ME Building

Indoor Measurements

Each room has been equipped with Aoetec Z-Wave Multisensors that monitor indoor temperature, presence, humidity, and lightning. All measurments are sent through a low-power protocol to a Sensor Node.

Weather Forecast & Record

Weather information is retrieved from an external website, called Wunderground, by means of a standard API. Obtained data comprise both recorded measurements as well as future predictions for both outside temperature and solar global irradiance.

Actuation

Electric heaters have been installed in each room to provide the required thermal heat. The heaters are rated at 1850 Watts summing up to a total power capacity of 9250 Watts. To be able to modulate their power consumption the heaters were customized using additional hardware that allows PWM at 4Hz. Moreover, the heaters are equipped with power sensors to monitor their power consumption wich is sent to a Sensor Node.

The Sensor Node collects all measurments and stores them into an online database. The communication is handled using YARP [2].

When an experiment is performed, the main controller, implemented in MATLAB, performes the following tasks:

1. Retrieve all the necessary information from the database,
2. Decide on the power consumption for the heaters,
3. Communicate the power reference though Ethernet.

A crucial step towards the optimal control of a building is represented by the capability of predicting its future behaviour as a reponse to both actuation commands as well as external perturbation, e.g., occupancy, solar radiation, outside temperature, etc.

For these reason, a significant effort in the early stage of this project was spent in order to obtain reliable models.

Methodology

Standard black-box identification techniques were considered [3]. To obtain meaningful data, the heater power consumption was modulated with a PRBS signal. All external perturbations were recorded over a one-week period.

Each room was identified separately since the thermal coupling between them is weak.

As a result of this phase we obtained 4 second-order Auto Regressive model with exogenous inputs (ARX).

Results

On the left, the validation of the obtained dynamical model for one of the room is reported. As it can be seen in the top plot, the predicted temperature profile closely follows the actual measured temperature.