The values of the charge and discharge efficiency and power to energy ratios as given in [25] for lithium-ion batteries were used for all calculations reported here。 Since the system was taken to be not connected to the electricity grid, no inverter was included and it was assumed that the air-conditioner operated from the com- bined battery and PV DC output。 For all simulations PV panels were

orientated towards the west to maximise afternoon solar gain with a slope equal to the standard roof pitch of  22。5◦。

2。3。 Weather data

Simulations were run using the weather data provided by the NatHERS administrator [26] which consists of Typical Meteorologi- cal Year hourly values of ambient temperature, humidity, total and beam solar irradiance and wind speed for 69 climate zones across Australia。 This data is currently used in assessing the energy star ratings of new buildings in Australia。 Here the ‘A weighting’ ver- sion of the data was employed which is stated to be appropriate for detached low-rise buildings and which places a slightly higher weighting on capturing the typical mean solar  irradiance。

2。4。 Building model

To model the overall thermal response of the building a sim- plified representation consisting of a uniform temperature and humidity air node coupled to transfer function models of exter- nal walls, ceiling and roof cavity has been implemented based on the equations described in [27] §4。7。3。 This includes an approximate model of long-wave radiation transfer between surfaces within the zone based on their relative areas as well as accounting for radia- tion absorption and convection on external surfaces, convection at internal surfaces and heat conduction and transmission of radiation through windows。

The approach described in [27] was modified to include long- wave radiation transfer from the external surface of the roof by calculating an approximate sol-air temperature Tsaaccording   to;

Fig。 3。  Building internal loads over a day。

3。 Building model validation

The building model described above was validated by compar- ing the predicted temperature and humidity in an unconditioned building over a period of 7 days with the values output by the soft- ware package TRNSYS using the detailed Type56 building model [28]。 The Type 56 model has been previously validated both with other software packages and with measured data from real-world building modelling experiments  [29]。

The base ‘house’ used for validation consisted of an uninsulated brick veneer construction with dimensions 5 × 5 × 2。4m, interior thermal capacitance of 1080 kJ/K, one North facing unshaded sin- gle glazed window with 2。5m2 area, and an uninsulated ceiling with pitched tile roof over a ventilated sub-space。 Natural infiltration in the zone was set at 2 air-changes per hour。 Radiant, sensible and latent  internal  loads  designed  to  replicate  a  common occupancy

pattern resulting in once-daily morning and evening load peaks were included as shown in Fig。 3。 These heat loads were based   on a peak three person occupancy in the space combined with oper- ation of several small electrical appliances and lighting。 External absorptivity of all surfaces was set at 0。5 and internal reflectance of walls and ceiling was set at  0。5。

For the Type 56 model, the emissivity of the external surface of the walls (but not the roof) was set to zero since long-wave emission from the vertical walls is dependent on the surrounding structures but generally has only a small influence on the results。 Internal surface heat transfer coefficients and wind dependent external con- vective heat transfer coefficients were specified the same as those used in our model。