The hybrid/ electric vehicle (H/EV) market is very dependent on battery models. Battery models inform cell and battery pack design, critical in online battery management systems and can be used as predictive tools to maximise the lifetime of a battery pack. Battery models require parameterization, through experimentation. Temperature affects every aspect of a battery's operation and must therefore be closely controlled throughout all battery experiments. Today, the private-sector prefers climate chambers for experimental thermal control. However, evidence suggests that climate chambers are unable to adequately control the surface temperature of a battery under test. In this study, laboratory apparatus is introduced that controls the temperature of any exposed surface of a battery through conduction. Pulse discharge tests, temperature step change tests and driving cycle tests are used to compare the performance of this conductive temperature control apparatus (CTCA) against a climate chamber across a range of scenarios. The CTCA outperforms the climate chamber in all tests. In CTCA testing, the rate of heat removal from the cell is increased by two orders of magnitude. The CTCA eliminates error due to cell surface temperature rise, which is inherent to climate chamber testing due to insufficient heat removal rates from a cell under test. The CTCA can reduce the time taken to conduct entropic parameterization of a cell by almost 10 days, a 70% reduction in the presented case. Presently, the H/EV industry's reliance on climate chambers is impacting the accuracy of all battery models. The industry must move away from the flawed concept of convective cooling during battery parameterization.

This paper presents a framework for all-state estimation of Lithium-Sulfur (Li-S) battery cells based on a Long Short-Term Memory Recurrent Neural Network (LSTM RNN) model. Under the proposed framework, the LSTM RNN model is calibrated into the single task of State of Charge (SoC) estimation for fresh Li-S prototype cells. The Adaptive Moment Estimation (Adam) solver is used. Data sets for training and testing are derived from experiments using the WLTP duty cycles. The calibrated LSTM RNN structure is described for the purposes of training and testing with experimental datasets, so as to generate a network that can be deployed in real-time system. The demonstration of the training and testing results has shown robustness of the proposed approach against nonlinearities of the experimental datasets and uncertainty in initial SoC. The approach gave satisfactory estimation performance with an acceptable tradeoff between estimation accuracy and convergence speed.