Verification and validation of the Gaussian Process Regression model to predict the State of Health in Lithium-Ion Batteries

Garay F.; Huaman W.; Munoz J. · Proceedings of the 2022 IEEE 29th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2022 (2022) · DOI: 10.1109/intercon55795.2022.9870158

Resumen

Currently, lithium-ion batteries are the definitive standard used for the storage of electrical energy in modern vehicles, passenger aircraft, and smart mobile devices. For this reason, the industries seek to manage these components due to their high variability and atypical behavior. By other hands, the capacity inside shows an indicator used as a qualitative measure to determine the capacity nominal to store and provide energy to a system or device. Nowadays, using artificial intelligence and advanced resource computational, it is possible to obtain models data-driven, combining mathematical algorithms that optimize the management of the life cycle and prognostic the state of health. The Gaussian process regression algorithm is proposed to develop a model and validate it in another dataset, it guarantees the performance, reliability, and safety of the battery. For the verification stage, the lithium-ion battery dataset of the NASA research repository was used, and to ensure the effectiveness of the model, the dataset was separated into 30% for the training and 70% for the testing during the research. For the validation stage was added a noise, two datasets were used, and compared their results using metrics quantitative. The experimental showed results R2 > 98% and MAE and RMSE < 0.02 these confirmed that the proposed methodology demonstrates high prediction accuracy of the proposed in practice.