Particle swarm optimization method, system and device based on GPU end acceleration and medium

The invention relates to a particle swarm optimization method, system and device based on GPU end acceleration and a medium, the method firstly migrates data initialization operation to a GPU end to reduce IO loss, then a thread self-adaption and thread multiplexing module is added, and the execution efficiency of a traditional PSO algorithm in the parallel computing process of a large number of particles is further improved from the thread management angle; the system, the device and the medium are used for realizing the particle swarm optimization method based on GPU end acceleration. According to the method, the practical applicability of the PSO algorithm is improved, and the method has an absolute advantage in efficiency in a complex optimization problem. 一种基于GPU端加速的粒子群优化方法、系统、设备及介质，该方法首先将数据初始化操作迁移到GPU端，减少IO的损失，随后又加入了线程自适应与线程复用模块，从线程管理角度进一步提高了传统PSO算法在大量粒子并行计算过程中的执行效率；本发明的系统、设备及介质用于实现上述基于GPU端加速的粒子群的优化方法；本发明提高了PSO算法的实际适用性，在复杂优化问题中具有效率上的绝对优势。