安全与环境学报

针对海上溢油事故初期围控阶段应急调度中油膜动态特性与海域生态敏感差异协同不足的问题，提出了一种融合生态敏感度分级与油膜时变特性的协同调度方法。首先，基于生态脆弱性、社会经济价值及污染恢复能力构建环境敏感指数(Environmental Sensitivity Index, ESI),实现海域敏感等级划分与保护优先级确定；其次，结合扩散模型与漂移模型刻画油膜时变过程，动态更新围油栏需求与航行路径；在此基础上，构建面向围控作业的“动态路径规划-敏感区优先调度”模型，以调度成本和环境价值损失最小化为双目标。针对模型的NP-hard特性，设计融合ESI导向的种群预处理和变异参数调整、2-opt局部优化及自适应变异机制的改进遗传算法进行求解。以渤海、黄海海域溢油事故为仿真对象开展算例分析，结果表明：引入ESI分级策略后，小规模场景环境价值损失降低15.40%～19.87%,中大规模场景降低4.8%～8.2%;改进算法在寻优精度、收敛速度与稳定性方面均优于传统遗传算法。灵敏度分析表明，提升船舶航速与围油栏部署效率可显著降低环境损失，船舶容量主要影响调度成本，油膜漂移速度与成本及损失呈正相关关系。

This study aimed to optimize emergency dispatch for marine oil spill containment by jointly considering time-varying oil film diffusion and the spatial heterogeneity of ecological sensitivity. An integrated scheduling framework was developed to minimize operational costs and losses in environmental value during the early containment stage. First, an Environmental Sensitivity Index(ESI) was constructed by weighting ecological vulnerability, socioeconomic value, and pollution remediation capacity to classify marine areas into five sensitivity levels. Second, nonlinear diffusion and drift models were employed to characterize the temporal evolution of oil film diameter and position, facilitating dynamic calculations of containment boom demand and navigation distances. Based on these models, a dual-objective optimization model integrating dynamic route planning and sensitivity classification mechanisms was formulated. The model incorporated capacity constraints, time-dependent distances, deployment timings, and task assignment constraints. To address the NP-hard nature of the problem, an improved genetic algorithm was developed. This algorithm introduced ESI-oriented population preprocessing, 2-opt local path optimization, and an adaptive mutation mechanism guided by population diversity and high-sensitivity loss ratios. A dual-layer integer encoding scheme was utilized to represent routing sequences and fleet assignments, while a weighted fitness function was constructed after normalizing cost and environmental loss indicators. Simulation experiments were conducted using real spill data from the Bohai and Yellow Seas as well as extended Solomon benchmark instances. Comparative results demonstrated that the introduction of an ESI-based classification strategy reduced environmental value losses by 15.40% to 19.87% in small-scale scenarios and by 4.8% to 8.2% in medium-and large-scale scenarios. Ablation tests indicated that each enhancement module contributed to improved convergence speed, solution stability, and optimization accuracy. The proposed algorithm consistently outperformed conventional genetic algorithms and their variants. Sensitivity analysis revealed that increasing vessel speed and containment deployment efficiency significantly reduced pollution losses, while vessel capacity primarily influenced operational costs. These findings demonstrate that integrating ecological sensitivity classification with time-varying oil film dynamics can effectively enhance containment dispatch performance.