Yang Song’s scientific contributions

Multi-energy systems (MES) exploit advanced physical information technology and innovative management pattern to achieve collaborative control of multiple heterogeneous energy. The refined control of MES bursts massive delay-sensitive computing tasks, requiring precise system modeling and collaborative computing strategies. However, the unbalanced spatial and temporal distribution of resources and real-time requirements of tasks further increase the difficulty of computing. To address these challenges, a precise modeling and optimization method of collaborative computing in MES is proposed. First, an integrated system framework for regional multi-energy systems (RMES) in northeastern China is designed. Then, the collaborative computing problem with energy and delay constraints in RMES is modeled. It is proved that the problem can be transformed into a joint convex and nonconvex optimization problem. Moreover, an improved simulated annealing-based joint resources optimization (ISA-JRO) scheme is proposed. Finally, extensive simulation results show that ISA-JRO significantly improves the computational resource utilization ratio and reduces the total cost in MES. ISA-JRO reduces at least 25 % of the total cost compared with the traditional methods.

The network candidate routing data stream is restricted to reroute the data stream of same origin and same destination, which brings great limitation to the application of the algorithm. This paper proposes a multi-objective optimization oriented rerouting algorithm for congested data streams. The algorithm detects and aggregates data streams on congested links, and reroutes the detected data streams. When there is no data stream, the data streams are arranged from large to small in the congested links, and the appropriate number of data streams are selected as candidate data streams, and small streams are aggregated by source and destination addresses. In order to determine the maximum network traffic and the optimal routing of candidate data streams, multi-objective optimization is introduced to jointly optimize the network load data stream maximization and network load balancing. Finally, a non dominated genetic algorithm with elite strategy is designed to solve the multi-objective optimization problem.

Network operation and maintenance (O & M) activities of data centers focus mainly on checking the operating states of devices. O & M engineers determine how services are running and the bearing capacity of a data center by checking the operating states of devices. However, this method cannot reflect the real transmission status of business data; therefore, engineers cannot fully comprehensively perceive the overall running conditions of businesses. In this paper, ERSPAN (Encapsulated Remote Switch Port Analyzer) technology is applied to deliver stream matching rules in the forwarding path of TCP packets and mirror the TCP packets into the network O & M AI collector, which is used to conduct an in-depth analysis on the TCP packets, collect traffic statistics, recapture the forwarding path, carry out delayed computing, and identify applications. This enables O & M engineers to comprehensively perceive the service bearing status in a data center, and form a tightly coupled correlation model between networks and services through end-to-end visualized modeling, providing comprehensive technical support for data center optimization and early warning of network risks.