Machine Learning Assisted Optimization of Perovskite Thin Film Fabrication Process and Assessment of Feature Importance
The efficiency of perovskite solar cells has been improved to 26.1%in just ten years,which is very close to the certification efficiency of crystalline silicon solar cells(26.81%).This demonstrates the significant po-tential for industrialization.Currently,efforts are still being made to further enhance the efficiency of perovskite solar cells.However,various inseparable factors affect the performance of perovskite solar cells during the device prepara-tion process.Traditional methods often rely on trial and error to optimize the preparation process,resulting in time-consuming procedures.Bayesian optimization,a global optimization algorithm,has achieved remarkable success in addressing artificial intelligence,s black box problem.In this work,the Bayesian optimization is employed to opti-mize four key process parameters involved in the perovskite layer:excess percentage of lead iodide(PbI2),anneal-ing temperature,annealing time,and vacuum extraction time.The costs of research and development have been sig-nificantly reduced,as well as the required time for such activities has also been shortened.The improvement was achieved through five rounds of experimental iterations and 34 sets of process conditions,ultimately resulting in the preparation of an inverse perovskite solar cell with a device efficiency rating of 23.56%.
perovskite solar cellsmachine learningprocess optimizationhigh efficiency
国家科技期刊平台
NSTL
万方数据
