Abstract: The current research in the field of brain-like computing is mainly focused on how to perform large-scale brain-inspired simulations with high performance and low power consumption. NEST brain-like simulator has complete application ecology, can support large-scale simulation and has good scalability, and is the most widely used simulator in the field of brain-like computing. Aiming at the problems of slow running speed and high power consumption in large-scale simulation of NEST simulator, a NEST brain simulator based on heterogeneous computing platform is designed and implemented. The overall performance of the system is optimized by utilizing some useful approaches, including neuron update, data reordering design, multithreaded design, hardware acceleration and software hardware collaboration and etc. To show the effectiveness of design, the NEST simulator is implemented on heterogeneous computing platform named as Xilinx ZCU102. The experimental results show that when perform simulations on the cortical visual model, the performance of neuron update part is 11.9 times that of AMD3600X and 1.2 times that of PYNQ cluster and the energy efficiency is 57.9 times that of AMD3600X and 3.1 times that of PYNQ cluster, while the overall performance of NEST simulator is 2.0 times that of AMD3600X and 2.1 times that of PYNQ cluster and the energy efficiency is 10.1 times that of AMD3600X and 5.8 times that of PYNQ cluster. These results provide a more energy-efficient way for large-scale brain-like simulation based on NEST.