论文部分内容阅读
原油调度是炼化企业生产过程中一个重要环节.在石油炼制过程中,为节约成本,降低原油性质的波动,并保证常减压装置的平稳运行,通常选择多种原油混炼.原油混输调度就是根据生产计划的需求,在满足各种操作约束和保证常减压装置正常运行的前提下,通过调节各种原油的混合比例及油罐的切换次序等,尽可能的降低调度过程的费用.本文首先论述了长周期原油混输调度问题,其中包括原油从油轮卸载到码头原油罐、由码头原油罐通过管道向厂区原油罐输送原油,以及厂区原油罐向常减压装置提供进料等过程.提出了一种与数学规划方法完全不同的基于事件树的方法来解决长周期原油混输调度的建模和优化问题.避免了数学规划方法建模复杂、求解困难等缺点.在本文中,事件树是表示系统在整个调度周期内系统可能状态的图形,其中每一个节点是系统在某一时刻的状态,包含有油轮的工作状态(在海上等待,卸油)、原油罐的工作状态(加油,输油,闲置)、各原油罐所存储原油的类型和储量、输油罐的输油速率等.该方法采用自然语言建模,利用连续时间的表达方法,根据事件发生的时刻来划分时间段,提高了求解的精度.在事件树结构单元的求解上,利用了事件驱动规则,推动事件树的生长.即当有事件发生时,根据事件发生的时刻及相应的事件处理方法,计算出事件发生后系统所有可能的状态,即当前节点的子节点.同时为了降低系统的规模,引入了分解集成策略,把整个调度系统分解为码头调度子系统和厂区调度子系统,对两个子系统进行优化,最终将两部分的最优结果集成起来便可得到整个系统的最优解.在整个事件树的搜索方法上,采用了深度优先搜索算法和剪枝策略来提高算法的搜索效率.通过深度优先搜索方法可快速得到可行解,以得到的可行解中的最优的解作为依据,与后续的解进行比较,可以省略大量节点的计算,极大地提高算法的计算效率.最后通过4个工业规模的例子证明了本方法的有效性.在调度周期为4周,7艘油轮,6个码头原油罐,6个厂区原油罐,2套常减压装置,6种原油类型的情况下,CPU(AMD3000+,2.0GHz)的计算时间不超过70s.
Crude oil dispatching is an important step in the process of refining and petrochemical enterprises.In the process of petroleum refining, in order to save costs, reduce fluctuations in the nature of crude oil, and ensure the smooth operation of atmospheric and vacuum unit, usually choose a variety of crude oil mixing. Transport scheduling is based on the needs of production planning, to meet the various operational constraints and to ensure the normal operation of atmospheric and vacuum unit under the premise, by adjusting the mixing ratio of various crude oil and tank switching order, as much as possible to reduce the scheduling process Cost.This paper first discusses the long-period crude oil mixed transportation scheduling problems, including the unloading of crude oil from the tanker to the terminal crude oil tanker, crude oil tanker terminal through the pipeline to the factory crude oil tank crude oil, and plant crude oil tank to atmospheric and vacuum Device provides the process of feeding and so on.This paper proposes an event tree-based method which is totally different from the mathematical programming method to solve the modeling and optimization problem of long-period crude oil mixed transportation scheduling, avoids the modeling complexity and solving difficulty of the mathematical programming method Disadvantages. In this paper, the event tree is a graphic representation of the system’s possible states throughout the scheduling cycle, where each node is the system at some point Including the working status of the tankers (waiting at sea and unloading oil), the working status of the crude oil tank (refueling, transporting oil and idle), the types and reserves of the crude oil stored in the crude oil tanks, the losses of the oil tanks Oil rate, etc. The method uses natural language modeling, using continuous time expression method, according to the time of the event to divide the time period, to improve the accuracy of the solution.In the event tree structure unit solution, the use of event-driven rules, Promote the growth of the event tree, that is, when an event occurs, calculate all possible states of the system after the event, that is, the subnodes of the current node according to the moment when the event occurs and the corresponding event processing method, and in order to reduce the size of the system, The disassembly and integration strategy is introduced to decompose the whole dispatching system into terminal dispatching subsystem and factory dispatching subsystem. The two subsystems are optimized and finally the optimal result of the two parts can be integrated to obtain the optimal solution of the whole system. The search method of the entire event tree, using depth-first search algorithm and pruning strategy to improve the search efficiency of the algorithm through the depth-first search method can quickly get Based on the optimal solution of the feasible solutions obtained, compared with the subsequent solutions, a large number of nodes can be omitted and the computational efficiency of the algorithm can be greatly improved.Finally, The effectiveness of the method. In the case of a scheduling cycle of 4 weeks, 7 tankers, 6 wharf crude oil tanks, 6 crude oil tanks in the plant, 2 sets of atmospheric and vacuum units and 6 crude oil types, the CPU (AMD3000 +, 2.0 GHz) calculation time does not exceed 70s.