Deep learning based total transfer capability calculation model

A total transfer capability (TTC) calculation model based on stacked denoising autoencoder (SDAE) is proposed in this paper, considering static security, static voltage stability and transient stability constraints. The TTC calculation model consists of feature pre-screening, SDAE and the regression layer. Fast correlation-based filter (FCBF) is used to eliminate irrelevant and redundant features to improve the training efficiency of SDAE. SDAE takes advantage of the deep structure to extract high-order features relevant to TTC from original features. The regression layer is utilized to create the mapping between high-order features and the TTC value. Experiment results of a real power system demonstrate that the proposed TTC calculation model has higher computational accuracy than shallow machine learning models and feature pre-screening decreases the training time of the TTC calculation model obviously.