Report

Energy-Efficient Cognitive Heterogeneous Networks Powered by the Smart Grid Authors: Shengrong Bu, F. Richard Yu and Yi Qian Presenter: Ran Zhang Main Reference Shengrong Bu, F. Richard Yu, and Yi Qian, “Energy-efﬁcient cognitive heterogeneous networks powered by the smart grid,” INFOCOM’13, IEEE Proceedings, 2013. 2 OUTLINE Introduction • Background and Contributions System Model • • • • Cognitive Heterogeneous Mobile Network Model Electricity Consumption Model for BSs Real-Time Pricing in Demand Side Management (DRM) Complete System Model Problem Formulation • • Three Stage Stackelberg Game Utility Functions for Different Levels Analysis of the Proposed Game • Backward Induction Method Simulations Conclusions 3 OUTLINE Introduction • Background and Contributions System Model • • • • Cognitive Heterogeneous Mobile Network Model Electricity Consumption Model for BSs Real-Time Pricing in Demand Side Management (DRM) Complete System Model Problem Formulation • • Three Stage Stackelberg Game Utility Functions for Different Levels Analysis of the Proposed Game • Backward Induction Method Simulations Conclusions 4 Introduction – Background Energy Efficiency • • Energy cost account for almost half of its annual operating expenses (cellular) CO2 emissions Heterogeneous Networks (HetNets) • • • Smaller cells overlaid with macrocell – effective solution to energy efficiency Macrocell: large coverage and mobility management Femtocell: user-deployed, share the same channel with macrocell BS • • Higher data rates, more energy efficient Increase the handoff rates, inter-cell interference 5 Introduction – Background Cognitive Radio Technology • • Originally proposed to improve spectrum efficiency Mitigate interference and improve energy efficiency Smart Grid (SG) • • • Create two-way information exchange flows via communication technologies, greater flexibility and more important role for customers. Electricity price can be dramatically fluctuating (negative price) What kind of information should be sensed in HetNets powered by SG? • • Radio spectrum environment Smart grid environment 6 Introduction – Contributions Real-time pricing for demand-side management • • Multiple retailers, real-time prices; BSs of both marcro- and femto- cells dynamically decide from which retailer and the amount of electricity they will buy (energy-efficient power allocation) Price decision model • • Homogeneous Bertrand game with asymmetric costs Electricity cost formulation Interference price • Offered by macro BSs to femto BSs to mitigate\control interference from femtocells Three-level Stackelberg game • • • • Electricity price decision (retailer level) Power allocation of Macro BS and interference price decision (MBS level) Power allocation of Femtocell BS (FBS level) Backward induction method is proposed to achieve the equilibrium solution 7 OUTLINE Introduction • Background and Contributions System Model • • • • Cognitive Heterogeneous Mobile Network Model Electricity Consumption Model for BSs Real-Time Pricing in Demand Side Management (DRM) Complete System Model Problem Formulation • • Three Stage Stackelberg Game Utility Functions for Different Levels Analysis of the Proposed Game • Backward Induction Method Simulations Conclusions 8 System Model Cognitive Heterogeneous Mobile Networks with Femtocells Powered by Smart Grid 9 System Model – Cognitive HetNets Model One MBS and multiple FBSs (wiredly connected) MBS is aware of spectrum access of FBSs and FBSs can monitor the spectrum environment and randomly access the spectrum Slotted manner Multiple subchannels. users use OFDMA to communicate with BSs Assumptions • • • Macro- and femto- cells share spectrum There is one scheduled active femtocell user in each slot in each femtocell No interference between femtocells, only interference between femtocells and macrocell – scarcely distributed 10 System Model – Electricity Consumption Model For energy efficient communications • Energy-efficient metric: the weighted transmission rate minus the weighted electricity cost • Electricity cost: amount of consumed electricity times the real-time price • Amount of consumed electricity (transmission power and other consumptions) total tx power Efficiency of PA Dynamic Power Consumption Static Power Consumption 11 System Model – Real-Time Pricing Demand-Side Management (DSM) • A set of programs implemented in utility companies • Help utilities operate more efficiently, reduce CO2 emissions, decrease the cost of customers • Each retailer competes with each other and aims to maximize its own utility given the prices offered by other retailers 12 System Model – Real-Time Pricing Complete Model • R retailers, K femtocells, one macrocell user • How the system operates: Retailer: offer real-time price to MBS and FBS MBS: Energy-efficient power allocation, lowest price, issue interference price to FBS FBS: Energy-efficient power allocation, electricity price and interference price 13 OUTLINE Introduction • Background and Contributions System Model • • • • Cognitive Heterogeneous Mobile Network Model Electricity Consumption Model for BSs Real-Time Pricing in Demand Side Management (DRM) Complete System Model Problem Formulation • • Three Stage Stackelberg Game Utility Functions for Different Levels Analysis of the Proposed Game • Backward Induction Method Simulations Conclusions 14 Problem Formulation Three-level Stackelberg Game Goal: maximize the utility of retailers, MBS and FBS. Stage I • • Leader: retailers; follower: MBS and FBSs Retailers offer real-time price xr to MBS and FBSs Stage II • • • Leader: MBS; follower: FBSs MBS decides which retailer it buys electricity from and the amount of electricity (i.e., transmission power pm decision), based on real-time price xr. Offer interference price y based on the received interference from FBSs Stage III • Each FBS decides which retailer to buy from, the amount, based on xr and y. 15 Problem Formulation 16 Problem Formulation- Smart Grid Level Goal: maximize the utility function • • • • • cr: electricity cost (e.g., purchase cost, CO2 taxes) Pmf: additional power consumption Bmv: 1/η pm: MBS tx power; pk: FBS k tx power Srm, Srk: {0,1} Maximize its individual benefit 17 Problem Formulation- MBS Level Goal: maximize the utility function (three parts) • • • • • • W: subchannel bandwidth hm: channel gain from MBS to macrocell user gkm: channel gain from FBS to macrocell user y: interference price α,β: relative weight over transmission rate Tradeoff: interference revenue and transmission rates Maximize its individual benefit 18 Problem Formulation- FBS Level Goal: maximize the utility function (three parts) • • • hk: channel gain from FBS to FBS user srk: {0,1} indicates whether FBS k buys electricity from retailer r μk,λk: relative weight over transmission rate Maximize its individual benefit 19 OUTLINE Introduction • Background and Contributions System Model • • • • Cognitive Heterogeneous Mobile Network Model Electricity Consumption Model for BSs Real-Time Pricing in Demand Side Management (DRM) Complete System Model Problem Formulation • • Three Stage Stackelberg Game Utility Functions for Different Levels Analysis of the Proposed Game • Backward Induction Method Simulations Conclusions 20 Analysis of the Proposed Game Goal: to obtain the stackelberg equilibrium of the three-level game Method • • • Dependencies among different stages Propose a backward induction method to capture the sequential dependence of the decisions FBSs MBS Retailer 21 Analysis – Power Allocation Game for FBSs Already known: interference price y, electricity price xr. Action: choose which retailer r*k, decide the transmission power p*k. Solution: 22 Analysis – MBS Level Game Already known: FBS tx power p*k, electricity price xr, Action: choose which retailer r*m, decide the transmission power p*m, give interference price y Solution: 23 Analysis – Electricity Retailers Already known: tx power p*m and p*k, prices xr of other retailers Action: choose best price to maximize individual profits. Solution: 24 OUTLINE Introduction • Background and Contributions System Model • • • • Cognitive Heterogeneous Mobile Network Model Electricity Consumption Model for BSs Real-Time Pricing in Demand Side Management (DRM) Complete System Model Problem Formulation • • Three Stage Stackelberg Game Utility Functions for Different Levels Analysis of the Proposed Game • Backward Induction Method Simulations Conclusions 25 Simulations How each FBS makes its power allocation decision based on the interference price y Observations • • Decrease tx power with higher interference price Given interference price y, tx power is lower if the lowest electricity price is higher 26 Simulations Utility of MBS vs. interference price Observations • • Piece-wise concave When interference large enough, the utility tends to be stable 27 Simulations Tx power of BSs vs. lowest offered price Observations • • Tx power decreases with the increase of price for both kinds of BSs MBS decrease more significantly as it consumes much more energy than FBSs 28 Simulations Stackelberg equlibrium Observations • • Tx power Converge due to the convergence of price offered by the retailers Equilibrium: retailer 2 set its price equal to its cost, retailer 1 sets its price a little smaller than retailer 2 29 OUTLINE Introduction • Background and Contributions System Model • • • • Cognitive Heterogeneous Mobile Network Model Electricity Consumption Model for BSs Real-Time Pricing in Demand Side Management (DRM) Complete System Model Problem Formulation • • Three Stage Stackelberg Game Utility Functions for Different Levels Analysis of the Proposed Game • Backward Induction Method Simulations Conclusions 30 Conclusions Heterogeneous mobile networks with cognitive radios and femtocells, powered by smart grid. Multiple retailers sell electricity and MBS and FBSs adjust their tx powers based on electricity price and interference price Three-level Stackelberg game is used to model the whole system and homogeneous Betrand Game is used to model the price decision A backward induction method is used to achieve the Stackelberg equilibrium Simulations show that the dynamics of smart grid can have significant impact on the decision process of power allocations. 31