Two-phase separation based spatiotemporal modeling of thermal processes with applications to lithium-ion batteries

Abstract

The thermal process of a battery belongs to distributed parameter systems (DPSs). Modeling of this process is beneficial for energy prediction and conservation. When its first-principle model is not known completely, data-driven spatiotemporal modeling is necessary. Because a DPS is infinite-dimensional and time/space coupled, Karhunen–Loève decomposition (KL) is often used for time/space separation. When modeling a thermal process owning two spatial dimensions, the traditional KL considers two spatial dimensions as a whole without separation. However, in this case, the correlation function, which is critical to time/space separation, would be difficult to evaluate. To remedy this shortcoming, a two-phase separation based spatiotemporal modeling method is proposed. First, one spatial dimension is separated from the process by a set of dominant spatial basis functions (SBFs). Then, the other spatial dimension and the time dimension are separated by another set of dominant SBFs. As a result, the infinite-dimensional time/space coupled process is reduced into a low-dimensional temporal system, which is modeled by radial basis function neural networks. Finally, by spatiotemporal reconstruction, the temperature distribution on the whole space can be predicted. By separating the spatial dimensions, the proposed method is more effective than the traditional KL. Numerical simulations and experiments on the thermal process of a lithium-ion battery have demonstrated the effectiveness of the proposed model.