A novel M-1 structured Bidirectional Long Short Term Memory-Rauch Tung Striebel Smoothing algorithm for the joint estimation State of Charge and multi-constrained sustained peak power of lithium-ion batteries

Abstract

The estimation of State of Charge (SoC) and State of Power (SoP) of lithium-ion batteries is an important part of battery management system. For the purpose of obtaining high-precision SoC and SoP estimation results, this paper proposes a novel M-1 structured Bidirectional Long Short Term Memory-Rauch Tung Striebel Smoothing (BiLSTM-RTSS) algorithm, which can effectively fit current, voltage, temperature and SoC by using the improved sliding window M-1 technique. The RTSS algorithm is applied to update the SoC estimation of the BiLSTM to facilitate the accuracy and speed of SoC estimation. Subsequently, to better estimate the SoP with duration of 30 s, 2 min, and 4 min, three key constraints including SoC estimation results, second-order RC model and battery limits, are simultaneously used in SoP estimation. According to experiment, the mean absolute error of SoC is 0.602, 0.619 and 0.548, 0.469 respectively, under the conditions of HPPC and BBDST. When the duration is 30s, the peak charging and discharging currents up to 170 A and 380 A. In addition, the estimation trends of sustained peak power and peak current are in the same direction. The experimental results verify the rapidity and reliability of the proposed algorithm for SoC and SoP.