Tool for Utilities to predict consumption patterns of its customers and cluster them based on their consumption trends without the need for centralized data analytics.

Understanding electricity consumption habits of customers provides valuable data for power grid operators and electric utilities for distribution planning, demand response, and market segmentation to enhance the reliability and efficiency of distribution networks.
Due to limitations in standardization & interoperability, scalability, memory consumption, and time complexity of hardware deployment, most studies consider offline customer data to characterize customer electricity consumption habits.

Researchers at the University of Utah have developed a technology that provides the grid operator with customer characterization data in a fully distributed manner that has been proven to be efficient in memory and time consumptions and is highly scalable when deployed in hardware.
It is a communication and clustering platform that uses electricity demand curves to cluster customers based on consumption habits. Feature construction reduces the dimensionality of the demand curve down to four dimensions, which is reasonable number of dimensions for clustering. Distributed clustering is an algorithm based on a k-means variant and uses a threshold and step size for clustering convergence. The algorithm takes in the global demand curve and a customer demand curve and performs feature construction using the curves. Customers within the network connect and shares data within that neighborhood. Then all individuals, asynchronously, contribute to a global ledger and cluster the entire dataset.

Proof of Concept

• Distributed Asynchronous Clustering
• Data Interoperability
• Highly Scalable

Dean Gallagher
(801) 585-0396
dean.gallagher@utah.edu