机构地区: 湖南大学电气与信息工程学院
出 处: 《湖南大学学报(自然科学版)》 2009年第4期38-42,共5页
摘 要: 提出了一种基于S变换、加窗插值快速傅里叶变换(FFT)和概率神经网络(PNN)的电能质量扰动检测和分类方法.应用S变换和加窗插值FFT对电能质量多扰动信号进行时频分析,获取信号的特征量.通过训练信号集上获得的特征量,训练了一个概率神经网络用于扰动分类.训练好的网络在测试信号集上的测试结果表明,对正常电压和常见的电能质量扰动,该方法具有较高的分类准确率,在训练样本数较少、噪声影响大和多扰动信号并存时仍能取得较好的分类效果. A new detection and classification method of power quality disturbances based on S transform, interpolating windowed fast Fourier transform (FFT) and probabilistic neural network (PNN) was proposed. S transform and interpolating windowed FFT was first applied to perform time - frequency analysis on power quality disturbance samples, and the features can then be extracted from the results. These features are then used to train a PNN for disturbance classification. Results of applying the trained PNN on a test set with common power quality disturbances show that the method has relatively high classification accuracy. In the presence of smaller training set, higher noise level, and multiple types of disturbances, the proposed method can still achieve good classification.
领 域: [电气工程]