本文首先介绍了有源电力滤波器的基本结构和工作原理，并对一些常规的谐波电流检测方法的优缺点进行了比较。其次，针对传统的谐波电流检测方法的缺陷，提出将神经网络与基于噪声抵消原理的自适应谐波检测相结合，利用径向基函数运算量小、收敛快、无局部极小值等优点，构造了一种基于径向基函数神经网络的谐波电流检测方法，仿真结果表明该检测方法具有很好的动态响应及畸变电流检测精度。最后，设计了一套实验系统，对本文所采用的系统方案进行了实验验证。仿真表明，本文所采用的有源电力滤波器检测系统方案切实可行，能够较好地实现动态检测谐波和无功功率的目的。 -This paper, firstly, introduces the basic structure and principle of APF, and then makes comparison among some conventional harmonics detections methods. Secondly, contraposes the limitation of the traditional harmonic current detecting method based on neural network, presents the method that combines the neural network and auto adaptive harmonic current detecting method based on the principle of noise each other eliminating.As the characteristics that are small calculative quantity, fast converge, without local minimal point of Redial Basis Function (RBF), this dissertation forms a new harmonic current detecting method based on RBF. The simulation results show that the method has many advantages, such as good dynamic response, high accuracy. At last, A system is designed to verify the proposed APF scheme .The results of simulation　show that the proposed APF scheme is practical and can detect harmonics and reactive power effectively.