新型虚拟磁链定向的Hopfield神经网络谐波电流检测

摘要

The real time performance and accuracy of harmonic current detection are directly related to the compensation accuracy of active power filter.Compared to the phase locked loop, virtual flux can obtain more accurate orientation as well as faster response even in the case of the grid voltage fluctuations.In this paper, the new algorithm is used in this paper which is consisted of a high-pass filter and transformed coordinate system, which improves the phase error caused by the integrator in the virtual flux algorithm.Using Hopfield neural networks to detect harmonic current, this method constructs the feedback neural network based on energy function, through continuous feedback convergence and reached balance in the end.Simulation results illustrate that the method of Hopfield neural network with virtual flux orientation can simultaneously detect the grid fundamental and total harmonic as well as the specific number of harmonic currents quickly and accurately.In the situation of load mutation and voltage fluctuation, the detection methods put forward by this paper has excellent real time performance, high accuracy, and more strong robustness compared to the results with ip-iq method.%谐波电流检测的准确性和快速性直接影响有源电力滤波器的补偿精度.相较于锁相环,虚拟磁链在电网电压波动的情况下也能准确定向,而且响应速度更快.文中采用由高通滤波器和变换坐标系组成的新型辨识算法,改善了虚拟磁链中由积分器引起的相位误差.Hopfield神经网络利用新型虚拟磁链观测到的电网电压定向角,对谐波电流进行检测.该方法引入能量函数构造反馈型神经网络,通过不断反馈收敛最后达到平衡,以实现对谐波电流的检测.仿真结果表明基于虚拟磁链的Hopfield神经网络方法能同时快速准确地检测出基波电流、总谐波电流和各次谐波电流,而且在负载突变、电压波动的情况下,与基于瞬时无功功率理论的ip-iq法相比响应速度更快,准确性更高,具有更好的鲁棒性.