基于小波神经网络的大功率电器识别技术研究
|
摘要
针对公寓用电中的大功率电器识别问题,提出利用小波神经网络对大功率电器进行识别。由于采集到的电网电流信号是基波信号和谐波信号的混合,因此需要进行信号分离。基于Mallat快速算法进行小波变换提取其中的谐波电流信号;将总电流的平均功率增量和谐波电流的平均功率增量经过归一化处理后作为大功率电器识别的特征向量,利用得到的特征向量对融合型小波神经网络进行基于BP算法的网络训练;利用训练好的小波神经网络对未知的电网电流数据进行识别,实现大功率电器的在线识别和预警。对比仿真实验表明:利用小波神经网络对大功率电器识别比传统的BP神经网络有更高的准确率。
In order to identify the high-power electrical appliances in apartment, a wavelet neural network is used to recognize high-power electrical appliances. Since the collected grid current signal is a mixture of the fundamental wave signal and the harmonic signal, the signal separation is required. The wavelet transform based on Mallat fast algorithm was used to extract harmonic current signal, and the average power increment of the total current and that of the harmonic current were normalized to be the feature vector for the recognition of the high-power electrical appliances, and the fusion type wavelet neural network was trained with the feature vectors based on BP algorithm. The trained wavelet neural network was used to identify the unknown data of the grid current, which could realize the online identification and early warning of high-power electrical appliances. The results of contrast simulation experiment show that the wavelet neural network has higher accuracy than the traditional BP neural network in the identification of high-power electrical appliances.
In order to identify the high-power electrical appliances in apartment, a wavelet neural network is used to recognize high-power electrical appliances. Since the collected grid current signal is a mixture of the fundamental wave signal and the harmonic signal, the signal separation is required. The wavelet transform based on Mallat fast algorithm was used to extract harmonic current signal, and the average power increment of the total current and that of the harmonic current were normalized to be the feature vector for the recognition of the high-power electrical appliances, and the fusion type wavelet neural network was trained with the feature vectors based on BP algorithm. The trained wavelet neural network was used to identify the unknown data of the grid current, which could realize the online identification and early warning of high-power electrical appliances. The results of contrast simulation experiment show that the wavelet neural network has higher accuracy than the traditional BP neural network in the identification of high-power electrical appliances.
引文
[1]周维龙,欧阳洪波,胡姣,等.大功率电器智能识别系统设计[J].湖南工业大学学报,2014,28(1):44-48.
[2]孙立辉,王海.基于单片机的宿舍多功能用电监控系统的设计[J].现代电子技术,2016,39(4):135-138.
[3]周维龙,肖伸平,陈刚,等.基于物联网的大功率电器监控系统设计[J].湖南工业大学学报,2012,26(5):95-99.
[4]杭佳卉.基于小波变换和HHT的微网建模及其谐波检测方法的研究[D].淮南:安徽理工大学,2016.
[5]党克,张超.基于小波分析的谐波电流检测[J].电气开关,2015,53(6):39-41.
[6]边静.基于小波多分辨分析的谐波检测应用研究[J].集宁师范学院学报,2015,37(1):109-111.
[7]曹永峰,赵燕君.基于GA-BP神经网络的计算机智能化图像识别技术探究[J].应用激光,2017,37(1).
[8]袁圃,毛剑琳,向凤红,等.改进的基于遗传优化BP神经网络的电网故障诊断[J].电力系统及其自动化学报,2017,29(1):118-122.
[9]刘爱国,黄泽平,薛云涛,等.基于遗传算法小波神经网络的光伏微网发电预测[J].电测与仪表,2017,54(7).
[10]公茂法,李美蓉,殷凡姣,等.基于小波包和改进BP神经网络的变压器励磁涌流识别方法[J].电测与仪表,2015,52(6):124-128.
[11]肖书敏,闫云聚,姜波澜.基于小波神经网络方法的桥梁结构损伤识别研究[J].应用数学和力学,2016,37(2).
[12]雷汝海,郝震.基于小波变换Mallat算法的电网谐波检测方法[J].工矿自动化,2014,40(12):65-69.
[13]盛梦娇.非侵入式电器识别算法的研究[D].青岛:中国海洋大学,2015.
[2]孙立辉,王海.基于单片机的宿舍多功能用电监控系统的设计[J].现代电子技术,2016,39(4):135-138.
[3]周维龙,肖伸平,陈刚,等.基于物联网的大功率电器监控系统设计[J].湖南工业大学学报,2012,26(5):95-99.
[4]杭佳卉.基于小波变换和HHT的微网建模及其谐波检测方法的研究[D].淮南:安徽理工大学,2016.
[5]党克,张超.基于小波分析的谐波电流检测[J].电气开关,2015,53(6):39-41.
[6]边静.基于小波多分辨分析的谐波检测应用研究[J].集宁师范学院学报,2015,37(1):109-111.
[7]曹永峰,赵燕君.基于GA-BP神经网络的计算机智能化图像识别技术探究[J].应用激光,2017,37(1).
[8]袁圃,毛剑琳,向凤红,等.改进的基于遗传优化BP神经网络的电网故障诊断[J].电力系统及其自动化学报,2017,29(1):118-122.
[9]刘爱国,黄泽平,薛云涛,等.基于遗传算法小波神经网络的光伏微网发电预测[J].电测与仪表,2017,54(7).
[10]公茂法,李美蓉,殷凡姣,等.基于小波包和改进BP神经网络的变压器励磁涌流识别方法[J].电测与仪表,2015,52(6):124-128.
[11]肖书敏,闫云聚,姜波澜.基于小波神经网络方法的桥梁结构损伤识别研究[J].应用数学和力学,2016,37(2).
[12]雷汝海,郝震.基于小波变换Mallat算法的电网谐波检测方法[J].工矿自动化,2014,40(12):65-69.
[13]盛梦娇.非侵入式电器识别算法的研究[D].青岛:中国海洋大学,2015.