Valve-regulated lead-acid (VRLA) battery, owning the huge market, plays an important role in all aspects of industries. A VRLA battery sometimes called sealed leadacid (SLA) or maintenance-free battery, however, the safety of VRLA has been a wide concern since it is prone to self-heating problems which generate extra cost or even cause accidents when the internal temperature (IT) of VRLA is out of range. To prevent potential hazards, effective internal VRLA temperature monitoring methods are in need of further management. In this paper, a narrowband (NB) Internet of thing (IoT) connected VRLA battery internal temperature prediction (VBITP) algorithm is developed to provide early warning of battery temperature. In VBITP, the internal temperature is estimated by ambient temperature (AT) and input current (IC) through a pre-trained prediction model. The measured temperature data will be sent to the backend server using NB-IoT. A kind of recurrent neural network, nonlinear autoregressive exogenous (NARX) is applied to find the potential relationship between the input AT, IC and the output IT and train this model. The experimental results show that VBITP could estimate the IT of VRLA battery with an error rate of 0.04.
Keywords VRLA battery, internal temperature prediction, NARX neural network, narrowband internet of thing