This paper presents a novel energy balance algorithm based on Bayesian Game for wireless sensor network, which divides each data forwarding process into two game phases. In the first stage of the game, a node constructs a static Bayesian game model according to its own energy levels and strategies of other nodes involved in the game. Then the node takes the solution of the optimized expected revenue function as the routing and forwarding packets optimal decision probability. In the second game stage, the source node and the neighbor nodes build game model according to the energy level and the mutual strategies and the optimal he number of packets transmitted.