The estimation accuracy of the battery state of charge (SOC) directly affects the reliable and stable operation of the battery management system (BMS). In complex operating conditions such as vehicle-mounted applications and energy storage power stations, systematic biases and random noise are common in voltage and current sensing measurement. Most existing studies on SOC error focus on the independent effect of individual measurement error and lack systematic explanation on the propagation mechanism and quantitative influence of multi-source errors. In this paper, based on the second-order RC equivalent circuit model, the analytical transfer model of SOC estimation error is established under the framework of the extended Kalman filter (EKF), and the transfer path of measurement error in the algorithm is clarified. Then, a battery test bench with “reference-measurement” dual-channel parallel measurement is built. Under the constant-current discharge condition, the contribution weight quantification and sensitivity analysis of multi-source measurement error to SOC estimation error are completed, and it is revealed that the voltage measurement bias is the dominant error source of SOC estimation. On this basis, an augmented-state modified EKF algorithm (AM-EKF) with online identification of voltage bias is proposed. The voltage measurement bias is incorporated into the state vector to realize the joint estimation and online compensation of the operating state and measurement bias. The results show that under the constant-current discharge condition, the improved algorithm increases the proportion of SOC estimation error within the ±3% qualified interval to more than 85% in the whole discharge period. The maximum estimation error is significantly lower than that of the traditional EKF algorithm, and the accuracy is close to the unbiased ideal condition. Under the modified DST dynamic condition, the improved algorithm controls the SOC estimation error within 5%, which is 72% lower than the traditional EKF calculation error with bias input. The scheme has high robustness, no additional hardware cost, controllable computational complexity and excellent engineering application value.

To address the challenges of motor system temperature rising and the failure of temperature-sensitive components in electric agricultural robots, this study focuses on the cooling design for the motor system of a specific electric weeding robot using simulation and experiment methods. Optimal cooling fan configuration was proposed, and the safe operational fan airflow volume range at varying working ambient temperatures was determined. First, the thermal properties of the physical model, including component materials, internal heat sources, and fluid conditions, were analyzed. A three-dimensional, full-domain simulation model was then developed, and steady-state fluid-thermal coupling field calculations were performed for three different fan configurations. These simulations identified the optimal fan configuration and highlighted the motor as the primary heat-generating component. Under this configuration, the cooling performance improved by 22.7% and 22.3% compared to the other two configurations. Furthermore, based on the optimal fan setup and the operational temperature characteristics of the electric weeding robot, the relationships between motor temperature, fan airflow, and ambient temperature were analyzed and modeled. Real-time, high-precision steady-state temperature experiments were conducted to validate the airflow-motor temperature relationship, leading to the establishment of a safe operational airflow range based on the maximum allowable motor temperature. The implementation of this safe airflow range reduces the probability of motor system failure by at least 60%. This work provides theoretical and practical insights for enhancing the cooling design of motor systems in electric agricultural robots.