改进人工蜂群算法在梯级水库群优化调度中的应用

发布时间：2018-05-18 05:37

本文选题：人工蜂群算法 + 优化调度　；参考：《南昌工程学院》2017年硕士论文

【摘要】：随着我国水电事业日益发展,越来越多的水电站不断被开发利用,梯级水库群被广泛应用于各级水利枢纽系统。如何对梯级水库群进行合理调度,提高整体发电量成为水力资源管理利用的核心内容之一。因此,研究梯级水库群优化调度,制定调度规则,具有十分重要的学术意义和应用价值。水库调度是高维、多时段的非线性优化问题。传统算法通过建立精确模型的方式能够解决单一水库的调度问题,但随着水库数目的增多,优化问题的计算量显著增大,造成“维度灾”,难以符合实际应用。随着现代人工智能技术的发展,大量智能算法被应用于解决复杂的优化问题,这为解决梯级水库群调度问题提供了新的途径。人工蜂群算法具有结构简单、鲁棒性强等优点,被广泛应用于众多工程领域。但是,该算法本身仍存在许多不足。本文以标准人工蜂群算法为研究对象,并对其进行改进,取得主要成果如下:(1)针对标准人工蜂群算法收敛速度慢的缺点,引进改进粒子群算法中狭义中心的概念,并对其进行改进。通过比较适应度,选取优秀的蜜源构成改进的狭义中心,使狭义中心具有更好的性质;其次,修改标准蜂群的更新策略,利用全局最优解引导,使雇佣蜂始终围绕当前全局最优点搜索,强化蜂群在最优点附近开发隐藏解的能力,提高算法的求解精度。由此提出一种改进狭义中心的人工蜂群算法。(2)在收敛速度提升的同时,算法极易陷入局部最优,因此引入虚拟蜜源思想。在初始化时将整个种群随机划分为两个子群,并采取不同的方法建立虚拟蜜源以代替原蜜源进化。由于虚拟蜜源拥有多个个体的信息,在蜜源进化的同时加强不同子群间的信息交流,达到综合学习的目的,构造了综合学习的人工蜂群算法。(3)为了改变单一进化模式导致算法搜索能力失衡的问题,采用多群策略对算法进行优化。首先,将雇佣蜂随机分为三个子群,分别对应三种进化策略。由于三种策略具有不同的特征,能够平衡算法的全局搜索与局部开发能力。其次,通过模仿粒子群算法,充分利用当前全局最优蜜源和随机邻域蜜源包含的丰富信息,优化了跟随峰的搜索策略。构建了改进的多策略人工蜂群算法。论文提出了三种改进算法。通过12个经典基准函数和28个CEC2013函数测试结果表明,三种算法具有较好的搜索效率和寻优精度。最后,论文以清江流域的梯级水库群(水布娅—隔河岩—高坝洲)为研究背景,以梯级水电站总发电量最大为目标函数,建立梯级水库群联合调度模型,将三种算法应用于梯级水库发电调度中,取得了良好的结果。
[Abstract]:With the development of hydropower industry in China, more and more hydropower stations are being developed and used. How to carry on the reasonable operation to the cascade reservoir group and how to improve the whole generating quantity become one of the core contents of the management and utilization of the hydraulic resources. Therefore, it is of great academic significance and application value to study the optimal operation of cascade reservoir groups and to formulate dispatching rules. Reservoir operation is a high-dimensional, multi-time nonlinear optimization problem. The traditional algorithm can solve the operation problem of a single reservoir by establishing an accurate model. However, with the increase of the number of reservoirs, the calculation of the optimization problem increases significantly, resulting in a "dimensional disaster", which is difficult to be applied in practice. With the development of modern artificial intelligence technology, a large number of intelligent algorithms are applied to solve complex optimization problems, which provides a new way to solve the cascade reservoir group scheduling problem. Artificial bee colony algorithm is widely used in many engineering fields because of its simple structure and strong robustness. However, the algorithm itself still has many shortcomings. In this paper, we take the standard artificial bee colony algorithm as the research object and improve it. The main results are as follows: 1) aiming at the shortcoming of the standard artificial bee colony algorithm, we introduce the concept of narrow center in the improved particle swarm algorithm. And improve it. Through the comparison of fitness, select the excellent honey source to form the improved narrow center, make the narrow center have better properties. Secondly, modify the renewal strategy of the standard bee colony, use the global optimal solution to guide, The employment bee is always focused on the current global optimal search to enhance the ability of the colony to develop hidden solutions near the best and to improve the accuracy of the algorithm. Therefore, an improved artificial honeybee colony algorithm .Y2 (narrow center) is proposed. It is easy to fall into local optimum when the convergence speed is improved, so the virtual honeycomb is introduced. The whole population is randomly divided into two subgroups in initialization, and different methods are adopted to establish virtual honey source instead of original honey source evolution. Since virtual honey source has more than one individual's information, it strengthens the exchange of information among different subgroups as well as the evolution of honey source, so as to achieve the purpose of comprehensive learning. In order to change the unbalance of search ability caused by a single evolutionary model, a synthetic learning artificial bee colony algorithm is constructed. In order to optimize the algorithm, a multi-swarm strategy is used to optimize the algorithm. First, employment bees were randomly divided into three subgroups, corresponding to three evolutionary strategies. Because the three strategies have different characteristics, they can balance the ability of global search and local development of the algorithm. Secondly, by imitating the particle swarm optimization (PSO) algorithm, the search strategy of the following peak is optimized by making full use of the abundant information contained in the global optimal honey source and the random neighbor honey source. An improved multi-strategy artificial bee colony algorithm is constructed. Three improved algorithms are proposed in this paper. The test results of 12 classical datum functions and 28 CEC2013 functions show that the three algorithms have better search efficiency and optimization accuracy. Finally, taking the cascade reservoir group (Shuibuya, Geheyan and Gaobazhou) in the Qingjiang River Basin as the research background, taking the maximum total generating capacity of the cascade hydropower station as the objective function, the combined operation model of the cascade reservoir group is established. Three algorithms are applied to cascade reservoir power generation operation, and good results are obtained.
【学位授予单位】：南昌工程学院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TV697.12;TP18

【参考文献】