Real-time classification of L-H transition and ELM in KSTAR

In order to automate the control of ELM and plasma density, an algorithm for classification of L-H transition and the first ELM burst in real-time has been developed with KSTAR campaign data. We applied a Long Short-term Memory (LSTM) which is a special kind of recurrent neural networks for storing long time-series data to classify the L-H transition phenomena and the first ELM burst. For training the LSTM classifier, we used D-alpha and line-averaged electron density diagnostic data that indicate the characteristic patterns of the L-H transition. After the training, we tested two versions for real-time experimental data and post-processed data. In addition, through the results related to the long intermediate states, we examined whether the statistical approach can reflect physical aspect of the KSTAR plasmas.