【作者】 姚静；

【导师】 方彦军；

【作者基本信息】 武汉大学 ， 机械设计及理论， 2013， 博士

【摘要】 负荷优化分配是电厂经济、安全运行的一项很重要的工作,其目的是在给定某运行时段的开停机计划后,在组合机组满足电力系统运行约束条件的基础上,合理分配各机组负荷使得发电成本、污染排放等目标尽可能小。本文针对这一优化问题对其数学模型和优化算法进行了深入的研究。首先对电厂普遍采用的经济性指标进行分析与讨论,确定标准供电煤耗率为其优化目标,研究了基本遗传算法求解负荷优化分配问题的缺陷,并提出了相应的改进方案,在基本遗传算法的基础上提出了边界约束的初始化方案,排序选择的选择算子、自适应交叉算子以及最优值保存策略,最终收敛速度得到大幅提高。提出了基于遗传禁忌混合算法的负荷优化分配方法,遗传算法具有较强的全局搜索能力,禁忌搜索算法具有较强的“爬山能力”,两种算法优势互补,结合后可以避免出现早熟现象。为了充分利用禁忌算法的局部搜索能力,而又防止太过频繁的调用禁忌算法造成时间复杂度大幅度提高的问题,本文提出2种算法结合的关键是在遗传算法趋向于早熟现象时通过禁忌算法跳出这种局部最优状态,利用适应度的样本方差来比较遗传算法种群的变化情况,并提出了简单可靠的早熟判断方法,在禁忌算法的设计上提出了邻域解产生的新方法。实例分析证明该算法求解大规模考虑阀点效应的负荷优化问题性能更优。将负荷优化分配这一带约束的单目标优化问题转化为多目标优化问题来处理,建立了双目标负荷优化分配数学模型,一个目标函数为：总煤耗函数,另一个目标函数为：违反约束条件的程度函数。在评价策略、遗传算子等方面对常规的多目标遗传算法进行了改进,利用Pareto强度值作为个体的评价指标,利用遗传算法实现种群的进化,最终找到最优解,为机组负荷分配的求解提供了新的有效算法。最后通过仿真对简单遗传算法、改进遗传算法、遗传禁忌混合算法以及多目标遗传算法的性能进行了分析比较。更多还原

【Abstract】 Load optimal distribution is a hot research issue all along in power system. On the premise of giving the unit commitment and all constraint conditions, load optimal distribution studies how to dispatch load to operating sets to makes the total operating fees lowest in power station. This paper has deeply studied optimal modes and optimal algorithms on load optimal distribution.At first, the paper analyzes the economic indexes which are generally adopted in power station, and defines standard coal consumption as the optimal objective. The shortage of simple genetic algorithm solving economic load dispatch is researched. On the basis of the simple genetic algorithm, some improved solutions are proposed, which include increasing boundary constraint to initial population, ranking selection operator, adaptive crossover operator and optimization preservation strategy. The improved genetic algorithm is applied to three generating units, and the results shown that the improved genetic algorithm has better optimization effect.Next, a real code genetic-tabu search hybrid algorithm is presented. Genetic algorithm is characterized by the capability of global searching, and tabu search is characterized by the capability of mountain climbing, so the advantages of two algorithms complement each other and the hybrid algorithm can avoid pre-maturity after combination. In order to fully utilize tabu search’s local search ability, and also avoid using tabu search too much bringing about time complexity increasing, this paper proposes the key to combination of the two algorithms is breaking local optimum by tabu search when the genetic algorithm tends to prematurity. A simple and reliable method is advanced to estimate. prematurity, which compare the population change by sample variance. A new method is put forward to produce the neighborhood solution of tabu search. The effect of optimization is compared by case analysis, and the results demonstrate the effectiveness and viability of the algorithm.Then, load optimal dispatch is a single objective optimization question with constraints. The question is turned into a multi-objective optimization question in this paper. One objective is the total coal consumption function, and the other is the constraint violation degree function, so that the two-objective mathematic mode is built. Some improved measures are proposed in evaluation strategy, genetic operators, etc. The evaluation function is the individual pareto strength, and the population evolution depends on genetic algorithm. This algorithm provides a new and effective method to load economic dispatch. Finally through the simulation the performance of simple genetic algorithm, improved genetic algorithm, hybrid genetic and tabu search algorithm are compared and analyzed.更多还原