Energy saving in smart homes based on consumer behaviour data
11 - Studentische Arbeit
Primary target group
Created while belonging to FHNW?
This paper discusses how energy can be saved in smart homes without lowering the comfort of the inhabitants, based on consumer behaviour data only. A recommender system was designed, that suggests actions for inhabitants without the necessity for installing additional devices, executing manual configuration or having any other interaction with the system. As a consequence of the devastating earthquake and the resulting nuclear disaster that struck Fukushima in March 2011, concerned members of the public and the government agreed on a major reconsideration of the energy policy. However, such a radical rethinking can only be achieved if private households increase their efforts to save energy. Nevertheless, most research approaches conducted in smart homes in the past years, dealt with convenience rather than with sustainability. The aim of this master thesis is to find a way to save energy without causing significant inconveniences for the consumer. Therefore, the following hypothesis was formulated: “It is possible to design a recommender system that can suggest actions in smart homes based on consumer behaviour, which will lower energy usage but not decrease comfort levels”. The approach followed in this paper, is to mine frequent (and/or periodic) patterns in the event data of the inhabitants electricity usages, recorded by a smart home automation system. These patterns are converted into association rules, prioritized and compared with the current behaviour of the inhabitants. If the system detects opportunities to save energy without decreasing the comfort level, it will send a recommendation to the residents. Because the most appropriate research design to prove this hypothesis is design science research, the project follows the methodology to design and implement a functional prototype of a recommender system. At the end of the project, the prototype is evaluated in smart homes under real conditions. The main findings of the project and the concluding field-test of the prototype were: The project succeeded in identifying possible actions, which can be recommended in smart homes to lower energy usage in smart homes. Investigations showed how patterns in the behaviour data of the inhabitants can be used to trigger these actions at the right moment, to not lower comfort levels for the inhabitants. A design has evolved for a recommender system that uses association rules and deterministic finite state machines. It was identified, that the confidence and the length of a pattern are significant measures to predict if a suggestion does lower comfort or not. Overall, it can be said that this master thesis could verify part of its statement: The prototype demonstrated that it is possible to suggest actions that lower energy usage, but do not decrease comfort levels, while using consumer behaviour data as single source. However, besides the useful recommendations, the system did still recommend actions that did not just lower energy usage, but also the comfort level of the inhabitants. The ratio of useful recommendations, which reached little over 11% during the final test of the prototype, must be increased before broader adaption of the system is possible. Nevertheless, the proof of concept provided by the prototype is the first important step for further research in this field.