Hubble Space Telescope battery voltage-current modeling

Is it possible to predict long-term trends of battery capacity based on a few orbits of spacecraft data? Our approach to answering this question has been to study the trends in Ni-H₂ battery properties as derived from in-orbit measurements. The behavior of batteries is driven by complex chemical and physical phenomena. The complexity of electrochemical processes prevents the use of fundamental equations of physics and chemistry in practical situations. Therefore, we have developed semi-empirical models with the ability to simulate the battery properties and to measure the degree of degradation without the need to explain electrochemical processes in battery cells. The limitation of these models is that their predictive capability depends on a set of known charge and discharge conditions. Mathematically, we have used the linear and nonlinear mappings and time-delayed nonlinear mappings between the integrated raw current and the voltage data with noise. We studied local models of the voltage and current relationship in terms of their predictive power and the sensitivity with respect to other measured data (e.g. temperature and cell pressure). We define a local model as one based on a few spacecraft orbits. We have studied the information content of other measurements such as temperature and pressure in functional local models based on the current, charge, voltage and their short term history