Residential home temperature prediction models that use space conditioning experiments and disparate sources of information

  • Nils Christian Bausch

    Student thesis: Doctoral Thesis


    A model to predict air temperatures inside a residential home was created, which used local and remote environmental sensor data. Space conditioning experiments were carried out in the residential home and models were created to describe the observed temperature data. Results from laboratory space conditioning experiments were used to create a new experimental prediction model.
    Static inputs of the experimental model were replaced with dynamic inputs and an improved model was created, capable of general prediction application in the residential home. A novel system for space conditioning control was designed, which applied the improved model for air temperature prediction.
    Initial prediction models showed prediction error margins of ±1°C. Residential home and laboratory space conditioning experiments were utilized to create non-linear temperature prediction models capable of application outside of the experimental scope. Improved prediction models were based on the experimental models and showed average error margins of ±0:12°C compared to observed temperature data. A novel system design was proposed, combining the improved model with a traditional heating control to create a new optimal start-stop heating application for residential homes.
    Date of Award2014
    Original languageEnglish
    Awarding Institution
    • University of Portsmouth
    SupervisorGiles Tewkesbury (Supervisor), David Sanders (Supervisor) & Shalini Ramlall (Supervisor)


    • smart environment
    • smart home
    • prediction
    • temperature modelling
    • domestic

    Cite this