提升型贝叶斯分类器在电力系统暂态稳定评估中的应用  

Power System Transient Stability Assessment Based on Boosting Bayesian Classifier

作　　者：卢锦玲[1] 朱永利[1] 赵洪山[1] 刘艳[1] 

机构地区：[1]华北电力大学电气与电子工程学院,保定071003

出　　处：《电工技术学报》2009年第5期177-182,共6页Transactions of China Electrotechnical Society

摘　　要：应用人工智能技术可实现电力系统暂态稳定的快速评估,朴素贝叶斯分类器作为人工智能方法的一种,其训练计算复杂度是线性的,是解决分类问题最实用、有效的方法之一,但由于它是建立在属性变量相对类变量独立的假设前提下,故存在一定的误分类率。本文采用Adaptive Boost(AdaBoost)算法对朴素贝叶斯分类器进行提升,有效地降低了误分类率,并将提升型贝叶斯分类器用于电力系统暂态稳定评估。选取能迅速反映电力系统暂态过程的特征量,作为贝叶斯分类器的属性变量,将系统稳定或不稳定作为类变量,采用数值仿真算法产生大量样本,并对属性的连续数据进行离散化处理,构造了用于暂态稳定评估的提升型贝叶斯分类器。对新英格兰10机39节点系统进行仿真,结果表明:提升型贝叶斯分类器用于电力系统暂态稳定评估可有效降低机器学习的复杂度和提高暂态稳定的分类精度。Transient stability can be rapidly assessed using the artificial intelligence technology. The Bayesian classifier is one of the methods of artificial intelligence. Due to its linear training computational complexity, it is also one of the most practical and effective way for classification. Naive Bayesian classifier-bound as a Bayesian network is built on the characteristics of the property relative to the type of condition attributes of an independent assumption. Therefore, there is the possibility of misuse of classified. In this paper Adaptive Boost algorithm is used to boost the Naive Bayesian classifier which effectively reduces the rate misuse of classified. And the boosting Bayesian classifier is first put forward to be used for power system transient stability assessment. In this paper, the characteristics which can rapid reflect power system transient are selected. They are looked as the attribute variables of Bayesian classifier and stability or instability of the system are looked as class variables. The numerical simulation algorithm is used to produce a large amount of samples. And the attributes variables are processed to discrete data, the boosting Bayesian classifier is construct for transient stability assessment. At last, simulation results of the New England 10 machine 39-bus system verify that the power system transient stability assessment based on boosting Bayesian classifier can effectively reduce machine learning complexity and improve classifier precision.

关 键 词：贝叶斯分类器 暂态稳定评估 数据离散化 ADAPTIVE Boost算法 

分 类 号：TM732[电气工程—电力系统及自动化]