Enhancing sustainability through data-driven anomaly detection : a use case for smart utility meters