Clustering of customers based on AMS-data
Today, typical daily load profiles for electricity usage of both households and industry are based on average profiles made several decades ago. Now that all electricity customers in Norway have smart meters installed which meter the electricity consumption every hour, there is an opportunity to estimate more accurate how the electricity consumption of different customers vary during the day.
One way of finding typical daily profiles by using Big Data techniques is to use clustering techniques on metered electricity consumption from a large number of customers. There are several interesting questions to answer here:
- Which clustering algorithm performs best?
- How many typical customer clusters are there?
- Can we find characterisations for the customers in different clusters?
- Which other data sources can be used to improve the result?
By answering these questions planning of the distribution grid can be improved, and the grid operators can know more accurately how much electricity is transferred to different customers.