WCSE

Immersive Smart Meter Data Analytics: Leveraging eXtended Reality with LSTM and LLMs

Abstract— The rapid advancement of smart grid technologies has led to an exponential growth in smart meter data, creating new opportunities for more accurate energy consumption forecasting and immersive data visualization. This study proposes an integrated framework that combines eXtended Reality (XR), Long Short-Term Memory (LSTM) networks, and Large Language Models (LLMs) to enhance smart meter data analytics. The process begins with the application of LSTM to capture temporal dependencies in historical electricity usage data. Subsequently, the Large Language Models (LLMs) are employed to refine these textual forecasts, offering better predictions and explanations that are easily understandable by end-users. Finally, the enriched insights are presented through an XR environment, enabling users to interact with smart meter analytics in an immersive and intuitive way. By visualizing data trends, predictions, and explanatory narratives in a spatial computing interface, users can explore complex information more effectively. This multi-modal approach facilitates better decision-making for energy management, promotes user engagement, and supports smart city initiatives aiming for sustainable energy consumption. The integration of XR, LSTM, and LLMs technologies demonstrates a promising direction for future research and practical applications in smart energy systems.

Index Terms— Long Short-Term Memory (LSTM), Large Language Models (LLMs), Extended Reality (XR), Smart Meters

Ru-Guan Wang, Chu Hsien Tsai, Mei Cheng Tseng, Rong-Cian Hong, HuaRong Syu, Chien-Cheng Chou
National Central University, Taiwan

[Download]

Cite: Ru-Guan Wang, Chu Hsien Tsai, Mei Cheng Tseng, Rong-Cian Hong, HuaRong Syu, Chien-Cheng Chou, "Immersive Smart Meter Data Analytics: Leveraging eXtended Reality with LSTM and LLMs", 2025 the 15th International Workshop on Computer Science and Engineering (WCSE 2025), pp. 32-36, Jeju Island, South Korea, June 28-30, 2025.