Neural Network based Short Term Forecasting Engine To Optimize Energy And Big Data Storage Resources Of Wireless Sensor Networks

摘要

Energy efficient wireless networks is the primaryudresearch goal for evolving billion device applications like IoT,udsmart grids and CPS. Monitoring of multiple physical eventsudusing sensors and data collection at central gateways is theudgeneral architecture followed by most commercial, residentialudand test bed implementations. Most of the events monitored atudregular intervals are largely redundant/minor variations leadingudto large wastage of data storage resources in Big data servers andudcommunication energy at relay and sensor nodes. In this paperuda novel architecture of Neural Network (NN) based day aheadudsteady state forecasting engine is implemented at the gatewayudusing historical database. Gateway generates an optimal transmitudschedules based on NN outputs thereby reducing the redundantudsensor data when there is minor variations in the respectiveudpredicted sensor estimates. It is observed that NN based loadudforecasting for power monitoring system predicts load with lessudthan 3% Mean Absolute Percentage Error (MAPE). Gatewayudforward transmit schedules to all power sensing nodes day aheadudto reduce sensor and relay nodes communication energy. Matlabudbased simulation for evaluating the benefits of proposed modeludfor extending the wireless network life time is developed andudconfirmed with an emulation scenario of our testbed. Networkudlife time is improved by 43% from the observed results usingudproposed model.