Research On Data-driven Architecture And Kye Technologies Of Smart Consumption Grid

The user-side energy management system(UEMS)is the key to realize the Smart Grid on the user side.The current UEMS is designed for monitoring relatively small system from the concept to the software development,and it has some shortcomings such as lack of online scalability,low efficiency of data transmission,information island and so on.Thus,the current UEMS cannot adapt to the future development of user-side Smart Grid with massive users,appliances,scenarios,and data.In this dissertation,smart consumption grid(SCG)is proposed as the user-side paradigm of Smart Grid.With the help of the rapid development of Internet of things,big data,artificial intelligence,edge computing,this dissertation proposes a systematic solution including architecture design,data sharing,data transmitting,software and hardware development.The main contributions are listed as the following:1)An autonomous decentralized system(ADS)-based data-driven architecture of the SCG is proposed.The six major characteristics and the four core factors of the SCG are concluded from the aspect of cyber-physical system(CPS).Then,the two major challenges of the SCG are proposed.The first challenge is that the system should have physically-distributed and logically-integrated architecture with online scalability and online robustness.The second challenge is that the data should be shared horizontally and efficiently transmitted vertically.Based on the above analysis,an ADS-based data-driven architecture of the SCG is proposed,and its design concept,logical architecture and characteristics are introduced.2)To solve the problem of data sharing,a communication protocol and a data meta-interface of the SCG based on MQTT/MQTT-SN are proposed,which realized spatial decoupling,time decoupling and synchronization decoupling of the data interation in the SCG.Firstly,the characteristics,terms,message formats,realization notices of the MQTT/MQTT-SN are introduced.In order to demonstrate that the MQTT/MQTT-SN is the most suitable for the SCG,the common communication protocols such as HTTP,Co AP,AMQP and DDS are compared with MQTT/MQTT-SN from the aspects of data exchange mechanism,application range,design concept,and technical feasibility.Then,the three common entity services,i.e.,measurement service,control service and upgrade service,are summarized,and the forms of the entity services as well as the access process are formalized,then a four-layer architecture data meta-interface is proposed.3)To solve the problem of the efficient data transmission,two efficient real-time data transmission methods are proposed.The first method is based on the Lebesgue Sampling(LS).The mathematical model of the LS is established,and its performance is compared with the traditional polling method.The LS-based method has very low requirements for computing and storage resources,and is applicable to any hardware platform and time series data.The second method is the Online Pattern-based Data Compression(OPDC)method.Different from the traditional methods based on predefined functions,the method is based on the shape patterns of the time series data,and the most appropriate shape patterns are extracted by data mining for compression and transmission.Its implementation is supported by an online adaptive segmenting algorithm,the similarity measure based on reconstruction error,the pattern curves extraction based on clustering ensemble,the time series indexing based on random forest,and a decompression algorithm based on the patterns.The temporal complexity and spatial complexity of OPDC method is O(n)and O(1),respectively.The performance of OPDC is compared with the 6 advanced online compression algorithms(FSW,Optimal PLR,Conn Seg Alg,Conn Alg I,AA and LS)based on the actual datasets of electrical appliances.The outperformance of OPDC is demonstrated based on the statistical methods.OPDC method has certain requirements on computing and storage resources,and will be affected by the characteristics of time series.4)The SCG has been put into use on monitoring more than 10,000 electrical appliances at Intel Asia Pacific R&D center and other places.Taking the SCGs for electrical heating applied in 9 schools in Cangzhou,Hebei province as examples,the typical application of the SCG is introduced.The application of the data-driven architecture and key technologies of the SCG are introduced including the online expansion,the online robustness,the data sharing and the efficient data transmitting.The smart socket is taken as an example to introduce the development of the core monitoring equipment of the SCG,and formal specification and verification are carried out based on UPPAAL for four key function modules of the smart socket,i.e.,connection module,measurement module,control module and upgrade module.A cooperative operation optimization strategy of the SCG is proposed.In order to improve the cyber-physical characteristics of SCG and guarantee the actual performance of the above strategy,a solution based on Qo S improvement and network topology optimization is proposed.Then,the average daily electricity consumption of each school with the optimization strategy decreased by 8.2%,and the maximum decreased 14.9%.