Research on Robot Battery State of Health Prediction Based on CNN-TCN Model

To ensure the stable operation of robotic equipment and the successful completion of various tasks, predicting the battery state of health (SOH) of robotic devices is crucial. Neural network algorithms have demonstrated excellent performance in battery SOH prediction; however, running such algorithms on robotic equipment often faces challenges of limited resources and insufficient computing power. To address this issue, an innovative battery SOH prediction method based on a CNN-TCN hybrid model is proposed. The model benefits from the parallel convolutional computations of Convolutional Neural Network (CNN) and Temporal Convolutional Network (TCN), enabling efficient utilization of computational resources, fast computing speed, and ease of optimization, making it well-suited for resource-constrained robotic devices. Based on an in-depth study of battery aging mechanisms, four key health indicators were screened and extracted from battery cycle usage data to serve as important inputs for the model. Experimental results show that, compared to standalone CNN and TCN models, the proposed model exhibits superior prediction accuracy and stability. This innovative work holds broad application prospects and significant practical value in intelligent robots, drones, and other battery-dependent automated systems.