Professional Documents
Culture Documents
A Method For State-Of-charge Estimation of Lithium-Ion Batteries Based On PSO-LSTM-2
A Method For State-Of-charge Estimation of Lithium-Ion Batteries Based On PSO-LSTM-2
A Method For State-Of-charge Estimation of Lithium-Ion Batteries Based On PSO-LSTM-2
Energy
journal homepage: www.elsevier.com/locate/energy
a r t i c l e i n f o a b s t r a c t
Article history: State-of-charge (SOC) estimation of lithium-ion battery is one of the core functions of battery man-
Received 17 April 2021 agement system. In order to improve the estimation accuracy of SOC, this paper proposes a long short-
Received in revised form term memory neural network based on particle swarm optimization (PSO-LSTM). Firstly, the key pa-
6 June 2021
rameters of LSTM are optimized by PSO algorithm, so that the data characteristics of lithium-ion battery
Accepted 11 June 2021
can match the network topology. In addition, random noise is added to the input layer of PSO-LSTM
Available online 15 June 2021
neural network to improve the anti-interference ability of the network. Finally, experiments show that
the proposed method can achieve accurate estimation under different conditions. The estimates based on
Keywords:
Lithium-ion battery
PSO-LSTM converge to the real state-of-charge within an error of 0.5%.
SOC estimation © 2021 Published by Elsevier Ltd.
Particle swarm optimization algorithm
Long short-term memory neural network
https://doi.org/10.1016/j.energy.2021.121236
0360-5442/© 2021 Published by Elsevier Ltd.
X. Ren, S. Liu, X. Yu et al. Energy 234 (2021) 121236
The LSTM network effectively solves the problems that tradi- Klaus greff et al. [26] explored the setting of LSTM related hyper-
tional neural network is easy to fall into local minimum, gradient parameters. The experimental results showed that the learning rate
vanishing and explosion. At the same time, it has strong modeling was the most critical hyper-parameter of LSTM, followed by the
and analysis ability for processing time series data [25]. As shown network size, while the momentum gradient had little effect on the
in Fig. 1, LSTM introduces a new memory unit (ct) to transmit final results. In order to match the LSTM structure with the data
2
X. Ren, S. Liu, X. Yu et al. Energy 234 (2021) 121236
Step4: Generating the estimated value yt. 5) Judge whether the termination conditions are met. If satisfied,
the algorithm ends and outputs the optimization result xbest¼
SOCðtÞ ¼ Wy ht þ by (7)
(ls, lr, ep, lrdf, lrdp); otherwise, return to 3) in the Step 2.
3
X. Ren, S. Liu, X. Yu et al. Energy 234 (2021) 121236
Fig. 5. Battery current and voltage data. (a) Battery current data collected under constant current and UDDS conditions at 25 C. (b) Battery voltage data collected under constant
current and UDDS conditions at 25 C.
Table 2 Table 3
Hyper-parameters of PSO-LSTM. SOC estimation errors under constant current condition.
constant current(without noise) 333 0.0140 622 0.0699 275 EKF 0.1310 1.1366 1.9610
constant current (with noise) 357 0.0111 640 0.3473 259 LSTM(without noise) 0.0190 0.1836 0.2441
UDDS(without noise) 238 0.0429 661 0.4713 118 PSO-LSTM(without noise) 0.0167 0.2228 0.2950
UDDS (with noise) 235 0.0323 509 0.1593 225 LSTM(with noise) 0.1497 0.6201 0.6990
PSO-LSTM(with noise) 0.0117 0.1608 0.2157
Fig. 6. SOC estimate results under constant current condition (without random noise): (a) SOC estimate results based on EKF, LSTM and PSO-LSTM. (b) Comparison of SOC estimate
error.
5
X. Ren, S. Liu, X. Yu et al. Energy 234 (2021) 121236
Fig. 7. SOC estimate results under constant current condition (add random noise): (a) SOC estimate results based on LSTM and PSO-LSTM. (b) Comparison of SOC estimate error.
Fig. 8. SOC estimate results under UDDS (without random noise): (a) SOC estimate results based on LSTM and PSO-LSTM. (b)Comparison of SOC estimate error.
Table 4 Table 5
SOC estimation errors under UDDS condition. Comparison results of SOC estimation.
Fig. 9. SOC estimate results under UDDS condition (add random noise): (a) SOC estimate results based on LSTM and PSO-LSTM. (b)Comparison of SOC estimate error.
6
X. Ren, S. Liu, X. Yu et al. Energy 234 (2021) 121236
Credit author statement ion battery state of charge with alternate adaptive extended kalman filter and
ampere-hour counting methods. Energies 2019;12.
[10] Snihir I, Rey W, Verbitskiy E, Belfadhel-Ayeb A, Phl Notten. Battery open-
Xiaoqing Ren: Conceptualization, Methodology, Software, circuit voltage estimation by a method of statistical analysis. J Power Sour-
Formal analysis, Data curation, Writing e original draft, Visualiza- ces 2005;159(2):1484e7.
tion. Shulin Liu: Validation, Investigation, Writing e review & [11] Jiang C, Wang S, Wu B, Fernandez C, Xiong X, Coffie-Ken J. A state-of-charge
estimation method of the power lithium-ion battery in complex conditions
editing, Supervision, Project administration, Funding acquisition. based on adaptive square root extended Kalman filter. Energy 2021:219.
Xiaodong Yu: Supervision, Project administration. Xia Dong: Data [12] Sun D, Yu X, Wang C, Zhang C, Bhagat R. State of charge estimation for
curation, Funding acquisition. lithium-ion battery based on an Intelligent Adaptive Extended Kalman Filter
with improved noise estimator. Energy 2020:214.
[13] Ye M, Guo H, Cao B. A model-based adaptive state of charge estimator for a
Declaration of competing interest lithium-ion battery using an improved adaptive particle filter. Appl Energy
2017;190:740e8.
[14] Ning B, Cao B, Wang B, Zou Z. Adaptive sliding mode observers for lithium-ion
The authors declare that they have no known competing battery state estimation based on parameters identified online. Energy
financial interests or personal relationships that could have 2018;153(JUN.15):732e42.
appeared to influence the work reported in this paper. [15] Kang LW, Zhao X, Ma J. A new neural network model for the state-of-charge
estimation in the battery degradation process. Appl Energy 2014;121(may
15):20e7.
Acknowledgments [16] Yang F, Li W, Li C, Li C, Miao Q. State-of-charge estimation of lithium-ion
batteries based on gated recurrent neural network. Energy 2019;175(MAY
15):66e75.
This work was supported by the Natural Science Foundation of
[17] Zhao X, Xuan D, Zhao K, Li Z. Elman neural network using ant colony opti-
Shandong Province [ZR2020QF064] and [ZR2020ME206]. mization algorithm for estimating of state of charge of lithium-ion battery.
The Journal of Energy Storage 2020;32:101789.
References [18] Shen S, Sadoughi M, Chen X, Hong M, Hu C. A deep learning method for online
capacity estimation of lithium-ion batteries. The Journal of Energy Storage
2019;25:100817.
[1] Sun D, Yu X, Wang C, Zhang C, Bhagat R. State of charge estimation for [19] Li W, Sengupta N, Dechent P, David Howey, Anuradha Annaswamy, Dirk
lithium-ion battery based on an Intelligent Adaptive Extended Kalman Filter Uwe Sauer. Online capacity estimation of lithium-ion batteries with deep long
with improved noise estimator. Energy 2020:214. short-term memory networks. J Power Sources 2021;482:228863.
[2] Seo M, Song Y, Kim J, Paek SW, Sang WK. Innovative lumped-battery model [20] Yang F, Zhang S, Li W, Miao Q. State-of-charge estimation of lithium-ion
for state of charge estimation of lithium-ion batteries under various ambient batteries using LSTM and UKF. Energy 2020:201.
temperatures. Energy 2021;(8):120301. [21] Fasahat M, Manthouri M. State of charge estimation of lithium-ion batteries
[3] Jiang C, Wang S, Wu B, Fernandez C, Coffie-Ken J. A state-of-charge estimation using hybrid autoencoder and Long Short Term Memory neural networks.
method of the power lithium-ion battery in complex conditions based on J Power Sources 2020;469:228375.
adaptive square root extended Kalman filter. Energy 2021:219. [22] Chen J, Feng X, Jiang L, Zhu Q. State of charge estimation of lithium-ion battery
[4] Liu S, Dong X, Zhang Y. A new state of charge estimation method for lithium- using denoising autoencoder and gated recurrent unit recurrent neural
ion battery based on the fractional order model. IEEE Access 2019;(99). 1-1. network. Energy 2021;227(9):120451.
[5] Liu S, Cui N, Zhang C. An adaptive square root unscented kalman filter [23] Bian C, He H, Yang S. Stacked bidirectional long short-term memory networks
approach for state of charge estimation of lithium-ion batteries. Energies for state-of-charge estimation of lithium-ion batteries. Energy 2020;191:
2017;10(9):1345. 116538.
[6] Xu W, Xu J, Yan X. Lithium-ion battery state of charge and parameters joint [24] Cui X, Goel V, Kingsbury B. Data augmentation for deep neural network
estimation using cubature Kalman filter and particle filter. Journal of Power acoustic modeling. In: 2015 IEEE international conference on acoustics,
Electronics 2020;20(1):292e307. speech and signal processing. ICASSP); 2014.
[7] Zheng L, Zhu J, Lu Dah Chuan, Wang G, He T. Incremental capacity analysis and [25] Hochreiter S, Schmidhuber J. Long short-term memory. Neural Comput
differential voltage analysis based state of charge and capacity estimation for 1997;9(8):1735e80.
lithium-ion batteries. Energy 2018;150(MAY1):759e69. [26] Greff K, Srivastava RK, Koutnik J, Steunebrink BR, Schmidhuber J. LSTM: a
[8] Xiong R, Cao J, Yu Q, He H, Sun F. Critical review on the battery state of charge search space odyssey. In: IEEE transactions on neural networks and learning
estimation methods for electric vehicles. IEEE Access 2017;(6):1832e43. systems. pubMedId; 2017. p. 27411231.
[9] Liu Z, Zhe L, Zhang J, Su L, Hao G. Accurate and efficient estimation of lithium-