On a Novel Hybrid Manta Ray Foraging Optimizer and Its Application on Parameters Estimation of Lithium-Ion Battery

Rizk Masoud Rizk-Allah, Mohamed I. Zineldin, Abd Allah A. Mousa, Sayed Abdel-Khalek, Mohamed Said Mohamed, Václav SnáselInt. J. Comput. Intell. Syst 2022 · DOI

TL;DR

This paper introduces MRFO-PSO, a novel hybrid optimization algorithm that combines the Manta Ray Foraging Optimizer (MRFO) with Particle Swarm Optimization (PSO) to improve both exploration and exploitation capabilities. The MRFO-PSO algorithm demonstrates superior performance in estimating parameters for lithium-ion battery models, offering a more accurate and efficient approach for battery management and prognostics.

What Problem Does This Paper Address?

Accurate modeling of lithium-ion batteries is crucial for applications ranging from electric vehicles to grid-scale energy storage. These models are essential for predicting battery state-of-charge (SOC) and remaining useful life (RUL), which are vital for safety, performance, and efficient operation. However, accurately estimating the numerous parameters within these complex battery models, such as Tremblay's model, is a challenging optimization problem. Traditional methods often struggle with non-linearities, discontinuities, and the risk of getting trapped in local optima. Existing meta-heuristic algorithms, while powerful, can sometimes exhibit a trade-off between exploration (searching broadly) and exploitation (refining solutions). This paper aims to address this by developing a hybrid optimization algorithm that balances these two aspects more effectively, leading to more precise parameter estimation for lithium-ion batteries.

Methods

The core of this paper is the development of the MRFO-PSO algorithm. It's a hybrid approach that merges two well-established meta-heuristics:

  • Manta Ray Foraging Optimizer (MRFO): Inspired by the foraging behavior of manta rays, MRFO uses three phases: chain foraging, cyclone foraging, and somersault foraging. It aims to find optimal solutions by mimicking how manta rays search for plankton.
  • Particle Swarm Optimization (PSO): A population-based stochastic optimization technique that simulates the social behavior of bird flocking or fish schooling. PSO uses velocity and position updates based on individual best-known positions and the global best-known position.

The novelty lies in how these two are combined. The MRFO-PSO algorithm incorporates the velocity concept from PSO into the cyclone foraging phase of MRFO. This integration helps guide the MRFO's search process by considering the velocity of particles, aiming to improve the balance between exploration and exploitation. Specifically, the velocity update in MRFO-PSO uses the first and second-best solutions found so far, enhancing the algorithm's ability to converge to optimal solutions. The paper also details the mathematical formulations for each foraging strategy and the PSO integration.

Datasets

The MRFO-PSO algorithm was evaluated using two main types of datasets:

  1. 23 Benchmark Functions: These are standard, well-known mathematical functions (unimodal and multimodal) used to test the general performance and robustness of optimization algorithms. The paper lists these functions and their properties (dimensions, range, minimum values).
  2. Lithium-Ion Battery Data: The algorithm was applied to estimate parameters for Tremblay's model using data from three different commercial lithium-ion batteries:
  3. Samsung Cylindrical ICR18650-22
    • Tenergy 30209 prismatic cell
    • Ultralife UBBL03 (type LI-7)

The battery data used for parameter extraction was obtained from datasheets and potentially from publicly available datasets (e.g., from Yong Wang and Lin Li, as referenced).

Key Insights and Ideas

  • Synergy for Better Balance: The core insight is that combining MRFO's foraging strategies with PSO's velocity-driven exploration significantly improves the balance between exploring new search spaces and exploiting promising regions. This hybrid approach overcomes limitations of using either algorithm alone.
  • Enhanced Convergence: The MRFO-PSO algorithm demonstrates a faster convergence rate towards optimal solutions compared to its constituent algorithms and other state-of-the-art methods.
  • Robustness Across Diverse Problems: The algorithm's effectiveness is validated across a wide range of benchmark functions, showcasing its ability to handle various optimization challenges, including unimodal and multimodal landscapes.
  • Practical Application in Battery Modeling: The successful application to Tremblay's model for three different Li-ion batteries highlights the algorithm's practical utility in a critical engineering domain. Accurate parameter estimation is key to reliable battery management systems.
  • Overcoming Human-Based Tuning: The paper emphasizes that using MRFO-PSO for parameter tuning overcomes the disadvantages of manual, human-based tuning, which can be subjective and time-consuming.

Key Results

The MRFO-PSO algorithm achieved highly competitive results across the 23 benchmark functions. It secured the optimal solution for many functions and demonstrated superior performance in terms of mean, median, maximum, and standard deviation for several others.

In the lithium-ion battery parameter estimation task:

  • Case I (Samsung ICR18650-22): MRFO-PSO achieved the minimum RSS (Residual Sum of Squares), indicating the best fit to the experimental data. It also performed well in mean, maximum, and standard deviation.
  • Case II (Tenergy 30209): MRFO-PSO provided very competitive results regarding RSS.
  • Case III (Ultralife UBBL03): MRFO-PSO achieved the minimum RSS, outperforming most compared algorithms.

Statistical tests (Wilcoxon signed-rank test) confirmed that the MRFO-PSO's superior performance was statistically significant across most comparisons, affirming its robustness and effectiveness. The discharge curves generated using the parameters estimated by MRFO-PSO showed excellent agreement with the datasheet curves for all three battery cases.

Limitations and Future Work

Limitations:

  • The paper notes that MRFO, in its original form, "lacks for memory to keep the best information of the previous trials." The MRFO-PSO hybrid aims to address this, but the extent to which memory is fully leveraged across all phases could be a point for further investigation.
  • While MRFO-PSO showed strong results, the Wilcoxon test indicated that its performance was not significantly different from MGEHO for one specific task, suggesting that other advanced variants can also be highly competitive.

Future Work:

  • The authors suggest extending the model to account for the volume of the battery pack for applications like electric vehicles, which often use large battery packs.
  • Investigating the future failure time of the battery pack is also proposed.
  • The paper implies that MRFO-PSO has potential for more complex search spaces, suggesting future work could explore its application in even more challenging optimization problems.

Summary updated

BibTeX
@article{rizkallah2022novel,
  author = {Rizk Masoud Rizk-Allah and Mohamed I. Zineldin and Abd Allah A. Mousa and Sayed Abdel-Khalek and Mohamed Said Mohamed and Václav Snásel},
  title = {{On a Novel Hybrid Manta Ray Foraging Optimizer and Its Application on Parameters Estimation of Lithium-Ion Battery}},
  journal = {Int. J. Comput. Intell. Syst.},
  year = {2022},
  doi = {10.1007/s44196-022-00114-4},
}
Figure