In this paper, we consider energy fairness problem in wireless sensor networks. However, the heterogeneous hostile operating conditions - different transmission distances, varying fading environments and distinct remained energy levels, have made energy balancing a highly challenging design issue. To tackle this problem, we model the packet transmission of sensor nodes as a game. By properly designing the utility function, we get the Nash equilibrium, in which, while each node can optimize its own payoff, the global objective - energy balancing can also be achieved. In addition, by imposing penalty mechanism on sensors to punish selfish behaviors, the delivery rate and delay constraints are also satisfied. Through extensive simulations, the proposed game theoretical approach is proved to be effective in that the energy consumption is balanced and the energy resources are efficiently utilized, which can significantly improve the network lifetime.