Consensus-based clustering for indoor sensor deployment and indoor condition monitoring

Abstract

Accurate indoor climate monitoring is important for identifying energy-efficiency strategies in buildings. However, existing indoor monitoring approaches typically use networks of fixed/rigid sensor nodes that are often deployed in an ad-hoc manner, which could be limited in covering the areas/zones of interest or in accounting for the profile variations of the physical parameters over time. To address this gap, this paper proposes a consensus-based clustering method for identifying more robust sensor deployment strategies. The proposed method clusters a co-association matrix of input partitions that capture the periodic variations of the physical parameters from multiple building locations to increase the robustness of the sensing strategy to the daily and hourly changes in the indoor climate conditions. To test the proposed approach, thirty sensing units that sense four physical parameters were deployed in three rooms. Two consensus-clustering-based approaches (hourly- and daily-consensus) and four different frequencies for profile generation were tested, resulting in 48 scenarios for evaluation. The experimental results showed that the proposed approach could be more robust to changes in indoor climate conditions compared to the baseline, achieving a higher strategy performance index by up to 59%, and that the daily-consensus approach is often more robust than the hourly approach. The results also showed that the time frequency affects the strategy robustness and that parameters whose profiles have seasonal components are more likely to outperform. The proposed method could be used to determine reliable indoor sensor locations for capturing the dynamic environmental state of buildings, towards efficient building operation and management.