Abstract
The design of heating systems for dwellings using new technologies, or new versions of old technologies, requires the ability to predict the temperatures in a dwelling. The temperature behaviour can be modelled, typically by differential equations which incorporate thermal driving forces and the thermal inertia of a dwelling. The development and characterisation of these models is usually based on fitting data accumulated over sufficient time to capture the behaviour of the dwelling under different conditions (summer, winter, etc.). Model fitting relies on assumptions about the behaviour of the system.
Optimisation can be used to examine these assumptions and gain insight into this behaviour. This paper describes the application of a nature inspired algorithm, known as the Plant Propagation Algorithm, a variant of a Variable Neighbourhood Search algorithm, to the problem of modelling a dwelling heated by an air source heat pump. The algorithm is evaluated using different population evolution strategies and implemented using a simple parallel computing paradigm on a multi-core desktop system. The results are used to identify potential sources of missing data which could explain the observed behaviour of the dwelling.