摘要
针对有源滤波器谐波检测实时精度高的要求,将线性神经网络应用于自适应噪声对消技术,采用最小均方(least mean square,LMS)误差算法对神经网络进行训练,通过线性神经网络实现的自适应格型滤波器,每个神经元对输入基波和谐波信号并行协同处理,对电网高次谐波分量进行滤波和预测,较常规滤波器有更好的实时性和鲁棒性.仿真和实验结果分析表明,自适应格型检测算法比常规检测方法有更好的实时性和精度,补偿后5、7次谐波电流含有率明显下降,电力有源滤波器APF投入后补偿效果良好.
High real-time precision of harmonic current detection is vital for the performance of active power filter （APF）.In this paper linear neural network was applied to adaptive noise cancellation technology, and the neural network was trained by least mean square （ LMS ） algorithm. Thus the adaptive lattice filter was realized with better real-time performance and robustness than traditional filter. Each neuron processes fundamental and harmonic signals in parallel to predict and filter the high harmonic currents. Simulation and experimental results show that the new adaptive lattice detection method has a better precision and real-time performance.After compensation, 5th and 7th harmonic current is obviously reduced. Therefore, APF has better compensation effect than traditional solution.
出处
《天津大学学报》
EI
CAS
CSCD
北大核心
2008年第4期408-412,共5页
Journal of Tianjin University(Science and Technology)
基金
高等学校博士学科点专项科研基金(20060056054)
天津市自然科学基金(05YFJMJC11500)
关键词
有源滤波器
谐波检测
线性神经网络
自适应格型滤波器
active power filter （ APF ）
harmonic detection
linear neural network
adaptive lattice filter
