Statistical modeling of SoH of an automotive battery for online indication

Abstract

Battery health is always a challenge for heavy mobile or any portable system, as it is one of the prime factors affecting the system reliability. An indication of remaining battery life facilitates timely action taken proactively in alarming conditions, reducing the risk of running into a disaster. State of health (SoH) of battery is defined as remaining life of the battery to support load in question. Normally, the SoH of a battery is modeled as Ampere- hour (Ah) capacity with the age of the battery where terminal voltage is the only observable parameter employed in the model. This dose not provides an overall behavior model and not at all advisable to rely on in case of critical applications. In the present paper we propose a statistical modeling technique which uses stepwise regression on discharge profile of automotive battery to model SoH of the battery as a function of ageing and on- time (run- time) consumption. Ageing can be inferred from battery parameters such as Internal Resistance (IR), Terminal Voltage (OCV) and Specific Gravity (SG). These parameters are not independent and have problem of autocorrelation. An attempt has been made to develop a model free from the problem of autocorrelation and which provides a clue to the relative importance of each of the parameters in modeling the behavior of dependent variables. The objective of the study is to develop an online device to indicate the SoH of an automotive battery with the use of optimum observation needs.Run-time consumption can be seen from the discharged ampere-hour (A-h) capacity and ageing can be inferred from various slopes of battery parameters such as internal resistance, terminal voltage and specific gravity. The regression model when employed on discharge profile of automobile battery indicates correlation between these battery parameters (variables). The paper covers stepwise regression, effect of interaction between dependent variables on the model and autocorrelation effect on the model of SoH of a battery. This work is an attempt to design & develop an online device to indicate SoH of battery which can be put on the panel of an automobile to enable to actuate a preventive action.