Abstract
Renewable energy planning, electrochemical battery storages and advanced energy management strategies are flexible solutions for transformation towards smart grids, whereas the complex battery cycling ageing is nonlinearly dependent on intermittent renewable supply, stochastic load profiles and dynamic charging/discharging behaviors. In this study, a nonlinear mathematical model is developed to explore effective strategies for smart grids. A general regression learner-based battery cycling ageing prediction method is proposed for quantifications of lifetime battery cycling ageing and battery replacement times, including the database preparation, surrogate model training with typical feature extraction and classification, cross-validation, and performance prediction in various battery groups. A machine learning (ML) algorithm selection approach is proposed through the statistical analysis, to guide the accurate surrogate model development, considering the diversity in dynamic charging/discharging behaviours and intrinsic cycling ageing performances of each battery. Afterwards, a novel battery discharging control strategy is proposed, to address the contradiction between the economic cost-saving and the associated battery replacement cost. Last but not the least, the machine learning-based models are thereafter integrated in the district energy community for technical performance analysis. This study can provide a regression learner-based battery cycling ageing modelling method, a machine learning algorithm selection approach, and a holistic framework for systematic integration with avoidance on techno-economic performance overestimation, which is critical to guide renewable energy planning, electrochemical battery storages, and advanced energy management strategies.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call