快速K-S变换理论与谐波时频参数自适应分析关键技术研究

谐波时频特征参数是谐波电能计量与谐波治理最主要的依据，但传统的FFT谐波分析算法只能进行单一的时域或频域分析，以S变换为代表的时频联合分析是未来电力谐波分析的发展方向，但受Gauss窗的制约，目前的S变换存在分辨率低、计算量大等问题。为此，本项目提出一种新的K-S变换理论，研究包括：①优化Kaiser窗，建立窗系数与特征频率的自适应调节机制，构建性能卓越的Kaiser自适应优化窗；②依据FFT→Kaiser自适应优化窗→IFFT的思路，推导时频分析自适应算法，建立K-S变换；③从变采样率、调窗形、剔除非特征信息等多方面高效简化变换的计算量，形成快速K-S变换算法，提高实时性；④构建谐波检测实验平台，利用DSP实现快速K-S变换，实现谐波时频特征参数的快速、准确辨识。本项目将大幅度提高电力谐波分析的效率，提出的快速K-S变换理论与谐波时频分析新方法可望成为复杂信号快速时频分析的最有效方法。

Abstract: Harmonic time-frequency characteristic parameters are the most important foundation for harmonic energy meter and harmonic treatment. However, the traditional harmonic analysis methods based on FFT can only conduct either time-domain or frequency-domain analysis. The time-frequency analysis methods represented by S transform, are the development direction for power harmonic analysis in the future. However, restricted by Gauss window, the S transform suffers the problems such as low resolution and large computational efforts, etc. Thus a new K-S transform is proposed in this project. The major studies are as follows. ① Adaptive optimizing Kaiser window with adaptive regulatory mechanism for coefficient of window based on characteristic frequency is built as the result of improving classical Kaiser window. ② Based on the idea as FFT→Adaptive optimizing Kaiser window→IFFT, the adaptive time-frequency analysis method is deduced and the K-S transform is established. ③ The amount of computation is reduced effectively by changing sampling rate, adapting shape of window or removing non-feature information, etc. Thus the fast K-S transform is built, enhancing real-time performance of harmonic time-frequency analysis. ④ The test platform for power harmonic measurement is built. The fast K-S transform is implemented on DSP, realizing the quick and accurate identification of harmonic time-frequency characteristic parameters. The efficiency for power harmonic analysis is improved substantially in this project. The proposed K-S transform and harmonic time-frequency analysis method are expected to become the most effectively tools for fast time-frequency analysis of complex signal.