Battery Management using LSTM for Manhole Underground System

Abstract

The supply of electricity to the battery, which is connected to several sensors mounted in the manhole, is one of the problem in the Underground Management System. Data collection and prediction are critical for underground maintenance in order to avoid any faults. It is difficult to coordinate and handle a large volume of underground sensor data efficiently. This paper describes a prediction procedure for estimating the battery capacity evaluation in the underground management system. The system explained in this paper prevents faulty operation and sudden battery failure. Furthermore, it can help to reduce recovery time and repair costs. We propose a forecast of battery voltage for the next hour to improve the state of the sensor within the manhole. The developed procedure is implemented using a deep learning algorithm known as long short term memory. The implementation collected data for a one-week duration by measuring the performance power of the battery voltage. The results show that the trained and validated model will provide higher quality predictive value.