Anomaly detection

[robertmcleod2]

first version of anomaly detection where we look for averaged daily values of smart meter data that exceed a z-value of 2 compared to the rest of the sample, and pass these as anomalies to the chatbot to use, along with displaying a graph to the user of their consumption with the anomalous days highlighted.

Future enhancements:

• provide more info on the anomalous and normal smart meter data to the LLM such as the average values, the values on the anomaly day, particular times of day where there was an odd behaviour
• Trying to classify the type of anomaly occuring, such as if there has been a constant increase in energy consumption, or if it is only on some days or times, etc. And reporting that back to the LLM.
• potentially making the anomaly detection an agent tool, so that the LLM can perform more detailed anomaly detection based on customer input.