【作者】 侯云鹤；

【导师】 程时杰； 熊信银；

【作者基本信息】 华中科技大学 ， 电力系统及其自动化， 2005， 博士

【摘要】 电力系统规划是电力系统研究和工程实践的重要课题。随着电力工业的发展,电力系统更加复杂,在很多情况下,电力系统规划面临非凸、非线性以及非连续的优化问题,这对电力系统规划提出了新的挑战。传统的数学优化算法在处理这些问题时,存在困难。研究新的优化算法成为工程实践和理论研究中必需解决的课题。人工智能方法逐渐成为求解电力系统规划问题的一种可供选择的算法。在人工智能算法中,有很大一类模拟群体行为的方法,这类算法被称为群体智能算法。虽然不同的算法有不同的背景,但是就抽象层面看,都是由一群仅具有简单功能的自治个体,通过相互之间的通讯和与环境的互动,从而实现复杂的功能。由于这些算法的通用性和有效性,它们已成为电力系统规划领域研究的热点之一。但是群体智能算法的研究成果相对分散,缺乏系统化的研究。为了提高其应用的效率,需要从理论上对其进行分析。电力市场稳定是电力市场机制设计和监管领域的重要课题之一。传统的市场稳定性理论是建立在普通商品市场的基础之上的,但是电力系统有其特殊性,如电力供需同时进行,电力需求在一定程度上具有周期性的特性,以及电力报价交易分时开展等。这些因素是关系电力市场稳定的重要因素。本文旨在研究群体智能算法及其在电力系统规划中的应用以及电力市场的稳定性,主要包括以下几个方面: 首先,建立群体智能算法的统一框架,对各个算子进行定义,并探讨了为使算法收敛,框架中不同搜索策略需要满足的充分条件。第二,在群体智能算法框架下,基于求解组合优化的蚁群算法,提出了一种用于求解一般形式的非凸、非线性约束优化问题的通用算法—广义蚁群算法。基于不动点理论,给出了该算法收敛的充分条件。该方法被用于电力系统经济负荷分配和电力系统无功优化中,取得了比传统方法好的优化结果。第三,提出了一种全局收敛的粒子群算法,基于群体智能算法统一框架,证明了算法的全局收敛性。为了提高算法的效率,在满足算法全局收敛充分条件的基础上,提出了广义蚁群算法和改进粒子群算法的结合算法。多个电力系统经济负荷分配的算更多还原

【Abstract】 Power system planning is one of the most important issues in the area of power system research and operation. With the development of power systems, the mordern power systems have become more and more complex. With the nature of nonlinear, nonconvex and discontinuous, it is already very difficult to deal with the power system planning for the conventional mathematical optimization algorithms. It is, therefore, necessary to develop algorithms to solve these problems. In the recent years, the algorithms based on the artificial intelligence (AI) have been introduced to the power system planning as alternative methods such as the swarm intelligence algorithms inspired from the swarm behaviors of the nature. Although different swarm intelligence algorithms are developed from different backgrounds, some things are common to all these algorithms. The purposeful behaviors of these algorithms are achieved with the cooperation of a number of individual agents, which follow the simple local rules and interact with the environment. In such a way some complicated goal can be reached. Because these algorithms show very strong robustness and effectiveness, they have attracted great deal of attention. However, the dispersive research result currently presented in the literature show that it is time to carry out theoretical research on these algorithms. When design a policy for electric power market, it is necessary to consider whether the market under this policy is stable. Theoretically, the foundation of electricity markets is based on the classical economic theory of competitive markets and their benefits. However, it is widely recognized that electricity market differs from any other commodity market in the following aspects. Firstly, electricity as one kind of the special commodity can not be stored, i.e., the market has to be cleared instantaneously. Secondly, the load demand in electricity market displays a cyclic pattern. Finally, the discrete biding strategy of the power market should be considered when investigating the market’s stability. The object of this paper is to investigate the power system planning by swarm intelligence algorithms and the stability of the power market. Following results are obtained. Firstly, a unified frame for swarm intelligence algorithms is proposed. Main operators in the frame are defined. Several sufficient conditions for the convergence of the algorithms are also given. Secondly, according to the frame proposed, a new versatile optimization algorithm called generalized ant colony optimization is developed to solve the discontinuous, nonconvex, nonlinear constrained optimization problems. Based on the fixed-point theorem, the sufficient conditions for the convergence of the algorithm are deduced. The effectiveness of the algorithm is verified by the power system economic dispatching and reactive power planning. Thirdly, a global convergence algorithm called enhanced particle swarm optimization algorithm is developed. Based on the stochastic analysis theory, a sufficient condition for the convergence of the algorithm is given. To increase the speed of convergence, the general ant colony optimization is integrated with the particle swarm optimization algorithm. The effectiveness of the algorithms developed above is tested by the power system economic dispatching. Encouraging results are obtained. Fourthly, a new swarm intelligence algorithm called quantum-inspired evolutionary algorithm is proposed to solve the optimal transmission network expansion planning. Some new operators are defined and used for the global searching. The performances of these operators are investigated by comparison. In order to increase the convergence, measures are proposed to optimize the parameters in the algorithm. Test results are used to show the accuracy and the efficiency of the algorithm. Fifthly, a new optimization algorithm called estimation of distribution algorithm is introduced to solve the transmission network expansion planning. Two kinds of EDA: population-based incremental learning (PBIL) and factorized distribution algorithm (FDA) are used. To increase the convergence, several enhanced operators are involved in the algorithm. Numerical results show that the proposed method is feasible and effective. Sixthly, the generation expansion planning of power system is solved by modified genetic algorithm and partheno-genetic algorithm. Several results of actual system validate the proposed algorithms.Seventhly, the stability of power market is investigated. Several realistic models with periodic factors of the electricity market that is characterized by dynamic system models are obtained. The problem of electricity market stability is transformed to the problem of the stability of a periodic solution of the dynamic system. To study the stability conditions, several lemmas are proven. Conditions for the existence of a unique stable periodic solution are obtained and the ultimate convergence boundary is estimated. Simulation results show the effectiveness of the models and the stability conditions obtained. Finally, conclusions about this dissertation are summarized and the further work is pointed out.更多还原