Data-driven model for divertor plasma detachment prediction

So far the most successful method of reducing divertor heat load in tokamaks is achieved by detachment and yet a fast and accurate detachment prediction model is not available. We present a data-driven surrogate model for divertor plasma detachment prediction leveraging the latent space concept in machine learning research [1]. Our approach involves constructing and training two neural networks - an autoencoder that finds a proper latent space representation (LSR) of plasma state by compressing the multiple diagnostic measurements, and a forward model using multi-layer perception (MLP) that projects a set of plasma control parameters to its corresponding LSR. By combining the forward model and the decoder network from autoencoder, this new data-driven surrogate model predicts a consistent set of diagnostic measurements based on a few plasma control parameters. Benchmark between the data-driven surrogate model and 1D UEDGE simulations shows that our surrogate model is capable to provide accurate detachment prediction (usually in a few percent relative error margin) but with at least 10,000 times speed-up, indicating that performance-wise, it is adequate for integrated tokamak design and plasma control. Data-driven surrogate model using 2D UEDGE simulations will also be reported.