ABSTRACT
Several municipalities have recently installed wireless ’smart’ water meters that allow functionalities such as demand response, leak alerts, identification of characteristic demand patterns, and detailed consumption analysis. The meter data needs to be error-free to achieve these benefits, which is not necessarily available in practice due to ’dirtiness’ or ’uncertainty’ of data, which is mostly unavoidable.
This paper investigates solutions to mine uncertain data for reliable results and evaluates the impact of dirty data on data analysis results. The evaluation results can be used for informed decision-making on water planning strategies. A systematic study of the errors existing in large-scale smart water meter deployments is performed in this paper and helps to understand the nature of errors.
Identifying customers contributing to a consumption peak is used as the primary filter in this study. Its outputs are combined with the domain expert knowledge to evaluate their accuracy, validity, and potential errors. Each error is analyzed, and its source is investigated. This procedure is applied progressively to ensure that all detectable errors are discovered and characterized in the data model. The proposed approach is evaluated using the smart water meter consumption data obtained from Abbotsford, British Columbia, Canada. The results present the sensitivity of the selected filter to the errors are illustrated.
