Condition Monitoring of Lithium-ion Batteries Providing Grid Services

Research output: Book/ReportPh.D. thesis

32 Downloads (Pure)

Abstract

As the global community grapples with issues of climate change and energy security, sustainable energy solutions become an imperative need for modern society. While renewables like wind and solar promise sustainable solutions, the intermittency of power generation and the spatio-temporal mismatch between energy supply and demand creates challenges for energy system design and operation. To counteract these, the battery energy storage system (BESS) is pivotal in maintaining grid stability. Furthermore, battery applications have grown significantly in all sectors in the trend of energy transition, including electric vehicles (EVs), electric two and three-wheelers, and consumer electronics. The battery state estimation and performance prediction have been of great concern whenever the batteries are used.

This thesis offers comprehensive studies on grid-connected BESS applications, battery health modeling, and real-world battery monitoring and inspection. Despite recent advancements, batteries are under the challenge of fluctuating usage patterns, complex operating environments, and erratic health dynamics. Therefore, our research accentuates the usage patterns of batteries through usage-based assessments. By harnessing state-of-the-art data-driven methodologies and analyzing battery operation records from various sectors, we elevate battery health research to greater applicability and precision. Moving beyond traditional physics-based modeling, data-driven techniques offer insights into battery health development in various conditions, bridging laboratory experimental research with industrial applications.

Structured across five chapters, the work covers the introduction, BESS in power systems, data-driven battery health modeling, real-world battery health prognosis, and a conclusion. First, a comprehensive overview of BESS application and integration in the power system over the last ten years is given, emphasizing the usage pattern assessment. Insights from our BESS development in DTU are shared by showing extensive testing results, including 99.2% capacity efficiency and 5% to 8% capacity degradation. Secondly, three open-access battery datasets are investigated, and parametric and non-parametric regression methods are implemented, including linear regression, Gaussian process regression (GPR), long short-term memory (LSTM), etc. With one-year frequency measurement in the Nordic synchronous area, the cycle-calendar combined life simulation of the batteries used for frequency regulation service is carried out. Extracting the features from the early discharging period of the battery operation, full-cycle capacity estimation is achieved with 1% root mean squared error (RMSE) by 30% information, and the accuracy capacity prediction can be made for the subsequent 300 cycles. Enlarging the scope from one-point capacity degradation prediction, our model predicts the discharge capacity at all voltage levels with the RMSE of 0.008 Ah and the mean absolute percentage error (MAPE) of 0.59% for the prediction step size of 100 cycles.

Lastly, quantification approaches are carried out based on the field monitoring data from fleets of 176 battery modules with more than 12000 module-charging cycles and 160000 cell-charging cycles. Tackling the arbitrary real-life usage pattern, the capacity degradation is quantified by the selected SOC range, which is 60% to 80%. The end-of-life prediction accuracy reaches 9.79 cycles with the first 10-cycle information for early prediction by random forest algorithm, and the battery module classification achieves more than 97% accuracy in classifying the modules with thermal runaway risk.

The Ph.D. work distinguishes itself by analyzing vast battery applications and operation records from laboratory tests to practical usage, demonstrating plenty of data-driven algorithms for battery health prognosis, and yielding insights into realistic battery condition monitoring results. Challenges such as limited data access and complex battery usage are emphasized, together with the opportunities provided by emerging data analysis technologies, a burgeoning battery application landscape, and advanced machine learning techniques. Given the centrality of energy storage in human society endeavors, understanding battery usage and health is paramount. Such knowledge not only improves power system reliability, hardware lifespan, etc., but also impacts the second life and decommissioning of battery systems. All of these aspects are crucial for building a sustainable future.
Original languageEnglish
Place of PublicationKongens Lyngby
PublisherTechnical University of Denmark
Number of pages212
Publication statusPublished - 2023

Fingerprint

Dive into the research topics of 'Condition Monitoring of Lithium-ion Batteries Providing Grid Services'. Together they form a unique fingerprint.

Cite this