AI pipeline uses LLMs for office energy anomaly detection and recommendations

Researchers have developed an agentic AI pipeline designed to improve energy anomaly detection in office buildings. This system combines deep time-series forecasting with variational anomaly detection and LLM-based reasoning to provide actionable maintenance recommendations. The pipeline uses a hybrid SSA and LSTM forecasting model, an LSTM VAE for anomaly flagging, and a LangChain framework with a Context Agent, Diagnosis Agent, and Report Agent. Evaluations show that dynamic retrieval methods significantly reduce context sources while maintaining performance, and a 7B-parameter model successfully handled all tested scenarios. AI

IMPACT This research demonstrates a novel approach to energy management using LLMs and agentic pipelines, potentially improving operational efficiency in commercial buildings.

RANK_REASON Academic paper detailing a novel AI system and its evaluation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• LangChain
• LLM
• LSTM
• VAE

• paper
• product
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →