The T-type 3-level converter is key piece of equipment connecting new energy to the grid. If it experiences a current sensor fault, this will lead to the failure of control strategy and affect the safe and stable operation of the grid-tied system. A model predictive control strategy based on current reconstruction is proposed to improve the reliability of the T-type 3-level grid-tied converter. The relationship between phase current and voltage vector is analyzed. Fault current is reconstructed by DC current, normal phase current, and predicted current, respectively. The voltage vectors required for fault current reconstruction are divided into two voltage vector sets according to reconstruction methods. The continuous use of the prediction current to reconstruct the fault phase current will lead to an accumulation error in the reconstruction current. Therefore, it is necessary to detect the voltage vector set at the current moment to ensure that different voltage vector sets are selected at the next moment to avoid current accumulation errors in reconstruction. Experimental results show that the proposed strategy can realize fault-tolerant operation after a current sensor failure of a T-type converter, offers a lower current total harmonic distortion rate, and improves the reliability of the grid-tied system. This work is supported by the National Natural Science Foundation of China (No. U2004166).