Partager

Approche d’imputation de données basée sur l’algorithme des k plus proches voisins différé pour les données de systèmes de bâtiments à flux multiples. 

Lagged-kNN based data imputation approach for multi-stream building systems data.

Numéro : 3182

Auteurs : PRADHAN O., HÄLLENBERG D., CHEN Z., WEN J., WU T., CANDAN K. S., O'NEILL Z.

Résumé

Increasing advancements in building digitization, smart sensing, and metering technologies have allowed large amount of data to be collected and saved for monitoring, analyzing, and controlling building systems. However, due to sensors or communications failure, the data collected are often incomplete and poor in quality. Data imputation approaches to replace the missing values, specifically based on both statistical and predictive models have been widely adopted for multivariate datasets. It is hence of interest to find an effective way to impute building system data by leveraging the mutual information from strongly correlated sensors. In this paper, we evaluate multiple data imputation approaches using data collected from a medium sized, mixed-use institution building situated in Stockholm, Sweden. Sensors with widely varying characteristics from the case study building were selected to test the imputation methods. Artificial test data with ground-truth was first created for validation by removing randomly selected portions of data. The imputation accuracy was computed for each method and the impact of the chosen method on a short-term building forecasting model was evaluated. Results demonstrate that incorporating time-lagged cross correlations within the knearest neighbor (kNN) model provide the most accurate imputations.

Documents disponibles

Format PDF

Pages : 10 p.

Disponible

Gratuit

Télécharger

Détails

  • Titre original : Lagged-kNN based data imputation approach for multi-stream building systems data.
  • Identifiant de la fiche : 30030212
  • Langues : Anglais
  • Source : 2022 Purdue Conferences. 7th International High Performance Buildings Conference at Purdue.
  • Date d'édition : 2022
  • Document disponible en consultation à la bibliothèque du siège de l'IIF uniquement.

Liens


Voir d'autres communications du même compte rendu (39)
Voir le compte rendu de la conférence

Indexation