【作者】 李明伟；

【导师】 康海贵； 周鹏飞；

【作者基本信息】 大连理工大学 ， 港口、海岸及近海工程， 2013， 博士

【摘要】 港口吞吐量预测是进行港口体系结构优化和基础设施建设的基础,对于合理确定港口规划布局、基础设施投资规模、集疏运系统的建设起着重要的作用。泊位、岸桥作为岸线的稀缺资源,其是否能够得到科学合理的分配和调度,对于提高运营期港口生产效益和服务水平具有重要的现实意义。在港口吞吐量预测和泊位-岸桥分配过程中,模型参数的确定以及泊位-岸桥分配等NP-hard问题的求解直接关系到优化模型的可行性和有效性,智能算法的提出为上述问题的解决提供了有效途径,但任何一种智能优化算法都不是完美的,受自身结构的限制都存在一定的缺陷。为更好求解港口规划与运营管理中的优化问题,本文将粒子群算法(Particle Swarm Optimization, PSO)、Cat映射和云模型进行有机结合,提出混沌云粒子群混合优化算法(Chaos Cloud Particle Swarm Optimization, CCPSO),并将其应用于我国港口规划管理中,对其在港口吞吐量预测和泊位-岸桥分配中的应用进行探索和研究,具体研究内容如下：1)通过对Cat映射的混沌特性分析,指出Cat映射具有更好的混沌特性,因而将其引入到混合优化算法中,用于对粒子群中较差个体的混沌扰动。考虑到PSO算法易陷入局部极值和进化后期收敛速度慢,而混沌映射具有更好的遍历性以及云模型的随机性和稳定倾向性优势,因此,通过引入混合控制参数mix_gen和种群分配系数pop_distr,将PSO算法、Cat映射和云模型三种算法进行有机结合,提出CCPSO算法,以期发挥三种算法的各自优势,提高优化性能。利用经典测试函数对CCPSO算法中的混合控制参数mix_gen和种群分配系数pop_distr的取值对优化性能的影响进行了分析,给出了两参数在应用于不同优化问题时的建议值。通过对CCPSO算法在经典函数测试、模型参数优选以及复杂整数规划模型求解中表现的分析,说明了算法的有效性。2)针对Guass-vSVR模型参数组合选取困难,用CCPSO算法对Gauss-vSVR模型参数组合进行优选,得到了Guass-vSVR-CCPSO模型。针对港口吞吐量时间序列及其影响指标的历史数据中的跳跃数据,将能处理跳跃数据的Guass-vSVR-CCPSO模型用于港口吞吐量的预测。预测过程中,用主成分分析法和相关性分析法确定预测模型的输入向量,设计算例对模型的可行性和有效性进行了验证。3)为使船舶在靠泊时尽可能靠近偏好泊位,缩短集卡运距,减少船舶在港时间,以船舶未按偏好泊位靠泊而产生的平均集卡运距和船舶平均在港时间最小为优化目标,建立了多目标离散泊位-岸桥分配模型。4)利用CCPSO算法求解建立的离散泊位-岸桥分配模型,开发了粒子可行-整数化处理模块,制定了粒子编码规则,确定了基于多目标函数的粒子历史极值和全局极值的计算方法,设计了用于泊位-岸桥分配模型求解的Cat映射全局扰动和云模型局部搜索策略,获得了基于CCPSO算法求解的多目标离散泊位-岸桥分配的新方法。根据集装箱码头船舶到达统计规律和码头装卸设备的技术参数设计实验算例,验证建立模型和求解算法的可行性和有效性。更多还原

【Abstract】 The port throughput forecasting is the fundamental of the port architecture optimization and infrastructure construction, it plays a important role in port planning-layout, the scale of investment in infrastructure, the development strategy and the transportation system. Berth and quay-crane are as scarce resources of the coastline, whether it can be scientific and reasonable allocation, has important practical significance in improving the efficiency of port production and the service level. In the process of the port throughput prediction and the berth and quay-crane allocation, the determination of model parameters and the solving of the berth and quay-crane allocation such as NP-hard problem have direct effect on the feasibility and effectiveness of optimization model. The intelligent algorithms provide an effective way to solve the above problems, but any kind of intelligent optimization algorithm is not perfect, it has some defects by its own structure. In order to better solve the optimization problem in port planning and operation management, a new hybrid optimization algorithm is proposed, based on the organic bond of the PSO algorithm, the Cat map and the Cloud model, the new hybrid optimization algorithm is applied in the port management, then the application of the new hybrid optimization algorithm in port throughput prediction and berth-quay-crane allocation of the port management is explored and discussed, the major work of this paper is as follow:1) The Cat map is introduced into the hybrid optimization algorithm, used for chaotic disturbance of poor individual in particle swarm, due to its better chaos characteristics. Considering the disadvantages of the poor diversity and the tendency to get trapped into local extremum and slow convergence in the late evolution stage of PSO algorithm, as well as the advantages of the ergodicity of Cat map and the randomness and stable tendency of Cloud model, this paper introduces mixed control parameter mix_gen and population distribution coefficient pop_distr to mix PSO algorithm, Cat map and Cloud model and proposes the CCPSO (Chaos Cloud Particle Swarm Optimization) algorithm, in order to take advantages of the three kinds of algorithm and improve the optimal performance. Classic test functions are selected to analysis the effect on the values of mixed control parameter mix_gen and population distribution coefficient popdistr, and the recommended mix_gen and pop_distr values are given for application to different optimizations. The performance of CCPSO in function optimization, model parameter optimization and integer programming model solving proves the effectiveness of the algorithm.2) Considering the difficulty for selection of the parameter combination, the CCPSO algorithm is used to optimize the parameter combinations of Guass-vSVR model, the Guass-vSVR-CCPSO model is proposed. The Guass-vSVR-CCPSO Model is applied to the prediction of port throughput, according to the structure of the port throughput sequence and the data of its influence factors, so as to deal with the jumping data in the sequence. The input vector of the Gauss-vSVR Model is selected by Principal Component Analysis (PCA) and correlation analysis. Then, the example analysis was made to assess the feasibility and effectiveness of the prediction model proposed in this paper.3) Considering the fact that the ships shall be close to their preferred berths when berthing can reduce the transportation distance of container truck and stay time of ships in terminal, a new berth and quay-crane allocation mode is established for diminishing the additional trucking distance incurred by failure of berthing at the preferred berths and stay time of ships in terminal.4) The CCPSO algorithm is used to solve the berth and quay-crane allocation mode, the feasible-integer processing module for particles is developed, the encoding rules of particles are established, the algorithm for the calculation method of the historical and the global extremum of particles is determined, the strategy of the Cat map global chaos disturbance and Cloud model local search for solution of berth and quay-crane allocation mode is designed, the algorithm based on the CCPSO algorithm for solving the allocation model is implemented. According to the statistical law of ship arrival at the container wharf and the technical parameters of handling equipment in the wharf, the numerical example is designed to verify the feasibility and effectiveness of proposed model and algorithm.更多还原