Development of a non-intrusive monitoring technique for appliance' identification in electricity energy management

Abstract

Electricity is one of the most popular forms of energy in modern society. In a smart house, the smart meter directly interacts with intelligent appliances to disaggregate the total electricity energy consumption. However, this results in high financial costs and low monitoring reliability due to the installation of massive meters for appliances. In this paper, a Non-Intrusive Appliance Load Monitoring (NIALM) system is proposed. The proposed NIALM system, which integrates feature extraction and feature optimization methods with pattern recognition techniques, is used to identify the operation status of individual appliances. Both energizing and de-energizing transient current waveforms of an unidentified appliance can be identified through analyses of the proposed system. Different potential transient features, extracted from an acquired transient current waveform, form a feature space to be selected. In pattern recognition, it is important to select the most effective features from the original feature set and to appropriately scale them by real-valued numbers simultaneously in order to improve the identification accuracy and construct a small and efficient identification system. In this paper, the proposed NIALM system employs the Genetic Algorithm (GA) with the Fisher criterion to search for the optimal sub-set of features. After the most influential features are selected, different identifiers, including k-Nearest-Neighbor Rule (k-NNR), Back-Propagation Artificial Neural Network (BP-ANN) and Learning Vector Quantization (LVQ) are used to perform load identification. The identification results confirm that the proposed system is able to identify the operation status of each appliance under different single-load and multiple-load operation scenarios at different real experimental environments. Besides, this paper comments that the k-NNR identifier is preferred for the development of NIALM, due to its simplicity.