Non-Technical Losses Detection Tool

A tool for detecting anomalous consumption based on the characterization of each customer’s historic behavior
Nube de etiquetas

Processing in two phases

“Training phase” (individual customer characterization:
historic load curve clustering, shape factors
calculation,…)

“Evaluation phase” (each new load curve is evaluated
against the historic behavior pattern of the customer –
known clusters-, and classified accordingly or marked
as “anomalous”)

Employed techniques

  • Meta-heuristic algorithms

  • Clustering techniques

  • Outliers and pattern detection

  • Multidimensional scaling