A novel Drosophila‐back propagation method for the lithium‐ion battery state of charge estimation adaptive to complex working conditions

Abstract

Accurate state of charge (SOC) for the lithium-ion battery is not only related to user experience but also the top target to avoid overcharge and overdischarge and to use it safely. The back propagation (BP) neural network is widely used in SOC estimation, but there exist some issues, such as easily falling local extreme value, converging slowly, or even unable to converge and even overfitting. The Drosophila algorithm has a simple algorithm and strong global optimization ability, but there is also a problem of direct inheritance to reduce the optimization ability. To solve these problems, an individual migration dynamic step Drosophila (Improved Drosophila) algorithm combined with the BP neural network is proposed to estimate the SOC of lithium-ion batteries and improve estimation accuracy. In addition, the performance of the proposed method is compared with that of its traditional algorithms and other commonly used functions. The experiments are carried out to verify the ternary lithium-ion battery under DST and BBDST conditions., the mean absolute error is less than 0.8%, and the root mean square error is less than 1.4%. The SOC estimation is carried out when the current data under the DST condition are missing, which also has good estimation performance, which shows the robustness of the algorithm. Compared to other algorithms, there is good estimation accuracy.

An Improved Self-Adaptive Flower Pollination-Backpropagation Neural Network for the State of Charge Estimation of Lithium-Ion Batteries with Output Sliding Average Window Strategy

“…21 Data-driven methods for SOC estimation usually make use of neural networks, 22 support vector machines, deep learning, and other data-driven techniques. 23,24 They train models to directly map the relationship between various parameters like voltage, current, and temperature measured during battery operation to the SOC, 25 without the need for mathematical battery models. Data-driven SOC estimation methods typically offer high accuracy and strong adaptability, 26 making them a preferred choice among researchers in the field of lithium-ion battery SOC estimation.…”

“…However, BPNN have slow convergence speeds, are sensitive to initial weights and thresholds, and require improvements in accuracy. 25,30 To address these limitations, three categories of enhancement methods have emerged: heuristic algorithms, 31 intelligent optimization algorithms, 32 and changing activation functions. 33,34 Heuristic algorithms are iterative methods inspired by nature, simulating various biological behaviors and physical laws to achieve optimization.…”

Comparative investigation of the usability of different machine learning algorithms in the analysis of battery thermal performances of electric vehicles

“…The dataset employed in this article was gained from the paper presented by Kalkan et al 25 The author sincerely appreciations them for their valuable contributions to EV and BTMS.…”

“…Xu et al 25 studied a relocation active phase Drosophila procedure integrated with neural networks in predicting the correct state of charge for an L‐iB. The efficiency of the obtained archetypal is associated with the commonly used models.…”

State of charge estimation for lithium‐ion battery pack based on real vehicle data and optimized backpropagation method by adaptive cross mutation sparrow search algorithm

“…To address these issues, many researchers have proposed various improvement measures. Xu et al 21 proposed an improved Drosophila algorithm that combines BP neural networks with individual migration dynamic step sizes and tested it on complex dynamic stress test (DST) and Beijing bus dynamic stress test (BBDST). Gong et al 22 proposed a data‐driven SOC estimation method based on deep learning, which consists of a long short‐term memory neural network and a BP neural network.…”