基于贝叶斯优化弹性网络回归的谐波状态估计方法

Harmonic State Estimation Method Based on Bayesian Optimized Elastic Network Regression

回归类算法在估计系统谐波状态时,谐波源间的高度相关性会引起法矩阵的病态,从而显著影响谐波估计精度。为了更加准确地估计系统谐波状态,提出一种基于贝叶斯优化弹性网络回归的谐波状态估计方法。首先,在用最小二乘法进行谐波状态估计时,将带权值的1范数和2范数同时加入惩罚函数中;另外,为了更加高效准确地估计系统谐波状态,将高斯过程和贝叶斯优化应用于1范数和2范数的权值选取;最后,在IEEE 14节点中验证了所提方法的有效性。结果表明:在谐波源间存在相关性时,所提方法仍能实现合理的谐波源定位及谐波责任划分。

When the regression algorithm is used to estimate the harmonic state of power system,the high correlation between harmonic sources will cause the ill condition of the normal matrix,which will significantly affect the accuracy of harmonic estimation.In order to accurately estimate the harmonic state of the power system,a harmonic state estimation method is proposed based on Bayesian optimization elastic network regression.Firstly,when the least square method is used to estimate the harmonic state,the weighted 1-norm and 2-norm are added to the penalty function at the same time;In addition,in order to estimate the harmonic state more efficiently and accurately,the Gaussian process and Bayesian optimization are applied to select the weight of 1-norm and 2-norm;Finally,the effectiveness of the proposed method is verified in IEEE14 node.When there is correlation between harmonic sources,the proposed method can still achieve reasonable harmonic source location and harmonic responsibility division.