In order to reduce the impact of lithium-ion battery sensor faults on the safety and performance of electric vehicles, an observer-based fault diagnosis scheme was presented to detect and isolate battery sensor faults in this paper. The proposed scheme constructed two extended Kalman filter (EKF) observers in combination with the coupling electro-thermal dynamic model of Li-ion battery to realize state estimation. The difference between the estimated value and the sensor measured value generated the residual. Then the residuals were evaluated by statistical cumulative sum(CUSUM) test that determined the presence of the faults. According to the respond of two residuals, the fault diagnosis and isolation (FDI) of the current sensor, the voltage sensor and the surface temperature sensor could be realized. The proposed scheme was tested to verify its effectiveness. The result shows that the proposed scheme can diagnose and locate three kinds of lithium-ion battery cell sensor faults in time and accurately, demonstrating excellent performance and easy implementation.