Topology optimization for cold plate using neural networks as proxy models

To solve the problems of the high temperature and poor temperature uniformity of lithium batteries, a liquid cooling topology optimization using a neural network as a proxy model is proposed. The reduction of average cell temperature and cold plate pressure drop are taken as the optimization objectives. The effects of basin volume fraction, Reynolds number and boundary conditions on the topological results are investigated. A proxy model is established using neural networks and transfer learning. The differences between the predicted and true values of the source model and the target model do not exceed 10% and 15%, respectively. The improved optimization algorithm is combined with the proxy model. When the number of inlets and outlets is both three, the fitness value reaches 5.33. Compared to before optimization, the target value increased by 37.7%. The maximum battery temperature was reduced by 2.3% and the maximum temperature difference was reduced by 35.3%.