Energy Consumption Anomaly Detection Workflow

This n8n workflow is designed to detect anomalies in energy consumption data through an automated process involving data collection, analysis, and logging. Starting with an HTTP webhook trigger, it receives energy data submissions via POST requests. The data then flows through a text splitter node that divides the input into manageable chunks. These chunks are processed to generate text embeddings using Hugging Face’s models, which are then stored in a Supabase vector database for efficient similarity searches.

The workflow allows for querying the database to identify patterns or anomalies against stored energy data. An LLM (language model) interacts with the queried data to define and analyze anomalies, providing contextual insights. A chat node facilitates interactive analysis, supported by an in-memory buffer to maintain state. The results can be documented automatically into a Google Sheet for record-keeping and further analysis.

This automation is particularly useful for companies or facilities monitoring energy consumption for irregularities or inefficiencies, enabling proactive maintenance and cost savings. It combines natural language processing, vector similarity search, and data logging into a seamless, automated anomaly detection system.