有关的应急方案提供科学的决策支持。具体内容包括：1)构建基于数据挖掘的溢油CA模型；2)集成数据同化的溢油CA模拟；3)耦合集合卡尔曼滤波和多参数估计的CA模型。实验表明模拟结果的总精度可高达96.6%，Kappa系数高达0.899。通过敏感性分析发现，用于训练因子权重的采样比例越高，模型的测试精度越高。但是采样比例越高，获取权重时的运算量越大。虽然采样比例有所变化，但是距离、海风和洋流等因子的权重是较稳定的，然而温度、盐度和常数的变化却相对较大。在对比实验中还发现，温度与盐度对模型的影响最小，其次是空间距离，对模拟影响最大的是海风和洋流因子，若不考虑海风与洋流的情况下，模拟结果会很快地偏离真实的观测结果，并无法模拟出漂移这一重要的溢油特征。