System identification of a domestic residence using Wireless sensor node data
Within the framework of full-scale experimental validation of the global building energy simulation software programme CLIM2000, an experimentation has been carried out in a 100 m2 real house from Oct 95 to May 96. First, we compare the simulated results with the experimental results. Then, we applied two different screening methods (sensitivity analysis) to the model in order to exhibit the most influent parameters and to calculate the output confidence interval (uncertainty analysis), and to compare the pertinence of each method in terms of results precision and calculation time. The experimental results are compared with the output uncertainty in order to see if they are included in the confidence interval. 1 INTRODUCTION The CLIM2000 software environment was developed by Electricity Applications in Residential and Commercial Buildings Branch in Research and Development Division of the French utility company EDF (Electricité De France). This software operational since June 1989, allows the behaviour of an entire building to be simulated. Its main objective is to produce economical studies, pertaining to energy balances over long periods as well as more detailed physical behaviour studies including stiff non-linear problems and varied dynamics. The building is described by means of a graphics editor in the form of a set of icons representing the models chosen by the user and taken from a library containing about 150 elementary models. Up to 1994, the experimental validation work on CLIM2000 was based on tests carried out in small cells where each parameter was controlled or was well-known. The objective is to validate the models in CLIM2000 with respect to physical phenomena taken one by one. The results have been acceptable and in some cases the improvement of some models has been achieved by using an error analysis method. Experimental validation using tests cells is invaluable for enhancing prediction power of models. However, these types of tests lack credibility among the wider audience, as they see as laboratory tests remote from reality. A new complementary way for EDF to treat the CLIM2000 validation is to address the end-users concerns, i.e. to treat the validation in a more global fashion with full-scale experimental building. To meet this, one house has been rented so that experiments could be carried out with particular attention paid to energy-related measurements made on the electric heating system. This house called Valeriane is new cottage situated at Lisses (30 km south of Paris) complying with the 1989 French thermal regulations, equipped with electric convectors. It is intended for global comparisons of energy consumption over a complete statutory heating season, namely from October 1 to May 21. The tests were carried out in Valeriane from October 1 1995 to May 21 1996. Each room is heated to obtain 19°C constant temperature. Last October 1995, in order to give credibility to our validation work, we have modeled Valeriane on CLIM2000  in blind way (no knowledge of experimental data and then no possible adjustment of the modelling). In 1996, the same modelling was used for the measurements/calculation comparison. In this paper, we will first describe the experimental house and the experimental sequence. We will present the comparison between simulated results and experimental results. We will present briefly the different method used for sensitivity analysis and uncertainty analysis, and the conclusions that can be drawn in our case. Then, the experimental data are compared with the output uncertainty in order to see if they are included in the confidence interval. 2 EXPERIMENT DESCRIPTION 2-a House description This house Valeriane is a new cottage complying with the 1989 French thermal regulations and is equipped with electric convectors. It is fully representative of all new dwelling units in France according to the survey carried out by SOFRES (floor area 100 m2) .