Comprehensive experimental design of intelligent control technology based on a battery testing platform

This study aims to develop a comprehensive experimental framework for an intelligent control technology course by introducing a battery testing platform. The experiment integrates battery modeling, intelligent algorithm design and simulation, and experimental validation within an engineering-oriented battery testing scenario. This approach addresses the limitations of traditional experiments that rely on open-source datasets and lack practical system-level case studies. This approach systematically improves the modeling, algorithm application, and engineering problem-solving abilities of students. It also enhances the hands-on skills of students, broadens their analytical thinking, and lays a solid foundation for them to become qualified engineering and technical professionals in the future.

This experimental design implements a closed-loop experimental process of “data acquisition–theoretical modeling–algorithm design–result optimization.” Using the NEWARE BTS-4008 battery testing platform, real-time voltage and current data of Panasonic 18650 batteries are collected under constant-temperature conditions. Based on these data, students establish a second-order RC equivalent circuit model and use the particle swarm optimization (PSO) algorithm for parameter identification. Two different methods are used to estimate the state of charge (SOC): one is the model-based extended Kalman filter (EKF) estimation method, and the other is the data-driven random forest (RF) algorithm. A weighted fusion strategy is used to process the estimates from the two methods, improving the accuracy and robustness of the overall estimation scheme.

The parameters identified using the PSO algorithm can effectively reproduce the terminal voltage curve, proving the accuracy and reliability of the parameters and providing a reliable foundation for subsequent SOC estimation. Under dynamic stress test conditions, the performance of fusion and individual methods was evaluated and compared. Compared with the single method, the mean absolute error of the EKF and RF–weighted fusion methods decreased by 16% and 47%, respectively, the root mean square error decreased by 7% and 48%, respectively, and the coefficient of determination (R²) was closer to 1, proving that the fusion method can overcome the limitations of individual methods and significantly improve estimation performance.

Students collected the necessary data through the experimental platform and constructed a battery model based on course knowledge. They then improved the accuracy of the model by adjusting algorithm parameters. This process enhanced their engineering application and modeling abilities while addressing the gap between theory and practice in traditional teaching. This experimental design overcomes the limitations of single algorithm application, achieving battery SOC estimation through the fusion of the EKF algorithm and data-driven RF methods and further guiding students to perform weighted fusion of the estimation results from the two algorithms. This design helps students understand the applicable scenarios, advantages, and limitations of different intelligent algorithms and breaks the rigid thinking of traditional single-solution approaches by integrating real engineering objects and multiple technical paths. It cultivates innovative thinking by encouraging students to analyze problems from multiple dimensions and integrate technical solutions, effectively meeting the training goals of engineering and technical talent development. In the future, the experimental scenarios can be further expanded to continuously improve the depth and breadth of course teaching and provide practical teaching support for talent cultivation in the field of intelligent control.