Parallel Models Based Spatial Abnormal Detection for Distributed Parameter Process
- Hanxiong LI (Principal Investigator)Department of Systems Engineering and Engineering Management
- Xin Jiang LU (Co-Investigator)
DescriptionThe temperature distribution in the battery system will affect the battery performance, and any abnormality in the temperature field could indicate the potential damage to the battery system. The thermal environment of the battery system belongs to a complex distributed parameter system (DPS) that is difficult to model. Due to the spatiotemporal nature of the DPS, the abnormal detection would be much more difficult than the traditional system described in ordinary differential equations.This project aims to develop parallel models based abnormal detection and spatial localization in surrounding thermal environment of the battery system. This spatial detection scheme could be easily added to the existing battery management system. The major work of the project can be briefly summarized as follows.• First, a novel spatiotemporal modelling method will be developed for the thermal environment of the battery system with independent temporal and spatial components extracted for abnormal detection. The model will be optimized from data-based learning, and the deep learning can be applied for sensor reduction and optimal placement. This DPS model will be further upgraded for online operation with the incremental adaptation.• The fundamental principle is to have two models in operations: one reference model for normal condition (fault-free) and one online model for real-time performance. When the system is in normal condition, difference of these two models should approach to zero mean. If any abnormality appears, the online model will deviate away from the reference model in the statistical sense. The residual signals from this model deviation can be further processed in both temporal and spatial domain for effective detection.• Finally, the probabilistic identification method will be designed to have a reliable detection and spatially localize the abnormality in the system.
|Effective start/end date||1/01/20 → …|