3D Temperature Field Reconstruction Method For Battery Stack Based On An Improved MRF-KFCM Effective Region Segmentation

As the core and carrier for steady state operation and energy management of microgrid,the energy storage system has the functions of peak and valley cutting,voltage and frequency modulation,emergency standby,etc.Lithium-ion battery has the advantages of large energy density,large power density,high charge and discharge efficiency,fast response speed,etc.It is an energy storage battery with better comprehensive performance and low cost.Temperature monitoring is an important basis for thermal management and safety protection of lithium-ion battery stack.In practical applications,most commonly used methods are contact temperature measurement,but it can only reflect the single point temperature of the battery,and the later maintenance costs are high.With the rapid development of computer technology and microelectronic technology,the improvement of infrared diagnostic technology has been promoted,which has brought opportunities for thermal management of battery stack.However,the use of 2D infrared images to express the temperature distribution of 3D objects has serious limitations,including only obtaining surface temperature information of objects in a single perspective,lacking 3D information of temperature anomalous regions,and inability to accurately locate hotspot locations.The Infrared Thermal Imager cannot monitor the internal temperature filed of the battery stack,which is more important.Aiming at this problem,a 3D temperature field reconstruction method based on surface temperature field and virtual heat source has been proposed.The effective region is extracted by the segmentation algorithm and mapped to the surface temperature field by calibration.Then,the 3D temperature field of the battery stack is preliminarily reconstructed,and then the temperature of sub-unit is modified by virtual heat source.Experimental results show that the proposed method can reflect the variation trend and local difference of the internal temperature of the battery stack,and can timely find the thermal faults in the system,and at the same time,the accuracy can meet the practical application requirements.The infrared image of the battery stack has low contrast and edge gray aliasing.It is very difficult to segment the "interest" region.In order to depress the influence on the 3D reconstruction which caused by the inaccurate region segmentation of infrared image of battery stack,comprehensive use of visible and infrared images,and the Kernel-Based Fuzzy C-Means Clustering algorithm under the constraint of Markov Random Field has been improved.In this method,the preliminary effective region of the battery stack in the visible light image can obtained by Otsu algorithm,and assigning pixels to cluster with different target information weight,then the accurate position in the infrared image is obtained by registration.The experimental results show that the algorithm can divide the target area more completely,and only introduces a small amount of background area.The improved MRF-KFCM-based segmentation algorithm proposed in this paper has good robustness,and can effectively segment images of known target centroids or possible regions of the target.The proposed 3D temperature field reconstruction method based on surface temperature field and virtual heat source provides new ideas and methods for solving heat conduction problems,converts inverse problems into positive problems,simplifies the calculation complexity of 3D reconstruction,and has certain practical value.