基于支持向量回归模型的电力系统谐波分析新方法

当前电力系统中的谐波问题日益严重,对谐波的准确检测和分析是抑制谐波畸变的重要依据。将基于改进的SMO算法的支持向量回归模型应用于电力系统谐波的检测,该算法克服了常规算法计算规模大和建模复杂的困难,通过引入一个中间因子,将原来问题的计算规模减半,并利用迭代算法求解中间因子,使得该算法简单可行。对三相桥式整流电路交流侧产生的特征谐波和非特征谐波电流进行了分析,仿真结果通过与FFT算法和ADALINE神经网络的检测分析结果对比,表明该方法无论是在理想情况下还是在考虑了各种影响因素的情况下,都具有很高的检测精度,可以满足电力系统的谐波分析的要求。该方法的不足之处是计算量会随着输入量分辨率的提高而增大。

New approach of power system harmonic analysis based on support vector regression with simplified SMO algorithm

The harm caused by harmonics is getting serious increasingly, and the exact harmonic detection and analysis can provide important foundation for controlling harmonics. An approach for controlling harmonics, based on simplified SMO algorithm for support vector regression was proposed. This algorithm overcomes the defects of the conventional algorithm, which demands a large amount of calculation. By introducing a simplified factor and solving it through iterative algorithm, this algorithm cut down the dimension of task. The algorithm is simple and feasible, which solves the factor through iterative algorithm. The analysis of characteristic and non-characteristic harmonic currents at the AC side of the three-phase bridge rectifier circuit was presented. The simulation results compared with FFT and ADALINE show that the proposed method has good analysis accuracy both in ideal case and complicated case, and it can satisfy the request of harmonic analysis in electric power. The shortcoming of the algorithm is that the amount of computation will be larger if the resolution of the inputs is high.