• Login
    View Item 
    •   KCA University Repository Home
    • Theses and Dissertations
    • School of Business & Public Management
    • View Item
    •   KCA University Repository Home
    • Theses and Dissertations
    • School of Business & Public Management
    • View Item
    JavaScript is disabled for your browser. Some features of this site may not work without it.

    A Long Short-term Memory (LSTM) Network Model For Predicting Water Consumption In Residential Properties Using Smart Water Meter Data

    Thumbnail
    View/Open
    Full text (2.318Mb)
    Downloads: 122
    Date
    2023
    Author
    Kamau, Stanely M.
    Metadata
    Show full item record
    Abstract
    Rapid urbanization in Kenya and the subsequent population increase have caused a severe imbalance between water demand and water availability. This imbalance poses serious challenges in managing water consumption in urban areas. Furthermore, water leakages and variable human activity generate non-linear patterns in domestic water consumption data which make traditional linear time series models such as autoregressive integrated moving average (ARIMA) ineffective. Using a case study research design with Nairobi City, the author developed a novel Long Short-Term Memory (LSTM) network model for predicting water demand through deep learning of smart water meters data. The model uses high frequency non-linear time series data collected between January and December 2022 from smart sensors within an Internet of Things (IoT) framework, alongside other information such as timestamp and temperature. Nine different variables were constructed from the study data and used to train and validate the LSTM network model for smart water meter data management. The model was then evaluated using root mean square error (RMSE) and the correlation coefficient. Although significant variation was observed in the daily and monthly patterns of domestic water consumption, the model outcomes were relatively accurate. LSTM generated values that mirrored observed values more closely than the ARIMA model. Evaluation metrics also indicated that LSTM had lower prediction errors. It is expected that the developed model will be generalizable for estimating future water consumption in other urban households in Kenya and other regions. The study is limited by a small sample dataset of 320 households and the lack of socio economic and demographic factors to determine water consumption. A more extensive study with multiple influencing factors is recommended to assist water authorities and service providers to properly distribute water, identify leakages, and take corrective actions to prevent degradation of the ecological environment.
    URI
    https://repository.kcau.ac.ke/handle/123456789/1532
    Collections
    • School of Business & Public Management [630]

    Copyright © 2020  | KCA University Library | Off-Campus Access |
    Send Feedback
     

    Browse

    All of KCA University RepositoryCommunities & CollectionsBy Issue DateAuthorsTitlesSubjectsThis CollectionBy Issue DateAuthorsTitlesSubjects

    My Account

    LoginRegister

    Statistics

    View Usage Statistics

    Copyright © 2020  | KCA University Library | Off-Campus Access |
    Send Feedback