Predicting resistive wall mode stability in NSTX through balanced random forests and counterfactual explanations

Abstract

Recent progress in the disruption event characterization and forecasting framework has shown that machine learning guided by physics theory can be easily implemented as a supporting tool for fast computations of ideal stability properties of spherical tokamak plasmas. In order to extend that idea, a customized random forest (RF) classifier that takes into account imbalances in the training data is hereby employed to predict resistive wall mode (RWM) stability for a set of high beta discharges from the NSTX spherical tokamak. More specifically, with this approach each tree in the forest is trained on samples that are balanced via a user-defined over/under-sampler. The proposed approach outperforms classical cost-sensitive methods for the problem at hand, in particular when used in conjunction with a random under-sampler, while also resulting in a threefold reduction in the training time. In order to further understand the model’s decisions, a diverse set of counterfactual explanations based on determinantal point processes (DPP) is generated and evaluated. Via the use of DPP, the underlying RF model infers that the presence of hypothetical magnetohydrodynamic activity would have prevented the RWM from concurrently going unstable, which is a counterfactual that is indeed expected by prior physics knowledge. Given that this result emerges from the data-driven RF classifier and the use of counterfactuals without hand-crafted embedding of prior physics intuition, it motivates the usage of counterfactuals to simulate real-time control by generating the β N levels that would have kept the RWM stable for a set of unstable discharges.

Prediction of multifaceted asymmetric radiation from the edge movement in density-limit disruptive plasmas on Experimental Advanced Superconducting Tokamak using random forest

“…[45][46][47] For the former labeling method, if denoting the abnormal event time as t event , then samples between [t event − τ class ,t event ] are labeled as positive and samples before t event − τ class are labeled as negative. In studies of disruption prediction on HL-2A [48,49] the classification time is 100 ms and on J-TEXT it is 50 ms. [40] In resistive wall mode prediction on NSTX the classification time is also set to 100 ms. [34] Similarly, for MARFE movement prediction, a classification time of τ class ∼ 100 ms regarded to MARFE movement time t MARFEmove is applied for sample labeling. For example, in Fig.…”

“…In recent years, machine learning (ML) techniques are popular to predict different instability events for whom the physics mechanism is complex and to whom multiple physics processes and physics parameters are related, such as thermal quench and current quench of major disruptions, [28,29] vertical displacement events (VDE), [30] and tearing modes prediction. [31,32] Among these frequentlyused ML methods, random forest (RF) is a kind of model that has been applied for different tasks such as tearing modes prediction and avoidance, [33] resistive wall mode stability prediction [34] and disruption prediction. [35][36][37][38] Besides, from comparison of various ML models the RF model works best for the task of fault-detection for magnetic probes in terms of speed and accuracy.…”