IBPSA2005

Past research of predictive optimal control of active and passive building thermal storage inventory has confirmed the importance of accuracy in the employed building model. In a subsequent investigation of modelfree learning control for the same application, a hybrid modelbased/ modelfree control scheme based on simulated reinforcement learning has been proposed.

This paper reports on the measured performance of a residential solar water heater system over a period of 22 years and the modeling of the system to simulate its degradation over that period. The system consists of three fixed flat-plate collectors with a total of 5 m2 of double-layer glass cover plates and black aluminum fin-tube absorber plates. The solar storage tank capacity is 303 liters, which is used as a preheater to a 114-liter conventional electric water heater.

We present a model for calculating the sky radiation values considering the sky radiance distribution for a simulated building. We use the sky luminance distribution model of the CIE standard general sky rather than the measured sky radiance distribution. In this model, different sky types of the CIE standard general sky are identified from values of horizontal global sky radiation Eeg, and normal direct solar radiation Ees without reference to the measured sky radiance distribution. 

A finite differences numerical model for buried pipe systems is presented, accounting for sensible as well as for latent heat exchanges, so as for fully three dimensional heat diffusion in soil and flexible border conditions. After description of the algorithm, extensive validation against an analytical solution as well as against several long-term monitored real scale installations will be discussed. 

Natural night ventilation is an energy efficient way to improve thermal summer comfort. Coupled thermal and ventilation simulation tools predict the performances. Nevertheless, the reliability of the simulation results with regard to the assumptions in the input, is still unclear. Uncertainty analysis is chosen to determine the uncertainty on the predicted performances of natural night ventilation. Sensitivity analysis defines the most important input parameters causing this uncertainty.

This paper presents first steps of a methodology for calibration of building simulation models through definition of the parameters that most affect the main electric end-uses of a building. The first step consists on a good definition of constant loads (plug loads, lighting and occupation) and its schedules. The next steps are directed to calibrate the envelope variables. Sensitivity analysis is applied over the estimated cooling and heating loads in order to specify more accurate values for those inputs that present great impact on the total thermal load.

While most of the existing artificial neural networks (ANN) models for building energy prediction are static in nature, this paper evaluates the performance of adaptive ANN models that are capable of adapting themselves to unexpected pattern changes in the incoming data, and therefore can be used for the realtime on-line building energy prediction. Two adaptive ANN models are proposed and tested:accumulative training and sliding window training. The computational experiments presented in the paper use both simulated (synthetic) data and measured data.

Many existing building energy performance assessment frameworks, quantifying and categorising buildings post occupancy, offer limited feedback on design decisions.

This paper presents a numerical model for the analysis of impact of building envelope porosity on energy. In the porous envelope, the infiltrating air entering a building can change in temperature, along the infiltration path due to heat exchange between the air itself and the porous insulation matrix; hence the envelope effectively behaves as a heat exchanger. The presented model is based on combined airflow and heat transfer through porous media. Microscopic energy equations are formulated for solid and gas phases separately.

Behavioural models derived from on-going field studies can provide the basis for predicting personal action taken to adjust lighting levels or remedy direct glare in response to physical conditions. SHOCC, a sub-hourly occupancy-based control model, provides building energy simulation programs, such as ESP-r, access to advanced behavioural models, such as the Lightswitch2002 algorithms intended for manual and automated lighting systems.