用于局部放电图象识别的统计特征研究

结合图象识别技术，提出一种采用局部放电灰度图象的统计特征区分局部放电类型的方法。局部放电灰度图象统计特征由图象的矩特征和相关统计特征构成；其中矩特征描述局部放电灰度图象基本灰度分布状态，相关统计特征描述局部放电正、负工频半波图象之间的相关程度。设计出模拟变压器内部放电与外部放电的五种放电模型，并通过试验获得大量放电样本数据，采用局部放电灰度图象统计特征和人工神经网络分类器，对于五种放电样本获得了较高的识别率，表明该方法具有良好的应用效果。

USING STATISTICAL FEATURES FOR PARTIAL DISCHARGE IMAGE RECOGNITION

This paper brings forward a new method to use statistical features of partial discharge (PD) gray intensity images for PD pattern recognition. The statistical features are consisted of moments and correlation statistical parameters, of which the former describe the basic geometry characteristics and the latter characterize the correlativity between images of positive and negative half cycles. Five different discharge models are designed to simulate discharges occurring inside and outside transformer and large quantities discharge samples are acquired by the model tests. By means of the statistical features and artificial neural network, high recognition probability is achieved, as demonstrates the practicability of this method.