A novel genetic weight-directed feed forward backpropagation neural network for state of charge estimation of lithium-ion batteries.

Hai, Nan; Wang, Shunli; Liu, Donglei; Fernandez, Carlos; Guerrero, Josep M.

Abstract

Precious estimation of state-of-charge has become a more important status to the lithium-ion batteries of electronic vehicles. Basically, a three-layer genetic algorithm based on feed forward backpropagation neural network model is established. Specifically, an adaptive genetic method that makes the 𝑝𝑐 and 𝑝𝑚 change and self-correct with the degree of adaptation F is proposed to improve the stability and accuracy. Then, the momentum volume Δ𝑤𝑗𝑖 1 (𝛿) and the inertial volume Δ𝑤𝑘𝑗 2 (μ) are introduced to the first and the second weight of the topology in weighting correction process of backpropagation to help reduce the convergence time and improve the matching of the system with the increase in data volume. Finally, a further performance comparison of variable algorithms based on the backpropagation neural network is made under different working conditions at variable temperatures with large data volumes to prove the effectiveness of the proposed methods. The experimental results showed that the maximum error reached 0.9%, 1.2% and 0.3% under BBDST at 35°C, 25°C and 0°C over 500000 data, similarly, it reached 0.18%, 0.1% and 0.69% under DST at 15°C, 25°C and 35°C over 200000 data.

Citation

HAI, N., WANG, S., LIU, D., FERNANDEZ, C. and GUERRERO, J.M. 2024. A novel genetic weight-directed feed forward backpropagation neural network for state of charge estimation of lithium-ion batteries. Journal of energy storage [online], (accepted).

• Journal Article Type: Article
• Acceptance Date: Apr 1, 2024
• Deposit Date: Apr 2, 2024
• Journal: Journal of energy storage
• Print ISSN: 2352-152X
• Electronic ISSN: 2352-1538
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Keywords: Lithium-ion batteries; State of charge; Weight-directed; Feed-forward backpropagation; Adaptive genetic algorithm
• Public URL: https://rgu-repository.worktribe.com/output/2293534