Proposta de um Modelo de Previsão do Estado de Carga de Baterias Chumbo-Ácido

Auxiliary power systems, such as stationary battery banks, are required at random times of main circuit failure, to ensure that the system is operating at such times is costly and represents a challenge, especially in sealed batteries, without access to elements that would aid in predicting their state of charge and state of health [1, 2]. This paper presents a study proposal of a model that is capable to predict the state of charge of lead-acid batteries, from values extracted of a quick impedance spectroscopy (IE) measurement. The data were collected in an experiment that featured ten EI measurements in a battery which charge states were from 100% to 10%, the measurements were performed by the impedance spectroscopy (EIS) technique. Regarding the linearity test, the results showed that the 50 mV value for the voltage sinusoidal signal perturbation was the most suitable for use in measurements, unlike the other two procedures, this step can be discarded in other measurements, provided that the option is to use the 10 mV value, recommended in the literature. The characteristic curves, resulting of the measurements, were adjusted with the support of an electrical equivalent circuit, as it was proposed by Randles in 1947. The data were linearized and a correlation table was structured for the definition of independent variables that best explain the relationship between the state of charge and the change of internal resistance of the battery. Statistical tools such as linear regression and hypothesis testing were used to prepare the forecasting model that consists of two independent variables. One related to pure resistance, mainly due the formation of lead sulfate on the plates during the discharge process, and another due the variation of the double layer capacitance. The load state prediction model, the main objective of this work, demonstrated that with only two of the five Randles parameters it is able to explain 99.8% of the load variable. Thus confirming that it is possible to determine the charge state of lead acid batteries by a spectroscopy impedance measurement. The values related to the performance statistics presented values very close to the upper and lower limits, thus attesting to the predictive capacity of the model. The results were satisfactory with the proposed objectives achieved and can be used as a basis for future works, as the relation between the forecast state of charge and state of health of the battery, development of cheaper measurement impedance equipment and model implementation which can compensate temperature variations in the measurement environment.