The traditional particle swarm optimization (PSO) algorithm still has shortcomings in terms of performance and efficiency of cloud computing task scheduling, such as low local search efficiency and limited search accuracy, which often makes it difficult to find the global optimal solution and easily falls into the local optimal solution. To solve this problem, an improved particle swarm optimization task scheduling algorithm(IPSO) was proposed. Firstly, a opposition-based learning strategy was used to create a more homogeneous initial population and the Rate of convergence of this algorithm was enhanced. Secondly, in the particle update process, the sine cosine algorithm(SCA) was introduced to enhance the optimization ability of the particles and balance the two processes of global search and local development. Finally, a search behavior based on average fitness was added to further expand the search solution space to find better optimal solutions and prevent falling into local optima. Experimental verification was conducted on the CloudSim simulation platform. The experimental results show that the improved particle swarm algorithm has significant advantages in reducing the cost and maximum completion time of system tasks. In particular, when the number of tasks reaches 500, IPSO improves the total cost by 10%, 4.6%, 8.6%, 9.2%, 8.2%, 10.4% and 11.3% respectively compared with adaptive particle swarm optimization (AdPSO), sine cosine algorithm-particle swarm optimization (SCA-PSO), simulated annealing particle swarm optimization (SAPSO), enhanced phagocytosis genetic algorithm (EPGA), competitive crossover mechanism genetic algorithm(C2PGA), opposition based learning-particle swarm optimization (OBL-PSO) and PSO, and improves the maximum completion time by 34.1%, 27%, 41.7%, 28.5%, 21.6%, 50.3% and 54.8% respectively, which verifies the feasibility and effectiveness of IPSO in solving cloud computing task scheduling problems under different task scales.