Classification of customers based on temporal load profile patterns
[EN] The deployment of Advanced Metering Infrastructure (AMI) is providing to utilities large amounts of energy consumption data from their customers, in form of daily load profiles with energy consumed per hour or a smaller period. These data can yield valuable results when analyzed, in order to ex...
| Autores: | , , , |
|---|---|
| Tipo de recurso: | artículo |
| Fecha de publicación: | 2017 |
| País: | España |
| Institución: | Universitat Politècnica de València (UPV) |
| Repositorio: | RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia |
| Idioma: | inglés |
| OAI Identifier: | oai:riunet.upv.es:10251/104883 |
| Acceso en línea: | https://riunet.upv.es/handle/10251/104883 |
| Access Level: | acceso abierto |
| Palabra clave: | Load profiles Dynamic clustering Pattern recognition Classification INGENIERIA DE SISTEMAS Y AUTOMATICA INGENIERIA ELECTRICA |
| Sumario: | [EN] The deployment of Advanced Metering Infrastructure (AMI) is providing to utilities large amounts of energy consumption data from their customers, in form of daily load profiles with energy consumed per hour or a smaller period. These data can yield valuable results when analyzed, in order to extract useful knowledge about the typical patterns of consumption of energy from the customers. The proper mechanisms and tools have to be developed and implemented for this objective. Big Data and Big Data Analytics systems will contribute to analyze this information and help to extract knowledge from the data, summarized in form of patterns or other mining knowledge, that will aid experts in decision support. In the present work a classification of customers based on their temporal load profiles is proposed. This classification procedure could be implemented in the current Big Data Analytics software systems, providing an added value to their statistical analysis options. Previous works in the literature present algorithms that allow to classify load profiles from customers by processing batch datasets and obtaining static patterns of load profiles. The proposed technique allows to analyze patterns not only in shape but also in their evolution or trend of energy consumption at each hour of the day through time. Specific quantitative indicators that characterize the patterns (and the consumers associated to them) are described and tested for this purpose. |
|---|