Digital twin battery storage in energy storage system

Zhao, Kai 2024. Digital twin battery storage in energy storage system. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF (Thesis) - Accepted Post-Print Version Download (4MB) | Preview
	PDF (Cardiff University Electronic Publication Form) - Supplemental Material Restricted to Repository staff only Download (2MB)

Abstract

Battery Energy Storage Systems (BESS) play a key role in supporting the transition to renewable energy by providing stability to energy grids. However, the increasing complexity of managing BESS presents significant challenges regarding real-time monitoring, accurate state estimation, and predictive maintenance. Estimating key battery states, such as State of Charge (SOC), State of Health (SOH), and Remaining Useful Life (RUL), is important for enabling the operational efficiency and longevity of these systems. Traditional methods often struggle to account for the complex and dynamic behaviour of battery systems, leading to inefficiencies in decision-making and system performance. This thesis proposes a Digital Twin (DT)-driven approach to enhance decision support for BESS, focusing on improving the accuracy of battery state estimations and optimising system operations through the integration of real-time data with advanced analytical models. The thesis begins by outlining the development of a DT framework tailored specifically for BESS. The proposed framework creates a digital model of the physical system, enabling continuous synchronisation of real-time data between the physical and digital environments. This integration allows for real-time updates of battery states, providing operators with a comprehensive view of system performance. The framework includes detailed data acquisition and preprocessing procedures, which are essential for keeping the accuracy of the DT model. Additionally, advanced deep learning algorithms are applied to enhance the framework’s capacity for decision support. This approach provides a robust foundation for improving operational decision-making by offering insights into potential outcomes based on various operational scenarios. Secondly, this thesis presents a detailed examination of battery state estimation iv Abstract methods, with a focus on advanced deep learning techniques. Temporal Convolutional Networks (TCN) and Long Short-Term Memory (LSTM) networks are used to estimate SOC and SOH, and predict RUL. These models are capable of processing both historical and real-time data, allowing them to adapt to dynamic changes in the operational environment. Compared to traditional methods, the TCN-LSTM model demonstrates improved accuracy in estimating battery states, which is critical for proactive maintenance and efficient resource allocation. The results of the experimental analysis validate the effectiveness of these models, highlighting their ability to provide reliable predictions that support the management of BESS. Thirdly, the thesis addresses the importance of situational awareness in managing BESS operations. Situational awareness is essential for managing multiple operational objectives, including load balancing, energy dispatch, and system reliability. A multi-faceted optimisation strategy is proposed, leveraging real-time data from the DT to address these objectives simultaneously. This approach can provide operators with a detailed understanding of system conditions and the ability to simulate various operational scenarios. The optimisation approach improves system efficiency under varying conditions, allowing for more informed decisions that reduce the risk of unexpected system failures. Finally, the thesis introduces a DT-supported decision support system designed to optimise BESS maintenance and operational efficiency. The proposed method extends DT to support operational decision-making by incorporating real-time health monitoring, fault detection, and predictive maintenance strategies. The decision support system presented leverages predicted RUL. These predicted RUL to inform a maintenance scheduling and spare parts ordering policy, aimed at minimising system downtime and reducing operational costs. Additionally, large language models (LLMs), are introduced to enhance the system’s capability for intelligent fault diagnosis and maintenance recommendations through the analysis of unstructured data, such as maintenance logs and technical documentation. Abstract v This thesis presents a comprehensive DT-driven approach for enhancing the management and operation of BESS. The proposed framework integrates data with advanced deep learning models, providing more accurate estimations of battery states and supporting more effective decision-making. The findings of this research contribute to the broader field of energy management by demonstrating the potential of DT to improve the reliability and efficiency of BESS operations. As renewable energy continues to play an increasingly important role in global energy systems, the adoption of DT will be critical in supporting the long-term sustainability and performance of energy storage infrastructures.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Schools > Engineering
Uncontrolled Keywords:	Digital Twin; Battery Energy Storage System; Situational Awareness; Deep Learning; Decision Support; SOC estimation, SOH estimation and RUL prediction
Date of First Compliant Deposit:	2 May 2025
Last Modified:	06 May 2025 09:26
URI:	https://orca.cardiff.ac.uk/id/eprint/178013

Actions (repository staff only)

Edit Item