In order to achieve the safe and reliable operation of grid-connected decommissioned power battery energy storage system, based on long short-term memory (LSTM), model predictive control (MPC) and carrier phase-shift modulation, a coupled LSTM-MPC control and carrier phase-shift modulation method was proposed as an optimal control strategy for decommissioned power battery energy storage system by combining LSTM-MPC coupled control and carrier phase-shift modulation method. Firstly, the realtime current and voltage were used to continuously predict the state of charge (SOC), voltage, current, and output power of each phase under different switching states at the next moment by using LSTM network and discrete model. Secondly, the cost functions corresponding to all the switching states of the converter were calculated, and the optimal control objective with the smallest cost function was established, so that the converter state with the smallest cost function value was used as the input state at the next moment. The converter state with the smallest cost function value was taken as the input state at the next moment. Thirdly, the ratio of the optimal output voltage to the total single-phase voltage of the system was calculated by using the system discrete model, which is used as the duty cycle of the carrier phase-shift modulation algorithm to generate the submodule switching signals, and the process was continuously iterated in a scheduling cycle to perform the feedback correction, so as to realise the optimal control. Finally, the effectiveness of the proposed control strategy was verified in terms of power tracking, SOC tracking, negative sequence current suppression, and power fluctuation suppression for both balanced and unbalanced grid voltage conditions. The simulation results show that the control strategy proposed can effectively achieve the power tracking and SOC tracking of the grid-connected retired power battery energy storage system.