Intelligent Edge Analytics for Load Identification in Smart Meters

Electrical load identification plays a significant role in smart grid applications. Smart meter readings are typically used for load identification. This paper proposes an intelligent edge analytics approach to perform load identification in smart meters. By pushing the data analytics towards the point of sensing, granular level transient features are economically extracted from the signal which is sampled at high frequency in smart meters. Discrete wavelet transform is used for the feature extraction process. Extracted features are fed into an embedded neural network that resides inside smart meter for load identification. Captured raw data is discarded after load identification and only the results are sent out to utility providers. Our experimental results show that, the proposed approach can achieve around 99% of accuracy in load identification as well as around 99.9% of data reduction.