Following decades of stability and comfortable margins, utility companies today face
strong pressure from regulatory bodies and competitors. As a response to the market
dynamics, many have initiated a transformation from a “provider” to a service
company, yet realize that their customer insights that would be necessary to
successfully develop and market new services are sparse. We argue that the required
information is contained in consumption data that is available to utility companies. We
demonstrate how data analytics and machine learning make sense out of such data and
add value to organizations. Using datasets containing annual electricity consumption
information of private households, we apply and test in field experiments a Support
Vector Machines algorithm that predicts probabilities of individual costumers to sign up
on an energy efficiency portal. We show that signup rates can be doubled and argue
that classification tools provide customer insights at low cost and at scale.
The aim of this paper is the identification of a superordinate research framework for describing emerging ITinfrastructures within manufacturing, logistics and Supply Chain Management. This is in line with the thoughts and concepts of the Internet of Things (IoT), as well as with accompanying developments, namely the Internet of Services (IoS), Mobile Computing (MC), Big Data Analytics (BD) and Digital Social Networks (DSN). Furthermore, Cyber-Physical Systems (CPS) and their enabling technologies as a fundamental component of all these research streams receive particular attention. Besides of the development of an eponymous research framework, relevant applications against the background of the technological trends as well as potential areas of interest for future research, both raised from the economic practice’s perspective, are identified.
Purpose - The emerging technology of Cyber-Physical Systems (CPS) promises new problem solutions in manufacturing, logistics and SCM. In order to enable a structured scientific discussion, a general definition of such systems is needed.
Design/methodology/approach - To deliver such a definition, scientific articles on the relatively young research field of CPS have been identified and analyzed. Methods like literature review and matrix analysis have been applied during research work.
Findings - On an abstract level, seven basic functions of CPS have been identified. Besides these functions, several phenomenological characteristics emerged as well. Combining the functional and the phenomenological view, a generally applicable definition has been constructed. Some first application examples from logistics, manufacturing and general Supply Chain Management illustrate the definition.
Research limitations/implications - The paper aggregates findings from different research streams. They are application and technology agnostic. It would, in a next step, thus be necessary to identify the specific technologies behind those functions, to describe the state of the art from an application perspective, and to develop a framework for future research.
Practical implications - Considering the great potential of CPS within an operational context, this paper aggregates definitional aspects from a consistent scientific perspective. The processing of the literature on hand addresses a tangible demand for economic practice, which seems to be overwhelmed by the sheer amount of different terms and definitions.
Original/value - The presented study provides strong insights into the emerging subject of CPS gearing towards a prospective research-focus on operational aspects.
Introduction: The knowledge about household properties (such as number of inhabitants, living area, heating type, etc.) is highly desirable for utility companies to pave the way to targeted energy efficiency programs, products and services. Raising individual household data via surveys or purchasing it is expensive and time consuming, and often only a small fraction of customers participate.
Recently, data mining methods have been developed to automatically infer house-hold characteristics from smart meter consumption data. However, the slow smart metering rollout hampers practical implementation of these methods in many countries. In this work, we present a machine learning approach that reveals household properties from conventional annual electricity consumption data currently available at a large scale.