Prototype Implementation of Dynamic Data Pruning in Smart Energy Meter

Abstract

Smart electricity meter is an example of the Internet of Things (IoT) device which is nowadays being installed in domestic and industrial premises to monitor the power consumption along with various other services such as billing, load forecasting, and dynamic pricing. Usually, these IoT devices sample at a high rate and report the whole data to the cloud. In this process a large amount of energy is consumed at the IoT node for transmission of the data and a large bandwidth is used for transmission over the Internet, though for storage saving some compression is performed at the destination. To this end, we aim to prune the IoT data at the source node by providing some intelligence, thereby aiding to the wireless IoT node’s energy sustainability and efficient channel bandwidth usage for scalable deployment. The IoT devices measuring/reporting single parameter or multiple parameters are judiciously pruned within an acceptable reconstruction error limit. In this paper, we report embedded implementation of data-driven dynamic pruning of multi-parameter smart meter data as an example demonstration of data-smart IoT nodes. Our performance results show that the energy and bandwidth savings with multiple univariate data pruning are respectively about 19% and 36%, whereas the savings with multivariate data pruning are respectively about 36% and 98%. The developed embedded data pruning module is 99.09% more energy efficient than the implementation on Raspberry Pi.