采用极限学习机改进遗传算法的分布式电源优化配置

发布时间：2018-06-12 18:53

本文选题：分布式电源 + 极限学习机　；参考：《长沙理工大学》2014年硕士论文

【摘要】：随着社会经济的发展,能源与环境问题的重要性日益凸显,而分布式发电由于其具有低碳环保,投资小,发电方式灵活等优点受到广泛认可和应用。但是大量风电、光伏发电等分布式电源不断接入电网对电网的安全可靠、经济运行带来了更多挑战。因此,如何合理高效地规划分布式电源接入系统就变的尤为重要。文章阐述了分布式发电的发展及研究现状,详细介绍了几种主要分布式电源的并网情况,并从网络损耗、电能质量、可靠性、潮流分布等方面对其并网后给电网造成的影响进行了详细分析。通过对现有关于分布式电源优化配置问题求解方法分析表明,传统算法普遍存在速度较慢、容易陷入局部最优等问题,因此提出采用基于极限学习机改进遗传算法来求解此问题；该算法利用了一种新型的单隐层前馈神经网络算法一极限学习机(Extreme Learning Machine, ELM),来对基本遗传算法进行改进；同时也利用其优良的非线性映射能力来模拟前后两代种群的进化过程,并与传统遗传算法相结合；通过合理的参数设定,进而达到提高算法的全局搜索能力与收敛速度目的。为了综合兼顾环境效益和经济效益,建立了以投资运行成本最小、网络损耗费用最小、环境效益最大的经济优化配置模型；针对传统分布式电源规划中所缺少的对候选安装节点选择问题,提出了采用计算各节点视在二次精确矩值大小,并进行排序,选择最优候选安装节点的方法。通过对某地实际35节点系统进行仿真,算例分析表明：文中所采用的极限学习机改进遗传算法在求解分布式电源优化配置问题时,计算精度、收敛速度和寻优能力均优于传统遗传算法,可以得到更合理可靠的配网优化配置方案；同时也验证了采用计算各节点视在二次精确矩值的方法可以极大的减少变量的维数,降低计算量,提高了算法效率。
[Abstract]:With the development of social economy, the importance of energy and environment is becoming more and more important. Distributed generation has been widely recognized and applied for its advantages of low carbon environmental protection, low investment, flexible power generation and so on. However, a large number of wind power, photovoltaic generation and other distributed sources connected to the grid are safe and reliable, and economic operation brings more challenges. Therefore, it is very important to plan distributed power access system reasonably and efficiently. This paper describes the development and research status of distributed power generation, and introduces in detail the grid-connected situation of several main distributed power sources, including network loss, power quality, reliability, etc. The influence of power flow distribution on power grid is analyzed in detail. Based on the analysis of the existing methods for solving the problem of optimal configuration of distributed power supply, it is shown that the traditional algorithms generally have some problems, such as slow speed, easy to fall into local optimum, etc. Therefore, an improved genetic algorithm based on extreme learning machine is proposed to solve this problem, which uses a new single hidden layer feedforward neural network algorithm, extreme Learning Machine (ELMU), to improve the basic genetic algorithm. At the same time, it also uses its excellent nonlinear mapping ability to simulate the evolution process of the two generations of population before and after, and combines with the traditional genetic algorithm. Through reasonable parameter setting, the global search ability and convergence speed of the algorithm can be improved. In order to give consideration to both environmental benefit and economic benefit, an economic optimal allocation model with minimum operating cost of investment, minimum cost of network loss and maximum environmental benefit is established. In order to solve the problem of selecting candidate installation nodes, which is missing in traditional distributed power generation planning, this paper proposes a method to select the optimal candidate installation nodes by calculating the second accurate moment value of each node and ranking them. Through the simulation of the actual 35 bus system in a certain place, the example shows that the improved genetic algorithm of the ultimate learning machine used in this paper is accurate in solving the problem of optimal configuration of distributed power supply. The convergence speed and optimization ability are superior to the traditional genetic algorithm, and a more reasonable and reliable optimal allocation scheme of distribution network can be obtained. At the same time, it is verified that the method of calculating the apparent accurate moment value of each node can greatly reduce the dimension of variables. The computational complexity is reduced and the efficiency of the algorithm is improved.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM73;TP18

【参考文献】