Speaker
Description
Impact
The results that we are obtaining with the use of EAs and GAs applied to the improvement of the equilibrium are very promising, representing a great improvement compared to other approaches. The large number of configuration parameters and the computational cost of the application used to modellize the equilibrium, represent a big problem even with our approach, but we have simplified this problem by means of the automation of the configuration of the algorithms, so any researcher can easily run the system using the grid and get good configurations for a stellarator fusion device. We present how we have improved the equilibrium of a stellarator by means of different algorithms.
URL for further information
http://www-fusion.ciemat.es/New_fusion/en/
Conclusions and Future Work
The improvements achieved with our GAs and EAs show how the grid can help to get good configurations for a nuclear fusion device . The obtained results are very promising. As future work we think of a distributed EA based on 'bees' algorithm which we hope can offer better results as long as it can be better adapted to the distributed paradigm of the grid.
Keywords
vmec, equilibrium, evolutionary, genetic, algorithm
Detailed analysis
The grid has been used for many scientific purposes in the last years. The fusion community is becoming more important as long as fusion energy is considered the next generation of energy. Many different modelling tools can be used to predict the behaviour of nuclear fusion devices, but most of these tools require ahigh computational costs that only distributed paradigms like grid computing or supercomputers can offer. We are using a code called VMEC to measure the equilibrium of a nuclear fusion device and a set of genetic and evolutionary algorithms (GAs and EAs) to look for good configurations for these devices. If we want to improve a nuclear fusion device, taking into account the equilibrium of confined plasma, the exploration of the whole solution space by means of a brute-force algorithm is not a good approach due to the long number of different configurations.
