基于模式搜索算法的电力系统机组组合问题  

A General Pattern Search Algorithm for Electric Power System Unit Commitment Problems

作　　者：黎静华[1] 韦化[1] 

机构地区：[1]广西大学电气工程学院,南宁530004

出　　处：《电工技术学报》2009年第6期121-128,共8页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(50867001);广西大学科研基金(DD020015)资助项目

摘　　要：机组组合问题是一个复杂的大型混合整数非线性规划问题,目前尚未找到理想的解决方法。本文提出一种用于混合变量的模式搜索方法(GPSMV)解决机组组合问题。该方法能求解非凸、非线性、不可微甚至不连续的规划问题,理论上能保证全局收敛于稳定的解,其仅需要求目标函数及由约束条件构成的障碍函数值而不需要对目标函数和约束条件求导。在优化过程中离散变量作为独立的变量,而不是把原问题分成离散和连续两层来处理,适合于求解离散和连续的混合变量问题。最后对10～100机,26机24时段等六个系统进行仿真试验验证该算法。Unit commitment belongs to mixed variables programming which is difficult to find the optimal solution in mathematics. This paper presents a general pattern search algorithm with mixed variable （GPSMV） to solve unit commitment problems at the first time. The proposed algorithm can solve problems for which the objective function is nonlinear, nonconvex, nondifferentiable, stochastic, or even discontinuous. GPSMV guarantees global convergence and it only needs values of objective function and barrier functions consisted by constraint condition while discards the information of their derivative. During the process of optimization, the discrete variables can be deal directly while not dividing original problem to be discrete and continuous part. Simulations are executed on six systems of 10-100 units and 26 units in 24 time intervals, and the results verify the effectiveness of the proposed algorithm.

关 键 词：机组组合 过滤算法 模式搜索算法 离散变量 混合变量 

分 类 号：TM732[电气工程—电力系统及自动化]