Power systems are highly nonlinear systems that exhibit undesirable oscillations following disturbances. Power system stabilizers (PSS) are usually incorporated to provide auxiliary excitation signals to damp these oscillations. Our objective is to improve the PSS performance via the use of fuzzy logic and LMI techniques. A power system is viewed as a polytopic model that can be adequately represented by a Takagi-Sugeno fuzzy system. A power system stabilizer based on the parallel distributed control principle is suggested. Typically, speed measurements are used as feedback signals. Consequently, a fuzzy observer is included to estimate the unmeasured states. LMI conditions that guarantee the stability and robust pole clustering of the closed loop system are derived. Simulation results of both single-machine and multi-machine models confirm the effectiveness of the proposed algorithm