To solve the problem of node's premature death in existing power control methods for Wireless Sensor Network (WSN), a new method called Self-Adaptive Fuzzy Control (SAFPC) was proposed. Firstly, the model of two level fuzzy controller with "input-output-feedback" mechanism was designed, whose main controller was responsible for the node transmission power adjustment, and auxiliary controller was responsible for the desired node degree adjustment, so as to adjust the transmission power adaptively according to the residual energy of the node. Secondly, the fuzzification, fuzzy rules and defuzzification process were described in detail. Finally, SAFPC was simulated and analyzed in terms of network convergence time, average energy consumption and network life cycle. The experimental results show that, compared with FCTP (Fuzzy Control Transmission Power method), SAFPC can increase convergence rate by 12.5%, the average energy consumption of the nodes is reduced by 3.68% and the network life cycle is prolonged by 7.9%. It can be seen that SAFPC can effectively prolong the network life cycle, as well as improve network dynamic adaptability and link robustness significantly.