In the current energy transition, electric vehicles (EVs) have emerged as a significant player, driving the need for advanced technologies that ensure optimal performance and sustainability. One of the critical requirements for widespread EV adoption is an accurate State of Charge (SoC) estimation algorithm, essential for maintaining battery health and maximizing efficiency. A recent breakthrough study presented by researchers from the Pandit Deendayal Energy University introduces the use of principal components-based feature generation and optimized Artificial Neural Networks (ANN) for SoC estimation in LiFePO4. This advanced method can solve the shortcomings of the existing SoC estimation methods that hinder their robustness, accuracy, or pose challenges in real-time implementation.
Effective and reliable data acquisition is a prerequisite for algorithm. The research team develops a custom-designed 12V, 4Ah battery pack equipped with a dedicated hardware setup for real-time data collection. Using a computerized battery analyzer, the team monitores the battery's voltage, current, and open-circuit voltage, while the DHT22 temperature sensor connected to a Raspberry Pi tracked temperature changes. Principal components are derived for the collected battery data set and analyzed for feature engineering. Three principal components were generated asinput parameters for the developed ANN.
To enhance the model's training efficiency, early stopping was implemented, and the researchers tested eleven combinations of ten different optimizers to minimize the loss function. The Adafactor optimizer, with its specific settings, demonstrated superior performance, achieving a remarkable Root Mean Square Error (RMSE) of 0.4083 and an R² Score of 0.9998.
Accurate SoC estimation can facilitate better integration of EVs into smart grids and support vehicle-to-grid (V2G) applications, where EVs can act as energy storage units to provide power back to the grid during peak demand periods.
Beyond automotive applications, accurate SoC estimation can also benefit energy storage systems in renewable energy grids, ensuring more reliable and efficient power distribution. As the world moves towards sustainable energy solutions, the importance of accurate SoC estimation will continue to grow, driving innovation in battery management and energy storage technologies.
In conclusion, this innovative approach represents a significant step forward in the evolution of smart charging strategies and efficient energy management for electric vehicles, providing valuable insights for both EV operators and electricity companies in their infrastructure investments and policy decisions. Future SoC estimation techniques should be capable of real-time adaptation to changes in driving conditions and energy consumption patterns. This adaptability will allow for dynamic energy management, ensuring that EVs operate efficiently under all conditions.
Reference
[1] Abdolrasol MGM, Hussain SMS, Ustun TS, Sarker MR, Hannan MA, Mohamed R, et al. Artificial neural networks based optimization techniques: a review. Electronics (Switzerland) Nov. 01 2021;10(21). https://doi.org/10.3390/electronics10212689. MDPI.
[2] Banguero E, Correcher A, Perez-Navarro A, García E, Aristizabal A. Diagnosis of a battery energy storage system based on principal component analysis. Renew Energy Feb. 2020;146:2438–49. https://doi.org/10.1016/j.renene.2019.08.064.
[3] Lee PY, Kwon S, Kang D, Cho I, Kim J. Principle component analysis-based optimized feature extraction merged with nonlinear regression model for improved state-of-health prediction. J Energy Storage Apr. 2022;48. https://doi.org/10.1016/j.est.2022.104026.
Author: Haifei Chaitali Mehta, Amit V. Sant, Paawan Sharma
Title of original paper: Optimized ANN for LiFePO4 battery charge estimation using principal components based feature generation
Article link: https://doi.org/10.1016/j.geits.2024.100175
Journal: Green Energy and Intelligent Transportation
https://www.sciencedirect.com/science/article/pii/S2773153724000276
