Graph coloring problem is one of the main optimization problems from the literature. Many real world problems interacting with changing environments can be modeled with dynamic graphs. Genetic algorithms are a good choice to solve dynamic graph coloring problem because they can adopt to dynamic environments and are suitable for problems with NP-hard complexity. In this paper, we propose a dynamic pool based evolutionary algorithm (DPBEA) for solving the dynamic graph coloring problem, which contains a partition based representation to adopt to the dynamic changes of the graph and carry the valuable information obtained in history. The proposed algorithm uses a novel special purpose pool based crossover operator that targets to minimize the number of colors used in the solutions and a local search method that tries to increase the diversity of the solutions. We compared the performance of our algorithm with a well known heuristic for solving the graph coloring problem and a genetic algorithm with a dynamic population using a large number of dynamic graphs. The experimental evaluation indicates that our algorithm outperforms these algorithms with respect to number of colors used by the algorithms in most of the test cases provided.