基于随机动态规划和支持向量机的混合动力船舶能量管理策略研究

项目名称： 基于随机动态规划和支持向量机的混合动力船舶能量管理策略研究

项目编号： No.51209134

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 水利科学与海洋工程学科

项目作者： 黄细霞

作者单位： 上海海事大学

项目金额： 25万元

中文摘要： 随着全球节能减排呼声的日益高涨，混合动力船舶已经成为世界各国和船舶制造厂商的战略选择，混合动力船舶还具有军事意义。能量管理策略根据混合动力系统特性和运行工况合理分配能量，以获得最大动力续航能力、最大燃油经济性、最低排放以及平稳的驾驶性能。针对特定的船舶行驶工况,研究基于动态规划方法进行混合动力船舶能量管理策略的设计方法，得到混合动力船舶燃料经济性和排放的极限值，为评价控制策略提供理论依据；针对运行工况、驾驶习惯随机变化的特点, 研究基于随机动态规划方法进行混合动力船舶能量管理策略的设计方法,将驾驶员的功率需求建模为Markov链，通过对驾驶员的需求功率的概率转移矩阵进行在线建模；研究基于支持向量机动态规划（SVM-DP）方法的混合动力船舶能量管理策略的设计方法,SVM-DP采用支持向量机估计状态空间中状态的最优值函数的通用结构和相应的参数向量，从而大大减少计算量和存储空间。

中文关键词： 混合动力船舶；能量管理策略；动态规划；随机动态规划；支持向量机

英文摘要： With the growing voice of the global energy conservation, hybrid Electric ship has become a strategic choice of ship manufacturers all over the world and it also has military significance. According to the characteristics of the hybrid system and driving conditions, energy management strategy makes a reasonable distribution of energy so as to obtain maximum power endurance, maximum fuel economy, lowest emissions and a smooth driving performance. In view of specific ship driving conditions, a design method for energy management strategy of hybrid ship based on dynamic programming was researched and the fuel economy and emission limits of hybrid ship was obtained to provide a theoretical basis for the evaluation of control strategies. In relation to the random changes in the driving conditions and driving habits, a design method for energy management strategy of hybrid Electric ship based on stochastic dynamic programming was studied. The driver's power demand was modeled as a Markov chain, and the on-line modeling of the transition matrix of the probability of driver's power demand was built. Based on the support vector machine dynamic programming (SVM - DP), a design method for energy management strategy of hybrid ship was researched. The SVM-DP used support vector machine to estimate the general structure and t

英文关键词： hybrid power ship；energy management strategy；dynamic programming；stochastic dynamic programming；support vector machine