Abstract:The accurate construction of tokamak equilibria, which is critical for the effective control and optimization of plasma configurations, depends on the precise distribution of magnetic fields and magnetic fluxes. Equilibrium fitting codes, such as EFIT relying on traditional equilibrium algorithms, require solving the GS equation by iterations based on the least square method constrained with measured magnetic signals. The iterative methods face numerous challenges and complexities in the pursuit of equilibrium optimization. Furthermore, these methodologies heavily depend on the expertise and practical experience, demanding substantial resource allocation in personnel and time. This paper reconstructs magnetic equilibria for the EAST tokamak based on artificial neural networks through a supervised learning method. We use a fully connected neural network to replace the GS equation and reconstruct the poloidal magnetic flux distribution by training the model based on EAST datasets. The training set, validation set, and testing set are partitioned randomly from the dataset of poloidal magnetic flux distributions of the EAST experiments in 2016 and 2017 years. The feasibility of the neural network model is verified by comparing it to the offline EFIT results. It is found that the neural network algorithm based on the supervised machine learning method can accurately predict the location of different closed magnetic flux surfaces at a high efficiency. The similarities of the predicted X-point position and last closed magnetic surface are both 98%. The Pearson coherence of the predicted q profiles is 92%. Compared with the target value, the model results show the potential of the neural network model for practical use in plasma modeling and real-time control of tokamak operations.

Load model of high power magnet power supply for EAST based on GMPSO-BP neural network

Composite Load Models Based on Field Measurements and Their Applications in Dynamic Analysis

A new generalized load model of power systems and its applications

Generalized Load Modeling of Distribution Network Integrated with Direct-drive Permanent-magnet Wind Farms

Automatic Post-disturbance Simulation Based Method for Power System Load Modeling

Research on Traction and Levitation Characteristics of High-Speed Maglev Train Based on BP Neural Network

Simplified Model of Distributed PV Generation and Influence Analysis on Load Characteristics

Modeling Methods for Electric Loads in a Real Power Grid

Influence of Load Models in Sending and Receiving Ends of Henan Power Grid on Tie-line Stability Limit of Sending End

Load Modeling for Wide Area Power System

Prediction for Tokamak High-power Pulse Load based on Elman Neural Network

Reconstruction of Poloidal Magnetic Fluxes on EAST based on Neural Networks with Measured Signals

Cogging force estimator based on BP neural network of PMLSM

Modeling and Optimization for Submersible Pump Tubular Linear Motor Based on Dynamic Load Analysis

Experiment data-driven modeling of tokamak discharge in EAST

On-Load Field Prediction of Surface-Mounted PM Machines Considering Nonlinearity Based on Hybrid Field Model

Research on Load Prediction of Low-Calorific Fuel Fired Gas Turbine Based on Data and Knowledge Hybrid Model

No-Load Characteristic Analysis of Single/Three-Phase Dual-Port Permanent Magnet Synchronous Generator With Eccentric Magnetic Metal Block Based on Nonlinear Magnetic Field Analytical Method

Prediction of Load Model Based on Artificial Neural Network

Research on electromagnetic torque and fault parameter identification of pumped storage machine under power generation state based on optimized network parameter method

Midterm Power Load Forecasting Model Based on Kernel Principal Component Analysis and Back Propagation Neural Network with Particle Swarm Optimization