Battery autonomy estimation method applied to lead-acid batteries in uninterruptible power supplies

Uninterruptible power supplies (UPSs) provide energy to critical loads in the event of a power outage. During this event, batteries supply the load for a determinate amount of time, which can vary from a few minutes to several hours, depending on the load and battery capacity. The knowledge of this time period is useful to anticipate a system shutdown and to turn on backup generators, if present. With this in mind, this work contributes with a method for predicting battery autonomy time during the operation of UPSs in backup mode. The proposed method is implemented as an algorithm that predicts autonomy. A Thevenin equivalent circuit with variable resistance is used to model the battery. The circuit is employed in learning and predicting battery terminal voltage behavior during discharge. To achieve this, successive curve fittings are made using a Levenberg-Marquardt algorithm. In this way, the predictions of the behavior of terminal voltage and remaining battery autonomy are continuously adjusted. Due to the used strategy, the method has high adaptability, working even with variations in temperature, load steps, and battery aging. The method is developed based on a commercial 1 kW Single-Office/Home Office UPS, that uses a pair of series-connected 12 V / 7 Ah lead-acid batteries. The compatibility with different battery banks is demonstrated, as well as the possibility of operation without measuring temperature and current at the batteries, two advantages in economic terms. Experimental results demonstrate the predictions of autonomy over time.