Identification and Error Analysis of Lithium-Ion Battery Oriented to Cloud Data Application Scenario

Fang Zhang, Tao Sun, Bowen Xu, Yuejiu Zheng, Xin Lai, Long ZhouBatteries 2023 · DOI

TL;DR

This paper tackles the challenge of accurately estimating lithium-ion battery capacity from real-world vehicle data, which is notoriously noisy and lacks labels. They propose a novel approach combining an improved two-point method with a Fuzzy Kalman filter to handle systematic and random errors in cloud-based data, successfully demonstrating its application in a new energy big data platform.

What Problem Does This Paper Address?

Estimating the capacity and predicting the lifespan of lithium-ion batteries is crucial for electric vehicles (EVs). However, real-world data collected from vehicles (cloud data) presents significant challenges compared to controlled laboratory settings. This data is "label-less" (lacks explicit ground truth for capacity), suffers from unpredictable driving conditions, varying sampling rates and resolutions, and potential data loss. These factors introduce systematic and random errors that hinder accurate modeling and capacity identification. Existing methods often struggle with this noisy, real-world data, making it difficult to calibrate battery models for vehicles in service and leading to unreliable capacity estimations.

Methods

The authors propose a multi-faceted approach to address the challenges of cloud data for battery capacity identification:

  • Improved Two-Point Method: They start with a foundation in the commonly used two-point method for capacity calculation. However, they adapt it for engineering applications by considering partial charging curves and deriving a formula where capacity estimation relies on the accuracy of both charge/discharge current integration and SOC estimation.
  • Error Analysis: A significant portion of the paper is dedicated to analyzing the sources of error in real-world vehicle data. This includes:
  • Systematic Errors: Primarily from sampling frequency and resolution. They discuss how high-frequency data can be managed and how low SOC resolution can lead to stacked metering points, affecting accuracy.
  • Random Errors: Including issues like current/time jumps, temperature fluctuations, and the impact of manufacturer-defined SOC correction strategies.
  • SOC Correction Strategies: They analyze and discuss various SOC correction strategies, particularly focusing on the high linearity of the OCV-SOC curve at low SOC and the challenges of accurate correction at high SOC.
  • Temperature Correction: A formula is provided to correct capacity estimations to a standard temperature (25°C) to account for temperature-induced variations in battery performance.
  • Fuzzy Logic Kalman Filter (FL+KF): This is the core of their proposed solution for robust capacity estimation. They integrate a Kalman filter with fuzzy logic rules to adaptively adjust the observation noise based on the charging/discharging state (specifically, SOCmin and SOCmax). This allows the filter to better handle the uncertainties and noise in the cloud data, leading to a more stable and accurate capacity estimation curve.

Datasets

The study utilizes real-world cloud data collected from 100 electric vehicles. This data was gathered over a period of two months. The paper does not mention the creation of a new benchmark dataset but rather uses existing operational data from vehicles.

Key Insights and Ideas

  • Real-world data is fundamentally different and more challenging: The paper strongly emphasizes that laboratory data is not directly transferable to real-world EV applications due to inherent noise and lack of labels.
  • Systematic and random errors are pervasive: A detailed breakdown of error sources highlights the complexity of working with vehicle data, moving beyond just simple noise.
  • SOC estimation accuracy is paramount: The accuracy of the capacity estimation is directly tied to the accuracy of the SOC estimation, underscoring the importance of robust SOC correction strategies.
  • Fuzzy logic provides adaptive noise handling: The integration of fuzzy logic into the Kalman filter is a key innovation, allowing the system to dynamically adjust its sensitivity to noisy measurements based on the battery's operating state. This is crucial for handling the varying reliability of data points.
  • Charging and discharging segments offer complementary information: The paper demonstrates that analyzing both charging and discharging data segments, and then combining their results, leads to a more comprehensive and reliable capacity estimation.

Key Results

  • The proposed Fuzzy Kalman filter-based capacity identification algorithm was successfully applied to cloud data from 100 real vehicles.
  • The algorithm demonstrated a consistent decay trend in battery capacity over mileage, aligning with expected lithium-ion battery aging patterns.
  • The identified capacity curve showed good agreement between charging and discharging segment estimations.
  • The developed algorithm was integrated into a new energy big data monitoring cloud platform, signifying a successful transition from academic research to engineering application.
  • The estimated mileage life for the battery system was projected to reach 200,000 km, with a potential battery life of about 9 years if users drive 60 km per day.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that some vehicles may exhibit premature capacity aging due to multi-layer consistency and potential faults, which might require further investigation and return inspection.
  • While the algorithm is validated, the exact quantitative accuracy metrics (e.g., percentage error) are not explicitly stated for the final integrated system.
  • Future Work:
    • The authors plan to further develop the research by predicting capacity based on current estimations and precisely forecasting the battery service period. This suggests a move towards more proactive battery health management.

Summary updated

Abstract

The label-less characteristics of real vehicle data make engineering modeling and capacity identification of lithium-ion batteries face great challenges. Different from ideal laboratory data, the raw data collected from vehicle driving cycles have a great adverse impact on effective modeling and capacity identification of lithium-ion batteries due to the randomness and unpredictability of vehicle driving conditions, sampling frequency, sampling resolution, data loss, and other factors. Therefore, data cleaning and optimization is processed and the capacity of a battery pack is identified subsequently in combination with the improved two-point method. The current available capacity is obtained by a Fuzzy Kalman filter optimization capacity estimation curve, making use of the charging and discharging data segments. This algorithm is integrated into a new energy big data cloud platform. The results show that the identification algorithm of capacity is applied successfully from academic to engineering fields by charge and discharge mutual verification, and that life expectancy meets the engineering requirements.
BibTeX
@inproceedings{zhang2023identification,
  author = {Fang Zhang and Tao Sun and Bowen Xu and Yuejiu Zheng and Xin Lai and Long Zhou},
  title = {{Identification and Error Analysis of Lithium-Ion Battery Oriented to Cloud Data Application Scenario}},
  booktitle = {Batteries},
  year = {2023},
  doi = {10.3390/batteries9040216},
}
Figure