The high?performance permanent magnet motor system has become a key and core component of hard alloy production molding equipment. The mismatch of motor system parameters will seriously affect the overall efficiency of alloy product molding equipment. In order to mitigate the effects of model parameter mismatch and one beat delay on the current control performance of permanent magnet synchronous motor (PMSM), a robust deadbeat predictive current control (POC?DPCC) method with online parameter correction is proposed. Firstly, the parameter sensitivity of conventional deadbeat predictive current control is analyzed. Then, a multi?parameter error online identifier based on Adaline neural network is designed, and a new identification structure with parameter mismatch error as the neural network weight is proposed to improve the tracking performance of parameter changes. Finally, the POC?DPCC method is proposed to improve the robustness of the system to the motor parameters by updating the control voltage coefficient matrix, and the sampling current is replaced by the predicted current value at the next time to compensate the influence of the one?beat delay. The effectiveness and robustness of the proposed method under complex operating conditions are verified by comparing simulation and experimental results.