基于高斯过程替代模型的船舶能源系统建模与减排措施优选方法研究

Shipping contributes a large proportion of the global CO2 emissions. With the increasing concern about the global warming, it is important to reduce the ship emissions. Various mitigation measures have been proposed to reduce the total CO2 emissions from shipping. For instance, International Maritime Organization (IMO) has identified more than 50 mitigation measures, including both operational and technical measures. Despite the sheer number of mitigation measures available, the economic feasibility or the cost-effectiveness (CE) of these measures need to be further assessed. Therefore, it is important to rank and select the most cost-effective measures to reduce carbon emissions. The marginal abatement cost curve (MACC) is one of the commonly used approaches to rank the mitigation measures. IMO has applied this method to comprehensively evaluate the cost-effectiveness of 22 measures. Although MACCs are commonly applied for policy making, this method has several disadvantages. The first is that this method is not appropriate to rank the negative CE measures (i.e. the mitigation measures with negative cost). The second is the lack of uncertainty assessment for this type of method. The third is that the method does not consider the correlation among the mitigation measures. Due to the complexity of the ship energy system, some mitigation measures may be correlated with each other, which may significantly influence the ranking and selection results. To handle these problems, we propose a Gaussian process metamodel based multi-objective optimization method to rank and select correlated mitigation measures. This method has several advantages. Firstly, a Gaussian process metamodel is developed to represent the original complex ship energy system, which makes the analysis much more efficient. Secondly, the method takes into account the correlation among mitigation measures from system perspective. Thirdly, a multi-objective optimization approach is proposed to rank mitigation measures considering both abatement and cost objectives. Fourthly, an uncertainty quantification method is incorporated into the optimization method to rank mitigation measures under uncertainty. In summary, the proposed method is a general method which can be widely applied to rank and select correlated mitigation measures in different areas.

船舶排放会对全球气候变化和海洋环境造成持续的影响，为控制船舶排放，国际海事组织分析了超过50种船舶技术类和运营类减排措施。由于受到经济、环境等多方面制约，这些减排措施往往无法同时实施，且需要考虑减排能力和成本等多个目标，因此多目标下减排措施的优选问题尤为重要。国际海事组织绘制了22种减排措施的边际减排成本曲线，为减排措施的优选提供决策依据。然而，基于边际减排成本的优选方法存在许多不足之处，如无法优选‘负成本’减排措施、缺乏不确定性分析以及未考虑减排措施间的相关性。针对这些问题，本项目提出了基于高斯过程替代模型的多目标优选方法，从系统的视角解决具有相关性减排措施的优选问题。该方法具有如下优点：其一，高斯过程替代模型的构建代替了原始复杂的船舶能源系统，能够提供快速有效的分析；其二，采用系统方法分析了减排措施间的相关性；其三，考虑了多目标下的减排措施优选；其四，提供了不确定环境下的优选方法。

为实现到2050年海运减排50%的目标，国际海事组织提出了大量的减排措施，其中技术类和运营类减排措施是近中期实现船舶减排的主要措施。虽然有众多船舶减排措施，但是受成本、环境等多种不同因素的限制以及减排措施之间的相关性影响，不可能同时实施所有减排措施。此外，对减排措施的评估往往存在很大的不确定性。因此，考虑不确定性的不同类型减排措施多目标优选对于制定合理减排决策尤为重要。. 针对船舶减排措施的优选问题，项目主要研究了减排措施的能耗与排放预测模型、成本分析模型和多目标优选方法。能耗与排放预测模型用以分析减排措施的环境效益。项目结合了船舶能源系统的特性，综合考虑了船舶实测数据和计算机仿真数据，建立了基于多数据源的多层级高斯过程替代模型，同时还建立了反向传播神经网络和贝叶斯神经网络模型，对比分析不同减排措施的减排效果。成本分析模型用以分析减排措施的经济效益。项目分析了不同减排措施在不同船舶类型中的成本构成，建立了减排措施全生命周期成本的核算模型，评估了各减排措施的实施成本。以减排措施的减排效果和减排成本为目标，项目基于成本效益分析和帕累托最优提出了边际成本效益分析方法，对减排措施进行多目标优选。同时考虑到输入的不确定性，项目进一步提出基于统计分析和蒙特卡罗方法的不确定性量化方法，用以分析减排措施优选的不确定性。针对具有相关性的减排措施，项目提出了基于成本偏好的系统性排序和选择方法，避免了同时选择具有强相关性的减排措施。. 本项目建立的基于高斯过程和神经网络的船舶能耗和排放预测模型能够为船舶减排措施环境效益的评估提供有效的分析方法；基于能耗与排放模型和成本模型提出的多目标优选方法能够为航运企业合理选择减排措施提供科学的方法，为政府制定减排措施实施政策提供理论的依据；项目的分析结果为航运业实现减排目标提供合理可靠的减排措施选择决策依据。