Real-time axle-torque prediction of agricultural electric tractors using a kalman-algorithm-integrated predictive motor-speed controller

Real-time axle-torque prediction was accomplished by integrating the Kalman filter algorithm (KFA) with a motor-speed controller using a single-motor electric-tractor dynamic model. Statistical analysis confirmed that the measured and predicted axle torques with KFA corresponded closely and linearly for both front and rear axles (at all gear stages) with accuracies of 99 and 97 %, respectively. The predicted axle torques for both axles at all gear stages, excluding the KFA, were significantly different. The prediction accuracy was approximately 58 %, and it failed to adapt to torque fluctuations and sudden loads. The battery state-of-charge (SOC) for all gear stages with KFA could significantly save the battery energy of the electric tractor compared with the battery SOC of the measured and predicted without KFA. Furthermore, traction performance analysis confirmed plow-tillage (light-duty 2500 rpm) suitability. This algorithm is suitable for real-time axle torque prediction in actual agricultural working scenarios using a single-motor electric tractor. Since the hardware of the proposed prediction algorithm was not available for real-time applications in an actual tractor, the measured input dataset was used to simulate a real-world operation scenario. The proposed algorithm can be applied to unmanned agricultural vehicles or agri-robots to predict the required operation torques. This controller can dynamically adjust the motor output power of an electric tractor, ensuring smooth operation and minimizing energy wastage.