|
Thermonuclear fusion is the use of nuclear fusion reactions to produce energy. With higher energy density and less nuclear waste production than nuclear fission, thermonuclear fusion is considered to be a safer and more sustainable source of energy. However, the extreme conditions required to achieve thermonuclear fusion, such as high temperatures and densities, pose significant challenges to the design, construction and operation of fusion reactions. To overcome these challenges, iterative integrated design with training in thermonuclear fusion modelling and simulation has emerged as an important approach. With the continuous development of the crossover between artificial intelligence and physical design, combined with the powerful fitting capabilities of deep neural networks, deep learning was born, which utilises the highly anthropomorphic features of deep learning (DL) to learn itself through constant interaction and trial and error with the controlled object. Deep reinforcement learning has received great attention for its highly anthropomorphic features, pointto-point design ideas, and low a priori dependence. In this paper, deep learning is used to train and iterate on thermonuclear fusion models and their visualisation simulations to quickly obtain parameters for fusion reactions, significantly shortening the development cycle and providing more options and possibilities for the design and optimisation of fusion reactions, thus avoiding unnecessary costs and waste. The integrated design of thermonuclear fusion modelling and simulation training and iteration provides strong support for the research and application of thermonuclear fusion technology.
|
