基于正则化投影孪生支持向量机的电力系统暂态稳定评估

提出一种基于正则化投影孪生支持向量机的暂态稳定评估方法。将基于传统支持向量机进行暂态稳定评估的高维二项式优化问题转化为两个低维二项式优化问题,并在投影孪生支持向量机的目标函数中引入正则项来改善评估稳定性。首先,构建由系统特征和投影能量函数特征组成的初始样本集,通过特征选择对初始特征进行压缩,获取可有效表征暂态稳定性的最优特征集。然后,基于正则化投影孪生支持向量机的思想将暂态稳定状态分成稳定类与不稳定类,寻找各稳定状态的最佳投影坐标轴,使稳定类投影到稳定类投影超平面上后尽可能地聚成簇,而不稳定类投影到稳定类投影超平面上后尽可能远离稳定类聚成的簇,降低暂态稳定评估的计算时间,同时借助遗传算法进行参数选择以提高准确率。最后,通过IEEE-145和南方电网算例的仿真分析,验证了所提方法的有效性和准确性。

Transient Stability Assessment of Power System Based on Projection Twin Support Vector Machine with Regularization

A method is proposed to assess the power system transient stability using projection twin support vector machine with regularization(RPTSVM). The high dimensional binomial optimization problem of transient stability assessment(TSA)based on support vector machine is transformed into two low dimensional binomial optimization problems. The regularization term in the objective function of PTSVM is added to improve the stability of assessment. Firstly, a group of classification features are extracted from the power system operation parameters to build the original features set, such as feature of power system and the feature of projection energy function. The approach of feature selection is employed to evaluate the classification capability of the original features. Secondly, the optimal feature set is determined to effectively reflect the transient stability of the power system. Then, the features set are divided into stable classes and unstable classes based on RPTSVM. The best projection axis for stable classes and unstable classes are found, so that the class of stable are projected onto the projection hyper-plane of stable class and then clustered as much as possible. While the class of unstable is projected onto the class of stable projection super-plane as far as possible away from the cluster of stable clusters. The computation time of the transient stability assessment is reduced. At the same time, the genetic algorithm is used for parameter optimization and the accuracy of transient evaluation methods is improved. Finally, the simulated results of classic IEEE 145-bus system and China Southern power grid demonstrate the feasibility and validity of the proposed method.