自动电压控制中不良数据辨识的研究

自动电压控制（AVC）系统由于缺少对发电厂遥测量数据的真实性进行有效和准确地辨识，容易引起装置误动。支持向量机（SVM）是数据挖掘中一种具有优良模式识别性能的新方法。本文提出了利用SVM建立辨识发电厂遥测量中不良数据的模型：首先应用SVM非线性回归对各种运行情况下发电厂正常的遥测数据进行曲线拟合（即训练）；然后运用训练好的模型对历史遥测数据(包括正常遥测数据和不良遥测数据)进行预测，取得预测值与实际遥测数据的误差百分比；再结合历史遥测数据正确与否，应用SVM分类功能训练分类网络；最后将实时遥测数据输入到训练好的曲线拟合网络和分类网络中，就能够迅速判断该遥测数据是否为不良数据。仿真算例验证了SVM模型的有效性和准确性。

The Research of Bad Measurements Detection and Identification in Automatic Voltage Controll

Automatic Voltage Controll (AVC)system may cause misoperation due to bad measurements without being identified effectively and accurately.Support vector machine(SVM) is a novel data mining method with excellent pattern recognition. This paper proposes a new bad measurements detection and indentification method- SVM : Firstly, apply SVM’s nonlinear regression for curve fitting, i.e., training. Secondly, use trained network to predict the history data (including correct data and bad data).Thirdly, according to the percentage error between the prediction value and the history data correct or not, apply SVM’s classification theory to train a classification SVM network. At last, put the real-time data into the curve fitting network and classification network to judge the data correct or not. The results of the simulation show effectiveness and accurateness of the SVM.