Selbstlernendes Empfehlungssystem zur Steigerung der Behaglichkeit

Within the project „Empower Citizens” technical as well as social and health aspects are combined in order to increase the subjective comfort inside a flat with respect to room temperature, CO2 concentration, humidity, and so forth. Therefore, a low-cost system advisory system is developed which reads the aforementioned data from sensors situated in the different rooms of a flat. Using these measurements advices are given in order to increase the resident’s comfort. The advices are based on the computation of the so-called predicted mean value (PMV) which aims at capturing the comfort with respect to different parameters, e.g., age, body weight, and height. To achieve this goal, first, a co-simulation using Matlab and IDA-ICE was conducted utilizing a detailed model of the investigated flat. The model is split into two different zones, living room and bedroom, where a zone is defined through the presence of dedicated sensors inside the zone. The sensor data are transmitted to a central station, which reads also outside temperature and humidity, and form the basis for the PMV computation. If the PMV exceeds pre-defined limits, an advice is given, either to open a window or the shadow a window. Additionally, an advice is output if the CO2 concentration is too high. Furthermore, the advisory system is capable of learning how well the given advices are followed by the resident. Thus, the total number and the times at which advices are output can be flexibly adjusted by the system itself to optimally suit the resident’s fondness for following the advices. To evaluate the developed advisory system a parameter study was conducted evaluating different reference cases with respect to the parameters affecting the PMV calculation, e.g., age and bodyweight. Additionally, different motivations to follow the given advices were modeled within these reference cases. The simulation results prove that following the advices given by the advisory system leads to increased comfort for the residents while keeping the increase in energy demand for heating occurring from more ventilation to a minimum. Additionally, the study shows that ventilation only in the morning results in high CO2-concentrations heavily influencing the resident’s comfort. urthermore, this study presents a first easy-to-install hardware prototype comprised of a RaspberryPi 3B+ and an ARDUINO MKR1000 where the developed algorithm for the advisory system can be deployed. This prototype includes also the required sensors to monitor room temperature and CO2 concentration. In order to output the computed advices, it is equipped with an LCD display acting as human machine interface.