The black hole pattern is a landmark achievement in the study of human moving patterns. However, the black hole pattern has limitations in the evolution modeling of human moving patterns. This paper proposes a black hole pattern with time evolution characteristics. The definition of the new pattern needs to meet the three requirements of group scale, spatial locality and time persistence. This paper proposes a dynamic spatial network model with time evolution characteristics. Based on this model, we define a new black hole pattern and propose a corresponding mining algorithm. In order to improve the efficiency of the pattern mining algorithm, we design a candidate pattern pruning algorithm based on spatiotemporal partitioning, which effectively reduces the searching cost of the mining algorithm in spatiotemporal dimension. Finally, experiments based on real data verify the effectiveness and efficiency of the proposed black hole pattern and mining algorithm.