Temperature estimation method for supercapacitor cell and module with less sensors

Abstract

Temperature monitoring is of great significance for the safe operation and state estimation of supercapacitors in practical applications. As excessive temperature can even cause thermal runaway, it is quite necessary to monitor the temperature of each cell, especially in transportation area. However, there are two problems in the temperature distribution estimation. Firstly, the hot-spot temperature is usually distributed inside the cell and is hard to be measured by sensors. Secondly, applying sensors to each cell in a module requires a large number of sensors, which are difficult and expensive to install. In this paper, aiming at difficulty in cell internal temperature measurement, the closed-loop observer is established which measures the surface and ambient temperature online; aiming at difficulties to reduce sensors in the module, a Long Short-Term Memory (LSTM) network is established to measure a small number of cells to predict the temperature distribution of the whole module, thus a method of hot-spot temperature estimation of supercapacitor is proposed. The result from experiment and estimation shows that the RMSE and MAE of the H-∞ and NN joint filter is <0.4 °C in most situations, and the absolute error of LSTM for module temperature distribution is <0.8 °C, the MAE and RMSE is <0.5 °C.