A new article in the magazine energies has studied various methods for estimating the state of charge of lithium-ion batteries, a key technology in applications such as electric vehicles. The research was conducted by scientists from Brazil.
To learn: Method for SoC estimation in lithium-ion batteries based on multiple linear regression and particle swarm optimization. Photo credit: cigdem/Shutterstock.com
The rise of electric and hybrid vehicles
The transport sector is a recognized contributor to anthropogenic climate change and air pollution. Internal combustion engines use non-renewable petrochemical-based fuels that are drivers of carbon emissions. To address the environmental and health issues associated with internal combustion engines, companies have developed electric and hybrid vehicles over the last few decades.
The market share of all-electric and hybrid vehicles has increased year on year, with several automakers now offering options for customers. A report by Hazra and Reddy estimates that electric vehicles will have a 7% market share by 2030. Governments have introduced policies to significantly reduce and eventually phase out new internal combustion engines over the coming decades.
Batteries are a central element of electric and hybrid vehicles, with lithium-ion batteries being the main types used by manufacturers. These battery variants have a long service life with 80% discharge over three thousand cycles, fast charging, high energy density and high voltage compared to other commercial battery types used in the automotive industry.
However, there are some critical disadvantages of electric vehicles compared to traditional ICE technologies. Challenges related to charging time, autonomy, charging infrastructure and low rates of improvement of battery technologies hinder the full commercialization of these alternative transportation options. The urgent need to commit the world to zero carbon emissions requires these electric transport challenges to be overcome quickly.
Battery SoC model based on the voltage and current information. Photo credits: Castanho, D et al., Energies
State of charge estimation
Improving the performance of lithium-ion batteries for EV applications is central to the commercialization of these alternative modes of transportation. Energy management control strategies aim to improve energy distribution in EV systems between energy storage components and traction motors. Studies have shown that power source degradation affects these strategies.
Efficient power management strategies help mitigate critical performance issues and extend battery life. Among the various strategies that have been developed in recent years, state of charge estimates are fundamental to the power management control performance. Real-time monitoring of the battery charge level is essential. Creating an accurate state of charge estimate is critical to the efficiency of control systems.
Several methods for estimating battery state of charge have been developed in recent years, with several computational models and algorithms being studied in research. Data-driven deep learning estimation methods, adaptive filter algorithms, analytical battery models including the Shepherd model, particle swarm algorithms, and backpropagation neural networks have all shown promise as state of charge estimation techniques.
initial capacity of the battery. (a) Battery SoC model. (b) Current and voltage profiles for the initial capacity. Photo credits: Castanho, D et al., Energies
The new study in energies has studied and compared different methods for estimating the state of charge of lithium-ion batteries that can be used in the electric vehicle field. Four different cable profiles were tested under different temperature conditions.
The multiple linear regression model both with and without spline interpolation has been used in research along with the generalized linear model. Three different optimization techniques (Differential Evolution, Particle Swarm Optimization and Genetic Algorithm) were used to calibrate the models. The free coefficients of the models were calculated using these bio-inspired metaheuristics.
Conclusions of the study
The aim of the study was to evaluate the performance of each model for state of charge estimation in order to find a low computational cost and a simple computational strategy with high power and accuracy.
The calculation results of the training and test set showed that a multiple linear regression with particle swarm optimization offers the best performance for state of charge estimation. This approach provides the best estimator for all case studies conducted in the research.
Remarkably, this optimized model outperforms current methods reported in the literature. The model requires relatively little computational effort, making it a competitive candidate for commercial applications such as electric vehicles.
Initial capacity curve for PSO training. Photo credits: Castanho, D et al., Energies
The paper made some recommendations for future research in this area. The authors have highlighted the need to conduct tests on different battery models and under different conditions. In addition, different C-rates should be tested. One approach that will support this effort is the development of new databases.
The research also found that the current literature contains over a hundred different algorithms than those evaluated in the research, providing significant opportunities for future researchers to evaluate their suitability for state of charge estimations. Further investigations will expand the knowledge base on this central research question.
Castanho, D. et al. (2022) Method for SoC estimation in lithium-ion batteries based on multiple linear regression and particle swarm optimization energies 15(19) 6881 [online] mdpi.com. Available at: https://www.mdpi.com/1996-1073/15/19/6881