Research On Electricity Consuming Activity Recognition And Feedback In Intelligent Family Environment

In home environment, the recognition and feedback of the user's electricity behavior on the power demand side is an important means to improve the environmental awareness and energy saving of the end user. It has aroused extensive research from domestic and foreign scholars. The existing intelligent environment has some shortcomings in the recognition of the behavior in the family. The selection of the intrusive equipment is difficult for residents to accept. The recognition method with non-intrusive equipment will cause that recognition accuracy is not high enough.In order to solve these problems, this paper adopted a more reasonable data acquisition method, established the ontology model on the user behavior, and uses an improved Hidden Markov Model as the behavior recognition method. The main contents of this paper are as follows:(1)User acceptance for the sampling method of present activity recognition system is low, the activity recognition study mainly adopts the existing data set, it does not take into account the actual situation of data errors and loss, so we designed an intelligent socket to get load data for behavior identification, proposed a load data verify algorithm and a LBboosting algorithm which fused two curve fitting methods to repair the data. The algorithms used Kernel Extreme Learning Machine to learn the characteristic of home circuit loss, used this characteristic to verify and repair data.The experimental results show that the LBboosting algorithm is better than the B-spline algorithm. Data missing has a significant impact on the recognition accuracy. Data quality assurance improves the accuracy of activity recognition.(2) Aiming at the problem that the recognition accuracy is not high enough, we added indoor seamless positioning module, and the ontology-based domestic user's activity model is established. Using a domain knowledge included in this model, a probability matrix is defined. HMM2K, which is based on knowledge and data hybrid driven, is proposed. Compared with the classical recognition method NB and HMM, the experimental results show that HMM2K shortens the training test time and the recognition accuracy is better than other methods.(3)Aiming at the problem that the current energy feedback has no clear guidance on the improvement of behavior, too subjective design ideas and fuzzy evaluation method. A activity-oriented interactive feedback model called EnergyAction is proposed. This model used focus + context (F + C) technology to highlight the focus object to help users know how to improve their behavior, and used the OZ paradigm(Wizard of OZ) to design and test our energy feedback prototype. We proposed a new evaluation standard called behavior improvement value (BIV), which compared to the traditional method rely solely on energy saving rate to evaluate is more effective.