The performance evaluation of lithium thionyl chloride batteries for long-life meter applications
thesisposted on 17.12.2013, 14:11 authored by Tim G. Saunders
A research project has been carried out to provide BG plc with service life predictions of lithium primary batteries capable of supplying a continuous pulsed power supply to two new electronic domestic gas meters over a desired design life of 11.5 years, in UK weather conditions. The paper study includes investigations of the range of suitable lithium technologies and test methods, and detailed reviews of the discharge processes, polarisation and self-discharge failure modes of the lithium thionyl chloride system. A new multi-channel load test rig and a high resolution measurement system, and software analysis tools were designed and constructed, and some 440 accelerated discharge tests were carried out at different stress levels on 4 cell types. The results provide a unique database of the voltage/temperature/load trends through discharge, and generate a ranking order of performance. Basic statistical analyses have been carried out to quantify the variability of performance trends. A hitherto unreported behaviour pattern is characterised. Qualitative models are postulated to account for deviations from normal behaviour exhibited by two cell types. The analysis suggests that catholyte additives could predispose a system to early failure (due to modification of the crystal structure of the reaction products), and that manufacturing tolerances define the degree of failure. Mathematical models of self-discharge rate for both low and medium rate discharge were developed from laboratory measurements. Meter load profiles were also measured, which together with the self-discharge model enabled prediction of operational energy utilisation rates. A sample of 50 batteries was extracted from customers homes, after operating in the field for periods of up to 2 years, and the battery capacity loss rates were measured by the residual capacity method. A comparison of predicted and actual capacity utilisation rates yielded a discrepancy of approximately 1.28. Analysis implied that the source of the discrepancy could be adduced to an under estimate of the impact of self-discharge, but that a factor of up to six times the predicted value was required. Evidence was provided to show that self-discharge rate under operational stresses could be significantly higher than that under the steady-state laboratory measurement conditions, but that that the amplitude and time constant associated with a selfdischarge peak was unknown and not predictable. Mean service lives of 14 and 10 years for the respective battery types in the two types of meter are predicted, the worst case (probability of 0.13% of the population) being failure within approximately 5.9 years.