Performance improvement for Python-based Dynamic System Simulation

PCS (plasma control system) is an important part of the tokamak. China Fusion Engineering Test Reactor has developed DSSim (Dynamic System Simulation) to develop PCS. DSSim allows users to perform simulations by dragging and dropping graphical blocks. Moreover, reinforcement learning has achieved some good results in the field of plasma control. To meet the simulation speed requirements of reinforcement learning, a variety of approaches including rewriting the solver were used for performance improvement. The simulation process of DSSim was optimized, and the time spent on the benchmark model was reduced by 65%. The performance improvement of continuous systems and discrete systems was also tested separately. In addition, the preliminary design of the reinforcement learning interface was done in DSSim. It will provide help for future reinforcement learning controller development.