The original K-means clustering algorithm is seriously affected by initial centroids of clustering and easy to fall into local optima. To overcome these shortages, this paper uses the quantum particle swarm optimization(QPSO) which has power ability of global search and quick convergence rate to optimize the initial clustering centers of the original K-means algorithm. As the QPSO algorithm can easily fall into the local optimum, the local attractor with Gauss disturbance is used to make the population jump out of the local extremum. To improve the convergence speed of the algorithm, the weighted average best position is used to take advantage of the elite particles. The contraction-expansion factors and random variables are combined in order to select the best parameter strategy. The simulation results on various benchmark problems show that the optimization accuracy, convergence speed and stability of the improved optimization algorithm are significantly improved. Experimental results on the typical UCI datasets show that the proposed method is superior to compared algorithms. Finally, this method is applied to the customer classification of table grapes, which shows the effectiveness and practicability of the proposed clustering algorithm. Through the empirical analysis, it is also proved that this model can be promoted and applied.