机组优化组合的人工神经网络拉格朗日混合方法  

ARTIFICIAL NEURAL-NET APPLIED TO UNIT COMMITMENT SCHEDULING

作　　者：张潮海[1] 周其节[1] 

机构地区：[1]华南理工大学自动化系

出　　处：《电力系统及其自动化学报》1995年第2期52-58,共7页Proceedings of the CSU-EPSA

摘　　要：本文提出了一种求解电力系统组合优化问题的混合神经网络─拉格朗日方法．至今，拉格朗日松弛法一直被记为是机组优化组合近似解的实用方法．这样．基于神经网络的监督学习和自适应识别概念，我们用神经网络来推测负荷需求与拉格朗日乘子的非线性关系，并且采用了优化的学习速率和势态项来加速网络的收敛．数值计算的结果表明本文的方法是可行的．This paper presents a hybrid artificial neural network (Arm) Lagrangian relaxation approach to combinatorial optAnisation problems in power systems, in pafticular to unit commitment. Until now, the lagrangian relaxation method has been studied as it appeared to be the most practical method for obtaining an approximate solution to unit commitment. Based on thd use of supervised learning neural-net technology and the adaphve pattern recognition concept, which presume the relahonship between power demand pattern and Lagrange multipliers (LMPs). The numerical results obtained are very encouraging.

关 键 词：机组优化组合 人工神经网络 拉格朗日松驰法 

分 类 号：TM310.13[电气工程—电机]