Recommendation System Towards Residential Energy Saving Based on Anomaly Detection

Abstract

This paper presents a recommender system to promote energy consumption reduction behaviors in residential buildings. The system exploits data stream processing methods jointly with machine learning algorithms on real-time residential data. Specifically, the data stream includes disaggregated power consumption, context, and weather conditions data. Internally the system converts time-series data streams into discrete ordered data points, which serve as inputs for training ML models. This method is used to predict power consumption anomalies. Gradually, the system helps to shape its users' activities into more energy-efficient ones. The experimental evaluation on real and simulated datasets demonstrates the promising performance of the proposed method, primarily when the K-neighbors neighbors' algorithm is used to classify the features extracted with interleaving current with the previous data points. The performance assessment of the machine learning algorithms shows the suitability of our implementation for Edge and Fog platforms in terms of accuracy, latency, and model size.