Evaluating the Importance of Samples in Deep Learning Based Transient Stability Assessment

Deep learning based transient stability assessment has achieved big success in power system analysis. However, it is still unclear that how much of the data is superfluous and which samples are important for training. In this work, we introduce the latest technique from the computer science community to evaluate the importance of the samples used in deep learning model for transient stability assessment. From empirical experiments, it is found that nearly 80% of the low importance samples can be pruned without affecting the testing performance at early training stages, thus saving much computational time and effort. We also observe that the samples with fault clearing time close to the critical clearing time often have higher importance scores, indicating that the decision boundary learned by the deep network is the transient stability boundary. This is intuitive, but to the best of our knowledge, this work is the first to analyze the connection from sample importance aspects. The ultimate goal of the study is to create a tool to generate and evaluate some benchmark datasets for power system transient stability assessment analysis, so that various algorithms can be tested in a unified and standard platform which could verify and compare the performance of the algorithms.