Connect with us

Hi, what are you looking for?

AI Generative

Smart Buildings Achieve 86% Energy Savings Using AI and Large Language Models

Virginia Tech researchers develop an AI-driven Building Energy Management System achieving 86% accuracy in energy control, revolutionizing smart building efficiency.

Researchers at Virginia Polytechnic Institute and State University are pioneering a new approach to energy management in smart buildings by integrating artificial intelligence, specifically utilizing Large Language Models (LLMs). This innovative framework for a Building Energy Management System (BEMS) aims to enhance energy efficiency by responding intelligently to user requests and interpreting building data in real time. Tianzhi He and Farrokh Jazizadeh, along with their team, have developed a prototype that moves beyond mere automation, creating a closed-loop system that learns from energy data to improve building operations.

The research leverages advanced LLMs, including ChatGPT and GPT-4, to manage key building systems such as HVAC, lighting, and appliances. This allows occupants to interact with their environments using natural language, tailoring automated routines to individual preferences. The findings indicate a shift from traditional energy management practices, with the team employing metadata schemas like Brick and Bot to provide structured data about building components and relationships. Vector databases, such as Milvus and FAISS, are utilized for efficient data storage and retrieval, while Retrieval-Augmented Generation (RAG) enhances the accuracy of LLM reasoning by merging it with external knowledge sources.

In addition to real-time management, the research delves into energy forecasting and optimization, enabling time series analysis and consumption prediction through LLMs. However, privacy and security concerns remain paramount as these models access sensitive building data. The exploration of multi-agent systems, wherein several LLM-powered agents collaborate on complex tasks, is also underway. To assess the viability of this technology, the researchers developed benchmarks like ElecBench, which evaluate LLM performance in building applications.

The framework comprises three modules: perception, central control, and action, forming a cohesive feedback loop that captures, analyzes, and responds to energy data and user requests. This structure differentiates it from traditional dashboard interfaces by embedding autonomous data analytics capabilities within LLMs. The prototype was rigorously tested against 120 user queries across four real-world residential energy datasets, focusing on metrics such as latency, functionality, capability, accuracy, and cost-effectiveness.

Results indicate high performance, with the system achieving 86% accuracy in device control, 97% in memory-related tasks, and 77% in energy analysis. Scheduling and automation tasks yielded 74% accuracy, demonstrating the system’s proactive energy management capabilities. However, challenges remain in complex cost estimation tasks, which resulted in a lower accuracy of 49%. The research team has identified this as a critical area for further development.

Statistical analysis through ANOVA tests confirmed the framework’s adaptability across diverse residential energy profiles, showcasing its potential for generalization in varying settings. The authors emphasized that while the system excels in user interaction and energy management, there is a trade-off between response accuracy and computational efficiency, warranting further exploration. Future work is expected to focus on refining cost prediction precision and optimizing system performance, ultimately enhancing the efficacy of LLM-driven BEMS.

This breakthrough presents a significant advancement in the integration of AI technology in energy management, bridging the gap between human-centered design and operational efficiency in smart buildings. As the landscape of energy management continues to evolve, the potential for LLMs to revolutionize the way buildings operate and interact with occupants is becoming increasingly apparent.

👉 More information
🗞Context-aware LLM-based AI Agents for Human-centered Energy Management Systems in Smart Buildings
🧠 ArXiv: https://arxiv.org/abs/2512.25055

See also
Staff
Written By

The AiPressa Staff team brings you comprehensive coverage of the artificial intelligence industry, including breaking news, research developments, business trends, and policy updates. Our mission is to keep you informed about the rapidly evolving world of AI technology.

You May Also Like

AI Marketing

BusySeed unveils Rankxa, a tool tracking brand visibility across AI-generated responses, revealing 90% of brands lack meaningful presence in this new landscape.

AI Technology

A1 Public Relations helps entertainment brands enhance AI visibility in 2026 by integrating structured content and fresh, authoritative media, ensuring they are recognized by...

AI Finance

More than 55% of Americans now turn to AI tools for financial advice, risking personal data exposure despite rising privacy concerns.

Top Stories

Google DeepMind's AI co-clinician outperformed GPT-5.4 in doctor tests, achieving 67 preferences in primary care queries and a remarkable 95% quality score in open-ended...

Top Stories

Apple's CarPlay now supports third-party voice assistants like ChatGPT and Perplexity AI, with Perplexity outperforming ChatGPT in navigation and calendar management.

Top Stories

Omnea launches its MCP Server, the first platform integrating supplier data with tools like ChatGPT and Claude to streamline procurement workflows and enhance productivity.

AI Marketing

OGS Media reports a staggering 2,000% increase in AI visibility for clients by leveraging trust signals on Reddit, transforming digital marketing strategies.

Top Stories

Bitcoin rebounds 27% to $76,128, but remains below critical resistance of $82,228 as AI insights reveal mixed market sentiment ahead.

© 2025 AIPressa · Part of Buzzora Media · All rights reserved. This website provides general news and educational content for informational purposes only. While we strive for accuracy, we do not guarantee the completeness or reliability of the information presented. The content should not be considered professional advice of any kind. Readers are encouraged to verify facts and consult appropriate experts when needed. We are not responsible for any loss or inconvenience resulting from the use of information on this site. Some images used on this website are generated with artificial intelligence and are illustrative in nature. They may not accurately represent the products, people, or events described in the articles.