电网信号谐波分析算法计算性能比较研究

Research on the performance comparison of harmonic analysis algorithms for power grid signals

电网信号谐波分析算法种类多,适用于不同的负荷场景,但它们的性能优劣,目前仍没有一个统一的性能评价指标体系,致使很多用户无法根据自己的特定需求,更合理、快速地选择出最适用的谐波分析算法。针对于此,文中比较研究了621种电网信号谐波分析算法的计算性能。具体地,对众多窗函数,按频域特性的不同进行了分类;并根据计算原理,将电网信号谐波分析算法分成了加窗FFT插值、加窗FFT和谐波分组算法等三大类,进而比较分析了这些算法的计算准确度和计算的硬软件开销等性能;探究了非整周期采样和噪声等对算法计算准确度的影响;基于对供电系统电压、电弧炉电流以及风电机组电流等典型信号的分析计算结果,选取出了推荐算法;最后,还对稳态谐波分析算法在被测信号发生典型动态变化下的计算性能进行了测试。所得结论对不同电力用户依据自己的特定需求、选用合适的谐波分析算法以及开展相关研究等方面,可提供有说服力的指导和帮助。

There are many types of harmonic analysis algorithms for grid signals,which are suitable for different load scenarios. However,their performance are different,there is still no unified performance evaluation index system,which makes many users unable to choose the most suitable harmonic analysis algorithm according to their specific needs more reasonably and quickly. In view of this,this paper compares the computational performance of 621 kinds of harmonic analysis algorithms for grid signals. Specifically,many window functions are classified according to different characteristics in frequency domain. According to the principle,the harmonic analysis algorithms for grid signals are divided into three categories: windowed FFT interpolation,windowed FFT and harmonic grouping algorithms. Furthermore,the computational accuracy as well as the computational cost of hardware and software of these algorithms are compared and analyzed. This paper also investigates the effects of nonsynchronous sample and noise on the computational accuracy of the algorithms.Based on the analysis results of typical signals,namely voltage of the power supply system,the electric arc furnace current and the wind turbine current,recommended algorithms are selected. Finally,the computational performance of the steadystate harmonic analysis algorithms on the measured signals with typical dynamic changes is tested. The conclusions obtained in this paper provide convincing guidance and help for different users to select the appropriate harmonic analysis algorithms and to carry out related research according to their specific needs.