Smart Sampling-based Scenario Selection for Real Power System Data

Abstract

This study develops a comprehensive smart sampling based scenario selection framework that incorporates feature extraction and hierarchical Latin Hypercube sampling with dependence (LHSD) and integrates it into a large scale real-world power system data. The framework allows identification and selection of a small set of representative power system scenarios that maintain the coverage of system characteristics and operation envelope. The issues of multiple data types, high dimensionality, and data unbalancing issues are addressed in the approach with exploratory data analysis and the hierarchical structure. The effectiveness and efficiency of the smart sampling approach is demonstrated with real-world data and a set of performance metrics that measure the capability of reproducing statistical multivariate joint behaviors of the power system.