A High-Precision Torque Control Method for New Energy Vehicle Motors Based on Virtual Signal Injection

The operating temperature of new energy vehicles fluctuates significantly, and variations in motor temperature lead to changes in parameters. These changes introduce errors into the motor’s mathematical model, reducing torque accuracy and causing deviations in the Maximum Torque Per Ampere (MTPA). This paper proposes a Gated Recurrent Unit (GRU) neural network-based torque observer that employs virtual signal injection. Specifically, this method innovatively injects a virtual constant signal into the d-q axis current inputs processed by the neural network to derive the partial derivatives of torque concerning the d-axis and q-axis currents. Subsequently, it calculates the derivative of torque concerning the current vector angle (β) using the total differential equation. By leveraging these partial derivatives, the motor parameters are identified online, and the MTPA current reference value is dynamically adjusted based on the identified parameters. Additionally, the GRU’s internal parameters are fine-tuned in real time using the least mean square (LMS) algorithm, which adjusts based on the derivative of torque concerning the current angle and the error between the observed and actual values, thereby enhancing the accuracy of torque observation, and bringing results closer to the true shaft-end torque. Finally, experimental validation confirms the effectiveness and superiority of the proposed method.