With the development of technology, the power electronics have widely be used. This would make power quality draw much attention. Among of them, the harmonic pollution is an important disturbance source, which would influence the operation of power electronic devices. For the general end users, it may only cause the short-term power interruption in the electrical apparatus. But for the large electrical equipments, harmonics could lead to fault and even cause economical problem. Therefore, the detection of power quality is a noticeable issue. In order to provide good power quality, it is necessary to perform power system harmonic analysis firstly and propose the improved strategies. Among numerous literatures, the fast Fourier transform (FFT) is one of the widely used approaches. However, most research has pointed out that it is necessary to meet the requirements when directly adopting the fast Fourier transform. Therefore, the adaptive linear neural network (ADALINE) with simple structure and fast convergence will be discussed in this thesis. In the experiment of harmonic detection, it is fount that once the fundamental frequency deviation is present, the estimation results would be inaccurate, especially for the detection of phase angles. To enhance the practicality of adaptive linear neural network, this thesis makes the improvement for fundamental frequency deviation. The fundamental frequency can be estimated accurately by Prony’s method. Finally, IEEE 1459-2000 trial-use standard would be adopted to perform the measurement of power quantities. In this thesis, the graphical development environment LabVIEW would be used to build the measurement system and to verify the performance of proposed approach. From the experiment, it is shown that the proposed algorithm not only can achieve fast convergence but also can be very accurate without the interference of the fundamental frequency deviation.
