Plan recognition is an important field in artificial intelligence. In this paper, a new plan recognition algorithm based on plan knowledge graph to predict future actions was presented. The algorithm is more powerful and simpler, comparing with other algorithms. It can be used to handle the condition of partial observation and predict future actions. The experimental results show that the algorithm is linear-time with the domain knowledge, and it powerful than Jiang's algorithms. The studies on event relations in plan recognition are not benefit to improve the algorithms in plan knowledge graph framework, but also are helpful for improving other recognition algorithms and developing new algorithms.