Graphite//LFP synthetic training prognosis dataset

Published: 06-05-2020| Version 1 | DOI: 10.17632/6s6ph9n8zg.1
Matthieu Dubarry


This training dataset was calculated using the mechanistic modeling approach. See the “Benchmark Synthetic Training Data for Artificial Intelligence-based Li-ion Diagnosis and Prognosis“ publication for mode details. More details will be added when published. The prognosis dataset was harder to define as there are no limits on how the three degradation modes can evolve. For this proof of concept work, we considered eight parameters to scan. For each degradation mode, degradation was chosen to follow equation (1). %degradation=a × cycle+ (exp^(b×cycle)-1) (1) Considering the three degradation modes, this accounts for six parameters to scan. In addition, two other parameters were added, a delay for the exponential factor for LLI, and a parameter for the reversibility of lithium plating. The delay was introduced to reflect degradation paths where plating cannot be explained by an increase of LAMs or resistance [55]. The chosen parameters and their values are summarized in Table S1 and their evolution is represented in Figure S1. Figure S1(a,b) presents the evolution of parameters p1 to p7. At the worst, the cells endured 100% of one of the degradation modes in around 1,500 cycles. Minimal LLI was chosen to be 20% after 3,000 cycles. This is to guarantee at least 20% capacity loss for all the simulations. For the LAMs, conditions were less restrictive, and, after 3,000 cycles, the lowest degradation is of 3%. The reversibility factor p8 was calculated with equation (2) when LAMNE > PT. %LLI=%LLI+p8 (LAM_PE-PT) (2) Where PT was calculated with equation (3) from [60]. PT=100-((100-LAMPE)/(100×LRini-LAMPE ))×(100-OFSini-LLI) (3) Varying all those parameters accounted for more than 130,000 individual duty cycles. With one voltage curve for every 100 cycles. 6 MATLAB© .mat files are included: The GIC-LFP_duty_other.mat file contains 12 variables Qnorm: normalize capacity scale for all voltage curves P1 to p8: values used to generate the duty cycles Key: index for which values were used for each degradation paths. 1 -p1, … 8 - p8 QL: capacity loss, one line per path, one column per 100 cycles. File GIC-LFP_duty_LLI-LAMsvalues.mat contains the values for LLI, LAMPE and LAMNE for all cycles (1line per 100 cycles) and duty cycles (columns). Files GIC-LFP_duty_1 to _4 files contains the voltage data split into 1GB chunks (40,000 simulations). Each cell corresponds to 1 line in the key variable. Inside each cell, one colunm per 100 cycles.