【作者】 李可；

【导师】 马孝义；

【作者基本信息】 西北农林科技大学 ， 农业电气化与自动化， 2010， 硕士

【摘要】 配电网络规划是一个具有不确定性、非线性、多目标性和动态性等特点的复杂系统优化问题。本文运用改进的遗传算法、可靠性价值理论和多目标规划理论进行了研究,其主要研究内容如下:首先探讨了过去研究中对配电网络中网架结构和馈线截面的分步优化,难以获得全局最优解的问题,重点研究了以用电节点为基础的配电网络布置与馈线截面同步优化整数遗传算法编码方法,基于供用电节点矩阵转换的线路路径自动识别方法和配电网络中功率、电压等约束的自动计算方法,并引入退火罚因子设计动态罚函数处理约束条件,由此建立了基于整数遗传算法的配电网络规划同步优化的数学模型,实现了配电网络规划中网架结构和馈线截面的同步优化。算例研究表明,论文方法克服了分步式优化方法不易得到全局最优解的缺陷,可以快速有效地得到配电网络的网架结构和馈线截面规划全局优化方案。其次,针对配电系统可靠性评估模型中没有考虑影响设备实时故障变化的影响问题,探讨电力设备实时故障率的变化规律,研究建立了考虑设备实时故障变化的复杂配电网络可靠性评估网络等值模型及其算法;探讨不同天气条件下对可靠性指标的变化规律,研究建立考虑多态天气因素的复杂配电网络的可靠性评估分块算法模型,使得修正后模型较完备地考虑了不同天气因素的对配电系统可靠性评估的影响;为克服以往缺电损失计算方法中存在的缺陷,试图尝试把间接缺电成本考虑进缺电成本计算中,研究改进了缺电成本的配电网络缺电损失评价的模型。算例验证表明,该估算模型能为分析电力系统故障和负荷点缺电引起的配电网络缺电损失评估提供方法。最后,针对配电网络多目标规划模型中难以全面考虑经济性优化子目标和可靠性优化子目标之间的相互作用、相互影响和整体协调性的问题,论文先把全局搜索性能良好、鲁棒性好的遗传算法对各子目标分开优化,然后对子目标间引入协同进化算法根据各约束条件进行协调和权衡,在此基础上建立了考虑缺电成本配电网络多目标规划数学模型;同时针对协同进化算法应用于大规模多目标问题的求解时,存在着难以收敛到最优pareto解,收敛速度慢和对初始解选取敏感等缺陷,通过引入退火动态罚函数保证算法收敛到全局最优解,精英保留措施的加入降低了算法对初始解的敏感性,促使算法朝着最优最优pareto解的方向快速收敛,并通过实例对其可行性和实用性进行了验证,本文所研究的多目标模型配电网络能够使配电网络经济性指标和可靠性指标相互协调和平衡,最终得到了最佳规划方案,可以使规划结果更具实际意义。更多还原

【Abstract】 Distribution network planning is a complex NP—hard optimal problem with characteristics such as uncertain, non-linear, multi-objects and multi-stages. Intelligent optimization algorithm, reliability worth theory and multi—objective were adopted in this paper to do the research as follows:Firstly, In the traditional optimization design of electricity distribution network , network structure and conductor cross sections were optimized step by step,which firstly optimize network structure of electricity distribution network, and then optimize cross section of electricity conductors on basis of the optimized network structure. it is can not get the global optimization scheme of network structure and conductor cross sections. In order to deal with this problem, a integer coding method based on a electricity node for network structure and conductor sections simultaneous optimization design was put forward; An automatical method for identifying electricity flow route based on electricity node matrix transformation was deduced, an automatically calculating program of electricity power and voltage of different nodes was developed, an dynamic penalty function was used to deal with the constrain conditions. Hence, an electricity distribution network structure and conductors cross sections simultaneous optimization model and its solution method based on integer code genetic algorithm were obtained. Two different electricity network examples were used to test this method, and its results were compared with optimized schemes based on traditional optimization method, which show that the simultaneous optimization model and algorithm can reduce the electricity conductors cost of distribution network, and it can effectively get global optimization scheme of electricity distribution network.Secondly, This paper was researched and established equipment real-time fault complexity distribution network reliability assessment of model and the algorithm, for which distribution system reliability evaluation model did not consider the impact of equipment and the rule of real-time fault change impact; Section algorithm and model based different weather conditions on complexity distribution network reliability assessment was established which fully consider the rule of different weather conditions for reliability assessment, the model which makes more completely consider reliability assessment; This paper tried to take into account the indirect costs of lack of electricity power shortage cost to overcome the shortcoming of electricity loss calculation method previously, and improved the model of costs of lack of electricity power. Example shows that the model could provide assessment methods for analyzing power system fault and the load point of power shortage.Finally, the distribution network for multi-objective programming model was difficult to fully consider economy optimization and reliability of optimization of the interaction and overall coordination. This paper firstly used genetic algorithms which has global search performance and robust separately searching performance, and then coordinated between various constraints by the introduction of co-evolutionary algorithm, and finally the multi-objective mathematical model for distribution network planning was built. In order to overcome defects, such as being difficult to getting the pareto, slowly speed and sensitive to initial solution for large-scale multi-objective problem, the multi-objective mathematical model were obtained through the introduction of annealing algorithm for dynamic penalty function and the elite retention measures. Electricity network examples were used to test this method, and its results were compared with optimized schemes based on traditional optimization method, which show that the multi-objective model and algorithm could get global optimization scheme of distribution network and make the planning results of more practical significance.更多还原