LAUSR

Abstract We apply a class of nonlinear semiparametric models to the problem of transient stability analysis of a large-scale power system. The mapping between the pre-contingency state and the transient stability boundary (TSB) is modelled by a block-oriented structure known as the single-index model (SIM). In this model one has a single dimensional projection that enters the unknown nonlinearity nonparametrically. Such models form a rich class of nonlinear mappings that includes classical models, e.g., linear and logistic regression, as special cases. The generalized case of the SIM is utilized that is taking into account the sparsity of the projection vector. This yields a low-dimensional nonlinear model suitable for high-dimensional data emerging in modern power systems. The parametric part of the model is obtained by the properly modified sparsity sensitive LASSO algorithm. The nonlinear nonparametric part of the model is estimated by the monotonically corrected kernel regression estimate. The precision of our modeling is verified for two fault cases of the 470-bus power system. It is shown that for the examined faults, the proposed modeling methodology exhibits a stronger prediction accuracy compared to the existing competing methods recently applied in the field, namely, kernel ridge regression, as well as LASSO linear modeling.