Non-intrusive sensing based multi-model collaborative load identification in cyber-physical energy systems

Cyber-physical energy systems (CPES) integrate information sensing and communication networks to the physical entities for energy efficiency applications in smart grid. Non-intrusive load monitoring (NILM) is a technique to identify the operating states of electric appliances and support demand response with minimal consumer inconvenience. The CPES real-time electrical load information sensing and communication networks provide an effective avenue for distributed NILM. This paper proposes a non-intrusive multi-model collaborative load identification method. CPES distributed load sensing networks are designed to collect real-time electrical load data in a manufacturing center. The power spectral density (PSD) and principal component analysis (PCA) are employed for feature extraction and reduction respectively. Gaussian process classifier is presented for NILM. Multiple identification models are used to enhance the applicability of the method and committee voting mechanism is designed to make the best global decision. Experiments result show that the proposed NILM method under CPES infrastructure can identify electrical load accurately and robustly.