Research On The Key Technology Of Energy Saving For Green Wireless Communication Networks

With the rapid development of communication technology,people's requirements for wireless communication are aslo increasing.Apart from the growing demand for high speed,high capacity and low delay,the number and types of user equipment are also increasing.In order to meet the growing demand for wireless communication,operators need to constantly deploy new network access nodes.Through spectrum reuse and reducing the transmission distance between the transmitter and the receiver,network densification can improve the network capacity,improve the number of user equipment and reduce the transmission delay effectively.However,with the increasing deployment of network access nodes,the energy consumption generated by the network operation is also growing.The research of the ways to reduce the energy consumption of the network has attracted the interests of industry and researchers.If the operators can use the clean and environment-friendly green energy as energy supply for network access nodes,they can reduce the traditional energy consumption during network operation,improve the utilization rate of network energy,and achieve the goal of energy conservation and emission reduction effectively.However,the green energy is inherently instable and unpredictable,which makes the traditional mathematical optimization algorithm which is based on the deterministic model not applicable,resulting in a lot of inconvenience in the use of green energy.In order to make full and reasonable use of green energy,improve the energy efficiency of the network and reduce the traditional energy consumption,this paper utilizes the game theory,reinforcement learning theory,operational research and other mathematical optimization theory,and introduces green energy into the relay and micro base station respectively,uses the relay to reduce the transmission power of the user terminal,and uses the micro base station to reduce the energy consumption of the macro base station,so as to improve the energy efficiency of the network.In addition,the energy harvesting technology is also introduced into the nodes of the Internet of things(IoT)that are difficult to be covered by the traditional power grid,so as to get rid of the dependence on the traditional energy.The research is carried out from three perspectives:the energy saving on the user side,the energy saving on the base station side and the energy saving on the IoT devices.In order to promote the energy saving and emission reduction of the network and solve the energy problems of the network,the main research contents and achievements of this paper are as follows:1.Research on green relay selection and power adjustment based on auction game theory.Considering the difficulties of network resource allocation problem(including relay node selection and relay transmission power allocation problems)in a multi-user and multi-relay scenario,a two-step relay selection and power optimization scheme is proposed based on game theory and convex optimization theory to minimize the energy consumption of user terminals and extend their remaining equipment lifetime.In this scheme,the relay selection process is modeled as an auction game process,in which the user is the buyer,who wants to purchase the cooperative transmission resources provided by the green relay,the green relay is the seller who sells its cooperative transmission services,and the base station acts as the auctioneer.The auction matching decision is made after the base station collects the user's bid and the relay's ask price.The initial price of the resources purchased by the user is related to the rate increasement brought by using the relay,and the initial price of the resources sold by the relay is related to the transmission rate to the base station itself.Then,considering the user's residual energy,the user can increase its bid according to the residual energy to improve the possibility of getting cooperative service;in addition,the relay can also reduce its price according to its instantaneous energy capture to promote the auction to provide services to more users.After calculating the price,the relay and the user send the price to the base station.The base station completes the relay selection based on the Hungarian algorithm.There are three different relay user pairing objectives:maximize the number of users who get the relay service,maximize the total transmission rate of the network,and maximize the total profit of the relays.After the base station completes the auction,the matching results are sent to the relay and the user,and the relay gets the user set it serves.Then,considering the QoS constraints of user uplink transmission,the relay adjusts its transmission power when forwarding each user's data,so that the total transmission power of the users is minimized while meeting the user transmission rate constraints.Finally,the simulation results show that three different relay user pairing targets have their own advantages and disadvantages,but all of them can improve the total capacity of the system and extend the residual lifetime of users.Through the power allocation,the user terminal can meet the requirements of SNR and further reduce its energy consumption,and the algorithm complexity is lower compared to geometric programming algorithm.2.Research on traffic offloading and power allocation of green heterogeneous wireless networks.To deal with the increasing energy consumption problem brought by the deployment of micro base station in the wireless communication network,the macro base station power adjustment scheme and the micro base station dynamic switch on/off scheme are studied to maximize the energy efficiency of the network.In order to reduce the traditional energy consumption of the network,the micro base stations are all powered by green energy.The macro base station is always on to ensure the basic network coverage,and it can adjust its transmission power during operation to reduce the energy consumption.And the micro base station decides whether to be in the work mode or switch to the sleep state according to its own energy state,user demand and channel state.Due to the complexity of channel state variety and the randomness of user demand,the traditional dynamic programming algorithm is not suitable,but reinforcement learning method has the ability to find the optimal action which can maximize the long-term energy efficiency in different states by interacting with the environment.In addition,this paper proposes two optimization schemes,which are distributed multi-agent reinforcement learning and centralized deep reinforcement learning.For the distributed multi-agent reinforcement learning scheme,we model the energy efficiency optimization problem as a Markov game problem.Each base station makes decision independently.Through the interaction with the environment,each station can adjust and optimize its strategy constantly until it is stable according to the reward,so as to find the strategy to maximize the long-term energy efficiency.For the centralized deep reinforcement learning scheme,each green micro base station sends its own energy state and channel state between itself and the user to the macro base station,and the macro base station makes a unified decision.Because the state space after the summary is too large,we use the deep neural network to estimate the rewards of different actions in each state,and propose an optimization scheme based on the deep reinforcement learning.The simulation results show that the energy efficiency of the scheme based on centralized deep reinforcement learning algorithm is higher than that of the scheme based on distributed multi-agent reinforcement learning and greedy algorithm,and the convergence speed is faster,but the scheme based on distributed multi-agent reinforcement learning consumes less communication overhead.3.Throughput optimization of IoT terminals with the energy harvesting abilitySince a large number of IoT terminals are widespreadly located in the world,the cost of connecting all these nodes to the power grid is high.In view of the energy supply problem of IoT terminals,wireless energy carrying transmission technology is considered to be applied to IoT terminals.This technology gives the nodes the ability to capture the RF(radio frequency)energy in the environment as the energy source.Then using a rechargeable battery to store the captured energy and a buffer to store the data,the transmission power allocation problem is studied to maximize the long-term throuput.Since the nodes only know the channel,energy and data related information at the current time,there is no prior knowledge about the future channel/energy/data,so we model the optimal power allocation problem as a reinforcement learning task.The state of the node consists of the data stored in the buffer,the energy stored in the battery,the new coming data amount,the amount of harvesting RF energy and the current channel state.The action is the selected transmission power.As the state space is composed of five elements,it is difficult and complex to traverse each state action pair and estimate all the state action value functions with the growth of the state space.Therefore,we use the deep neural network to estimate the state action value function,and use the scheme based on the deep reinforcement learning algorithm to solve this problem.Finally,the proposed scheme is compared with Q-learning based scheme,linear programming based branch and bound algorithm and greedy algorithm by simulation.The results show that the convergence speed of the scheme based on deep reinforcement learning is faster than that of the scheme based on Q-learning,and the final long-term throughput is close to that of the branch and bound algorithm based on linear programming(this algorithm is based on the assumption that the nodes know the data and the exact amount and time of energy arrival,as well as the channel conditions at each time,which can be regarded as the optimal upper limit),which is superior to the Q-learning based scheme and greedy algorithm.