摘要
针对当前基于S变换的电能质量方法计算开销大、不能实时识别电能质量扰动的问题,提出利用快速S变换与最小二乘支持向量机相结合的识别电能质量干扰新方法。该方法从快速S变换得到的一维向量中提取各频率段模系数的标准偏差、最大模系数及额定频率对应的模系数作为特征向量,利用最小二乘支持向量机对电压骤升、电压骤降、电压中断、暂态脉冲、暂态振荡、谐波等几种电能质量干扰进行分类和识别。研究结果表明:与传统的基于S变换的电能质量方法相比,该方法在2个方面节省了时间,一是减少了提取特征量所用的时间,二是由于特征向量数据较少,采用支持向量机样本训练时间减少;特别是当电压扰动信号持续时间越长时节省效率越高,在同样准确性下,对于长度为1 024点的扰动信号,节省了约99%的时间;除此之外,该方法对信号分类的正确率可达98%,同时还具有较高的抗干扰能力。
Focusing on higher computation cost and lack of realtime detection for all techniques based on traditional Stransform to identify power quality disturbances, a realtime approach combining fast Stransform with least squares support vector machine is proposed. The standard deviation of module coefficients, maximum module coefficient of each frequency band, and module coefficient corresponding to the rated frequency are extracted from the onedimensional vector of the fast Stransform of the original power quality signals as features, and the least squares support vector machine based on optimized parameters and the minimum output coding is used to classify and identify the voltage swell,voltage sag, voltage interruption,spike,transient oscillation and harmonic waves. Compared with the traditional approach based on Stransform, the proposed approach reduces the tasks in both extracting features and training of the support vector machine classifier due to fewer training samples. The longer the duration of the voltage disturbance signal, the higher the saving efficiency. To the same accuracy, for the disturbance signal with a length of 1 024 points, processing time can be saved by 99%. The classification accuracy of this approach gets up to 98% with higher antiinterference ability.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2013年第8期133-140,共8页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(61271091)
陕西省教育厅科学研究计划资助项目(2013JK09937)
关键词
电能质量
快速S变换
支持向量机
实时性
power quality
fast S-transform
support vector machine
real-time
