Research On Traffic Adaptation Of Multi-mode Heterogeneous Communication Network In Low-carbon Parks

Low-carbon smart parks have become a new trend to achieve carbon peaking and carbon neutrality goals.In particular,distributed photostatic electricity,energy storage units and loads are integrated into smart parks based on DC distribution and coordinated dispatch to reduce carbon emissions.New low-carbon services,such as carbon footprint monitoring,electricity spot market,and distributed renewable energy dispatching,put forward higher requirements for high concurrency,millisecond-level control,and data transmission delay and reliability.Therefore,a single communication media cannot meet the differentiated requirements.This paper conducts in-depth research on the above problems,and the specific research work is summarized as follows:(1)Build a heterogeneous access network for intelligent devices in low-carbon smart parks.For the low-carbon monitoring needs of power equipment such as distributed photovoltaics,distributed energy storage,electric vehicles,buildings,and DC distribution networks,multi-mode devices are installed for each electrical equipment to access heterogeneous networks.The collected data is transmitted to the energy management platform through 5G.In particular,when the multi-mode device is accessed,it is necessary to combine the application-layer traffic access control and physical layer traffic allocation according to the network status to construct a jointly optimized action space,which reduces the complexity of joint optimization by reducing the spatial dimension.(2)Build a multi-mode multi-layer communication network with efficient data backhaul.Aiming at the low latency requirement of backhauling data,a multi-mode multilayer communication network with efficient data backhaul is constructed,and the monitoring data is controlled by gateways and transmitted to servers through multi-mode heterogeneous networks,supporting low-carbon services such as comprehensive benefit analysis of electric carbon,carbon emission monitoring and measurement,and electric carbon spot market.(3)An intelligent traffic adaptation algorithm based on deep reinforcement learning is proposed.Deep reinforcement learning based algorithms can approximate Q value based on system status.Based on the estimated Q value,the optimal traffic adaptation is learned by balancing exploration and exploiting trade-offs.The learning strategy relies only on local information,which is updated based on observed rewards for rapid convergence.In the example,the average queue delay,throughput,and probability of extreme events are compared in detail,and the effectiveness of the proposed algorithm are verified.