Research On Big Data Technology In Smart Home For Automated Demand Response

At present,the power load gap between peak and valley becomes obvious,how to eliminate the gap and improve electricity efficiency poses new challenges to smart grid.By inducing users to consume electricity rationally,demand response relieves peak power grid pressure,realizes peak load shifting and improves the utilization of power resources.Among all the response terminals in the user side,smart home as a high-power load,is flexible demand for the electricity,whose power can be dynamically adjusted,and reasonable adjustment of its power can ease the power grid pressure.But smart homes are magnanimous,many species and ubiquitous distribution,so there is a need to build a platform to get them together,to realize the unified control and management of them.There are a number of challenges in bringing different vendors' equipment together and get them to participate in demand response,such as inconsistencies in communication protocols,high concurrent system design,and TB-level data storage and processing.Based on smart home,the paper designs the models of demand response capability,aggregator subscription capability,energy saving effect evaluation and user behavior analysis according to the business needs of the automatic demand response system,and the distributed solution of those models.These models need to be excavated from the analysis of TB-level historical data.Traditional methods can't meet the space and time requirements of business,so the paper puts forward and realizes the smart home big data platform.The platform provides big data support for automated demand response systems,composed of demand response terminal system,business data processing system and visualization system three subsystems.Based on MINA,demand response terminal system,supports for high concurrent network communication,and pushes the collected data to the Kafka to solve the mismatch between data reception and processing speed.Based on the concept of Lambda architecture,business data processing system integrates Hadoop with Spark;it uses Spark to get real-time data flow from Kafka for real-time forecasting analysis;it uses MapReduce to implement parallelized KNN algorithm,and classifies users according to the record of electricity consumption;it uses MapReduce to implement the parallelized algorithm of subscription capacity model,energy saving effect evaluation model.Visualization system calls the results of data processing,and provides a visual interface for querying and displaying data.