PaperlayerView exact revision on arXiv

Source-linked AI summary

Modeling and Analysis of Energy Harvesting and Smart Grid-Powered Wireless Communication Networks: A Contemporary Survey

Shuyan Hu, Xiaojing Chen, Wei Ni, Xin Wang, Ekram Hossain

arXiv:1912.13203v1eess.SYcs.SYmath.OC

TL;DR

Rising electricity bills amid growing wireless demand motivate energy-efficient networks that integrate renewable energy sources. This survey reviews energy harvesting and smart-grid coordination across wireless architectures, reporting that renewable energy can reduce providers’ bills and brown-energy consumption.

  • Problem

    Growing wireless-service demand is driving cellular operators’ electricity bills upward, motivating approaches that integrate renewable energy sources.

    Source: §X · p. 20

  • Method

    The article surveys energy harvesting across wireless architectures and reviews energy redistribution, planning, and trading with smart grids.

    Source: §I · p. 2, §XIII · p. 25

  • Results

    Renewable energy use can significantly reduce wireless service providers’ electricity bills and brown-energy consumption.

    Source: §I · p. 2

  • Takeaways & Limitations

    Renewable sources, wireless power transfer, and smart-grid coordination support green, energy-efficient wireless-network operations.

    Source: §I · p. 2

Abstract

from arXiv · show

The advancements in smart power grid and the advocation of ``green communications'' have inspired the wireless communication networks to harness energy from ambient environments and operate in an energy-efficient manner for economic and ecological benefits. This article presents a contemporary review of recent breakthroughs on the utilization, redistribution, trading and planning of energy harvested in future wireless networks interoperating with smart grids. This article starts with classical models of renewable energy harvesting technologies. We embark on constrained operation and optimization of different energy harvesting wireless systems, such as point-to-point, multipoint-to-point, multipoint-to-multipoint, multi-hop, and multi-cell systems. We also review wireless power and information transfer technologies which provide a special implementation of energy harvesting wireless communications. A significant part of the article is devoted to the redistribution of redundant (unused) energy harvested within cellular networks, the energy planning under dynamic pricing when smart grids are in place, and two-way energy trading between cellular networks and smart grids. Applications of different optimization tools, such as convex optimization, Lagrangian dual-based method, subgradient method, and Lyapunov-based online optimization, are compared. This article also collates the potential applications of energy harvesting techniques in emerging (or upcoming) 5G/B5G communication systems. It is revealed that an effective redistribution and two-way trading of energy can significantly reduce the electricity bills of wireless service providers and decrease the consumption of brown energy. A list of interesting research directions are provided, requiring further investigation.

I. INTRODUCTION

The survey addresses a gap in reviews of energy-harvesting wireless systems by examining constrained operations and their optimization alongside interoperability with smart grids. It spans renewable sources, communication architectures, energy management, and prospective 5G/B5G applications.

I. INTRODUCTION

  • Existing reviews do not comprehensively cover both constrained wireless operation powered by renewable energy and its optimization methods, alongside wireless-network interoperability with smart grids.The survey situates this gap amid rising cellular capacity demands and efforts to limit power use.Source: §I · p. 1
  • The survey reviews renewable-energy models, multiple energy-harvesting network architectures, SWIPT, energy management with smart grids, and potential 5G/B5G applications.It compares optimization methods and reports that energy redistribution and two-way trading can reduce operators’ electricity bills and brown-energy consumption.Source: §I · p. 2

A. Types of Renewable Energy

  • The surveyed renewable sources include solar, wind, electromagnetic radiation, thermoelectric, and biomass energy.The article describes energy harvesting as collecting and converting environmental renewable energy into electricity.Source: A. Types of Renewable Energy · p. 2

1) Solar Energy:

  • Solar photovoltaic cells convert sunlight into electricity, but usable harvest varies with time, weather, and other environmental and cell factors.The survey characterizes solar energy as uncontrollable, unpredictable, and nondispatchable.Source: 1) Solar Energy · p. 3

2) Wind Energy:

  • Wind turbines can harvest wind motion, and an anemometer-based approach can also convert lever movement into electrical power.For a 1 m rotor diameter and 8 m/s wind speed, the reported harvest is around 85 W.Source: 2) Wind Energy · p. 3

3) Electromagnetic (EM) Radiation Energy:

  • Electromagnetic energy harvesting uses near-field induction or magnetic resonance, or far-field RF and microwave signals received by antennas and rectifiers.Near-field transfer efficiency can exceed 80%; far-field sources can be transmitter signals or ambient radiation.Source: 3) Electromagnetic (EM) Radiation Energy · p. 3

4) Thermoelectric Energy:

  • Thermoelectric harvesting generates voltage across different conductors at different temperatures, with reported power densities from 10 µW/cm2 to 1 mW/cm2.Temperature gradients can occur in human bodies or machine operations.Source: 4) Thermoelectric Energy · p. 3

5) Biomass Energy:

  • For underwater harvesting, microbial fuel cells use bacterial metabolic activity to generate electricity from broken-down substratum.The passage reports an energy amount of 153 mW/cm2 for this approach.Source: 5) Biomass Energy · p. 3

B. Mathematical Models for Renewable Energy

The section describes models for uncertain renewable-energy supply and introduces optimization approaches for energy-harvesting wireless links. It covers uncertainty sets, stochastic-process models, and point-to-point transmission settings.

1) Uncertainty Sets:

  • Forecasted RES uncertainty can be bounded by polyhedral sets over subperiod energy totals or ellipsoids around nominal harvest, with historical statistics informing set design.The ellipsoid’s matrix determines its outline and thus prediction accuracy.Source: 1) Uncertainty Sets · p. 3

2) Stationary Process:

  • For independently and identically distributed RES, Weibull and Beta distributions model wind and solar variation, respectively, while Gaussian generation is another approximation.The passage associates Weibull with wind-speed oscillation and Beta with solar-power alteration.Source: 2) Stationary Process · p. 3

3) Markov Chain:

  • Markov models represent temporal dependence in RES and battery states, including slotted active/inactive sensor states and stationary or non-stationary harvesting processes.A transition occurs when energy is harvested; battery replacement can also be modeled as a state transition.Source: 3) Markov Chain · p. 4

III. ENERGY HARVESTING-BASED COMMUNICATION OVER POINT-TO-POINT WIRELESS LINK

  • For point-to-point links, convex programming, Lagrange/KKT methods, and dynamic programming are used; DP decomposes problems recursively but faces dimensionality, while traffic arrivals may be heavy or moderate.Heavy arrivals are assumed available before transmission; moderate arrivals occur during transmission.Source: §III · p. 4

A. Heavy Data Arrival

With saturated traffic, energy-harvesting studies optimize transmission schedules for throughput or completion time. The policies must account for energy causality, finite battery capacity, channel variation, and non-negligible circuit power.

A. Heavy Data Arrival

  • For saturated traffic, studies maximize total throughput or minimize completion time over [0, T], typically assuming data are always available.The reviewed methods include convex programming, Lagrange multipliers, and dynamic programming.Source: A. Heavy Data Arrival · p. 4

1) Idle Circuit Power:

  • Energy-feasible throughput schedules keep transmitted energy below harvested energy, within the battery-capacity bound, and use all harvested energy by deadline; the optimal policy is a shortest feasible path.The string-tautening construction places the transmitted-energy curve between the harvested-energy and battery-overflow curves.Source: 1) Idle Circuit Power · pp. 4–5, Figure 3 · p. 5
  • Directional water-filling adapts power to time-varying channels while energy causality limits first-slot use to H when H arrives at both 0 and T/2.Without that later arrival, 2H is available at time 0; stronger channel gains receive higher transmit powers.Source: 1) Idle Circuit Power · pp. 5–6, Figure 4 · p. 6

2) Non-Idle Circuit Power:

  • Non-negligible circuit power can make sleep periods optimal; on-off scheduling uses a minimum transmit power when active, and a two-phase scheme combines EE-maximizing and non-decreasing SE-maximizing transmission.The on-off policy switches off once harvested energy is used up when the ideal-case optimal power falls below the minimum level.Source: 2) Non-Idle Circuit Power · pp. 6–7
  • Directional glue-pouring allocates at least the circuitry-dependent minimum power within each epoch, with harvested energy determining whether transmission covers part or all of the epoch.The allocation is formulated as a convex optimization problem for time-varying channels.Source: 2) Non-Idle Circuit Power · p. 6
  • Battery-loss-aware scheduling combines Lagrange multipliers with backward-induction dynamic programming, while separate work characterizes the EE-maximizing power for the EE–SE tradeoff.The dynamic-programming method locates a zero-battery slot while accounting for battery-history effects.Source: 2) Non-Idle Circuit Power · pp. 6–7

B. Moderate Data Arrival

With bursty data and energy arrivals, schedules must satisfy both data causality and deadlines as well as energy constraints. The reviewed approaches use string-tautening and water-filling constructions, with extensions to online scheduling and grid power.

B. Moderate Data Arrival

  • Bursty arrivals and packet deadlines require the departure curve to remain below arrived data and above the data due by each time, jointly respecting data- and energy-domain constraints.The heavy-data-arrival schedule may be infeasible when the buffer lacks enough data.Source: 1) Non-Stationary data and energy… · p. 7
  • For moderate data arrivals with energy available initially, calculus-based string tautening minimizes transmission energy by placing the departure curve between arrival and deadline curves.The rate changes when the data buffer empties or deadline-approaching data have been transmitted.Source: 1) Non-Stationary data and energy… · p. 7
  • For time-varying channels, convex optimization and KKT conditions yield a minimum-power schedule with multiple water levels, visualized as a shortest path between arrival and minimum-departure curves.The corresponding figure shows the water-arrival curve, deadline curve, and water-filling with multiple levels.Source: 1) Non-Stationary data and energy… · p. 7, Figure 5 · p. 7
  • Dynamic string tautening handles bursty data and energy arrivals, including finite-battery and non-ideal-circuit settings; substituting Φ_t = r_t l_t makes the formulation convex.Only the data-departure curve needs optimization because energy constraints can be translated into the data domain.Source: 1) Non-Stationary data and energy… · p. 8
  • Dynamic string-tautening policies extend to online operation and time-varying channels, and models also examine balancing QoS guarantees against energy use and minimizing completion time.The online policy is generated in real time.Source: 1) Non-Stationary data and energy… · p. 8
  • With constant grid supply alongside harvesting, reviewed methods minimize total or grid power using staged water-filling or online stochastic dynamic programming and convex optimization.The multi-stage water-filling strategy allocates energy backward from the last frame and is generated offline.Source: 1) Non-Stationary data and energy… · p. 8

2) Stationary data and energy arrival processes:

For stationary data and energy arrivals, Lyapunov optimization supports online resource allocation while stabilizing queues; the section then turns to scheduling multiple EH devices that share a sink.

2) Stationary data and energy arrival processes:

  • Lyapunov drift-plus-penalty minimization targets low time-average energy cost while stabilizing data and energy buffers.The framework minimizes an upper bound per slot; the cited work also studies power-delay tradeoffs in the small-delay regime.Source: 2) Stationary data and energy arrival… · p. 8
  • Q-learning for Markov energy and data arrivals asymptotically converges to the global throughput optimum as learning time increases.The transmitter learns by exploring packet-drop or packet-transmission actions.Source: 2) Stationary data and energy arrival… · p. 8

IV. ENERGY HARVESTING-BASED MULTIPOINT-TO-POINT WIRELESS SYSTEM

  • Multipoint-to-point systems schedule multiple EH sensors to send data to one sink, centrally using queue and battery states or distributively through random access –.This energy-aware scheduling distinguishes these systems from point-to-point links.Source: §IV · p. 8

A. Centralized Scheduling

  • For EH multiple-access channels, backward water-filling supports deadline-constrained throughput and completion-time optimization, while iterative dynamic water-filling accounts for per-slot energy limits.The iterative method is reported to converge within a few iterations in practice.Source: A. Centralized Scheduling · p. 9
  • Smart-grid-powered MIMO resource allocation maximizes weighted or expected sum-rate using uplink-downlink duality and an offline Lagrangian-dual subgradient method,.Downlink covariance matrices are derived from their uplink counterparts.Source: A. Centralized Scheduling · p. 9

B. Distributed Scheduling

Distributed EH scheduling addresses unpredictable energy arrivals, while the section also surveys CoMP systems and energy-aware beamforming methods. Smart-grid coordination can trade energy and support higher sum rates under a fixed cost budget.

B. Distributed Scheduling

  • For synchronized Bernoulli energy arrivals, the distributed fractional-power policy is near-optimal; asynchronous arrivals yield greater throughput than synchronized arrivals.The policy's allocated power decreases toward the end of each renewal time.Source: B. Distributed Scheduling · p. 9
  • A smart-grid-powered base station buys and stores more electricity at low prices, draws on its battery at high prices, and can sell surplus; it achieves higher sum rates than a traditional grid-powered base station under the same cost budget.The online stochastic-subgradient schedule is asymptotically optimal and feasible without prior knowledge of the underlying randomness, under the stated i.i.d. assumption.Source: B. Distributed Scheduling · p. 9

V. ENERGY HARVESTING-BASED MULTIPOINT-TO-MULTIPOINT WIRELESS COMMUNICATION SYSTEM

  • CoMP coordinates geographically collocated transmit antennas to serve users and manage interference; RES can reduce reliance on traditional power sources in these energy-intensive systems.CoMP methods include Lagrangian dual, subgradient, cutting-plane, and proximal bundle approaches; cutting-plane iterations are fewer but more computationally demanding and not distributable.Source: §V · p. 10

A. Minimization of Total Energy Consumption

  • For smart-grid-powered CoMP, uplink-downlink duality and convex optimization yield an optimal energy-minimizing solution, while zero-forcing beamforming offers a lower-complexity alternative.Zero forcing cancels mutual interference and is implementable only when the number of users is below the total number of transmit antennas across base stations.Source: A. Minimization of Total Energy… · pp. 10–11

B. Minimization of Operational Cost

The section surveys CoMP energy planning that accounts for storage, renewable supply, grid transactions, and dynamic prices. It contrasts offline worst-case optimization with online methods for long-term average cost.

B. Minimization of Operational Cost

  • A centralized controller coordinates CoMP base stations with renewables, storage, and smart meters to plan grid purchases and sales under dynamic pricing and imperfect transmitter CSI.The formulation minimizes worst-case system-wide transaction expenditure while guaranteeing user QoS.Source: B. Minimization of Operational Cost · p. 11
  • Semidefinite relaxation and the S-procedure produce a convex energy-management and beamforming problem; subgradient iteration with a proximal bundle method solves its nonsmooth formulation offline [114].Ahead-of-time planning requires energy and data arrivals to be known beforehand.Source: B. Minimization of Operational Cost · p. 11

C. Minimization of Long-term Average Cost

  • Battery dynamics make long-horizon optimization difficult for traditional solvers such as dynamic programming because they couple decisions across time.A two-scale online method replaces time-coupled battery constraints with time-averaged constraints and uses Lyapunov drift-plus-penalty control to obtain asymptotically close-to-optimal allocations without prior statistical knowledge, with proper settings.Source: C. Minimization of Long-term Average… · pp. 11–12
  • For i.i.d. renewable supply and transaction prices, stochastic subgradient updates and virtual energy queues support low-complexity online cost minimization; the approach is asymptotically optimal over the long term –.Virtual energy queues may take negative values, unlike real queues.Source: C. Minimization of Long-term Average… · pp. 11–12

VI. ENERGY HARVESTING-BASED MULTI-HOP WIRELESS COMMUNICATION LINK

The multi-hop section examines information cooperation between EH sources and relays, then extends to energy transfer and SWIPT. The surveyed approaches exploit energy-arrival profiles, queue stability, or channel conditions to shape transmission.

A. Information Cooperation Between Nodes

  • Directional water-filling bounds performance in a dual-hop link with an EH source and persistent-power relay; optimal transmit power is constant within each interval, and source-energy and relay-data buffers empty at interval ends.The relaxed EH profile is modified to obtain the original non-relaxed power-allocation solution.Source: A. Information Cooperation Between Nodes · p. 12, Figure 8 · p. 12
  • For decode-and-forward EH relays with known energy profiles, delay-constrained throughput uses KKT-based power allocation, while non-delay-constrained traffic can exploit independent source and relay energy arrivals for larger capacity.The non-delay-constrained solution optimizes source power first and then relay power; relaxing decoding delay enables energy diversity.Source: A. Information Cooperation Between Nodes · p. 12
  • In collision-prone two-hop links where source and relay both receive traffic, queue-stability conditions yield tight inner and outer stability-region bounds for a given transmission probability.Source: A. Information Cooperation Between Nodes · p. 12

B. Information and Energy Cooperations Between Nodes

  • Joint power control, routing, and inter-node energy transfer converges toward a Pareto-optimal equilibrium; without energy cooperation, higher transmit power is assigned to noisier or higher-flow links.Source: B. Information and Energy Cooperations… · p. 13
  • With SWIPT, a centralized sequential-water-filling scheme serves better-channel users first to minimize outage, while a distributed auction balances system performance and computational complexity.The relay updates power allocation at each iteration after destination nodes submit bids.Source: B. Information and Energy Cooperations… · p. 13

VII. WIRELESS POWER AND INFORMATION TRANSFER

Wireless power transfer offers on-demand energy replenishment, while SWIPT systems jointly manage information transmission and energy harvesting. The survey reviews SISO and MIMO designs, including receiver policies, beamforming, and rate-energy trade-offs.

VII. WIRELESS POWER AND INFORMATION TRANSFER

  • WPT can replenish low-power devices on demand, and SWIPT studies how radio signals can transfer power and information simultaneously.WPT can use induction, magnetic resonance, or electromagnetic radiation.Source: §VII · p. 13

A. SWIPT-Enabled Single-Output Systems

  • Dynamic power splitting jointly optimizes receiver splitting factors and transmit power to maximize ergodic capacity while meeting an energy-harvesting requirement.The cited SISO design uses receiver-known channel state information and a Lagrange-duality approach for the globally optimal solution.Source: A. SWIPT-Enabled Single-Output Systems · p. 13
  • SISO studies optimize energy efficiency, rates, transmit power, and time allocation under power, data-rate, and energy-harvesting requirements.The surveyed designs include OFDM and NOMA policies, multi-antenna beamforming, harvest-then-transmit schemes, and nonlinear harvesting with fairness or secrecy objectives.Source: A. SWIPT-Enabled Single-Output Systems · pp. 13–14

B. SWIPT-Enabled MIMO System

  • For MIMO broadcast channels, separated energy and information receivers yield a rate-energy trade-off, while co-located receivers use time switching or power splitting.The rate-energy regions are illustrated for the separated-receiver case in Fig. 10 and for practical policies in the co-located-receiver case in Fig. 11.Source: B. SWIPT-Enabled MIMO System · p. 14, Figure 10 · p. 14, Figure 11 · p. 14
  • Collaborative MIMO SWIPT jointly designs uplink and downlink beamformers, operating time, and user activation to meet downlink QoS while maximizing uplink data rate.The user-selection approach has lower computational complexity with a marginal sum-rate penalty than a combinatorial integer-programming alternative.Source: B. SWIPT-Enabled MIMO System · pp. 14–15
  • Other MIMO designs optimize power splitting for mmWave coverage or use finite-alphabet-aware precoding to meet energy constraints with near-optimal beamforming.A switch-based receiver framework targets low power use; a specialized co-located-receiver algorithm approaches the general SDR solution with substantially lower complexity.Source: B. SWIPT-Enabled MIMO System · p. 15

VIII. SMART GRID-POWERED WIRELESS NETWORKS

Smart-grid-powered cellular networks combine renewable sources, storage, energy sharing, and grid transactions. The survey reviews utility and efficiency objectives alongside optimization approaches for user engagement and energy-consumption reduction.

VIII. SMART GRID-POWERED WIRELESS NETWORKS

  • Smart-grid-powered base stations can redistribute redundant energy and transact bidirectionally with the grid, with optimization studies targeting cost, consumption, efficiency, or operator utility.The survey groups methods including game theory, dynamic programming, and stochastic gradient descent, and distinguishes energy harvesting and sharing from dynamic-price purchasing and two-way trading.Source: §VIII · p. 15, Figure 12 · p. 16

A. Utility Maximization

  • A two-tier system shares renewable energy and battery storage through an aggregator while maximizing system energy efficiency subject to average cell sum-rate demand.The cited offline solution uses an extended convex-concave procedure followed by the Dinkelbach method, with polynomial-time inner optimization.Source: A. Utility Maximization · pp. 15–16

B. User-Engaged Energy-Efficient Communication Scheme

  • User-engaged designs jointly optimize association and power control under QoS constraints, while mobile-user energy sharing matches peers using mobility and energy availability.The cited two-tier NOMA design reports higher energy efficiency than alternatives; peer matching seeks pairs unlikely to undergo energy outage.Source: B. User-Engaged Energy-Efficient… · p. 16
  • Energy-aware traffic offloading and base-station association coordinate traffic and renewable-energy allocation to reduce grid use.The cited offline approach assumes infinite batteries and non-causal renewable-energy and traffic information; online heuristics use finite batteries and causal information.Source: B. User-Engaged Energy-Efficient… · pp. 16–17

C. Minimization of Energy Consumption

  • For base-station energy collaboration, deterministic renewable-energy and demand profiles yield a linear program, whereas causal stochastic profiles motivate a greedy online policy.Under stated loss-coefficient conditions, energy transfer is optimal; the surveyed analysis also finds local storage more cost-saving than transferring energy for storage elsewhere.Source: C. Minimization of Energy Consumption · pp. 17–18
  • Two-phase dynamic programming reduces computational complexity for base-station mode and resource-block decisions, while centralized and decentralized schemes target lower grid expenditure.The two-phase approach is optimal when traffic is uniform; the distributed method uses locally available information and is reported to reduce system expenditure.Source: C. Minimization of Energy Consumption · p. 18

D. Lyapunov-based Online Optimization

The survey describes online Lyapunov methods for energy and service-cost objectives, then reviews dynamic-price planning that adapts base-station operation and electricity procurement. It also covers utility optimization that incorporates operator profit and emissions.

D. Lyapunov-based Online Optimization

  • A leakage-aware Lyapunov strategy minimizes grid energy use in a queueing OFDMA system with battery leakage, using current channel and battery states without prior state knowledge.The minimum grid-energy value is shown to be attainable asymptotically.Source: D. Lyapunov-based Online Optimization · pp. 18–19
  • Lyapunov-based joint base-station association and power control minimizes long-term average service expenditure using instantaneous states, with asymptotic optimality as control parameter V increases.The per-slot decisions require solving a deterministic problem and do not require prior distribution statistics for channel or energy-harvesting processes.Source: D. Lyapunov-based Online Optimization · p. 19

IX. ENERGY PLANNING UNDER DYNAMIC PRICING

  • Dynamic pricing lets cooperating base stations plan purchases, storage, and sharing to reduce system-wide operational expenditure.The cited large-scale green-network design combines stochastic geometry and dynamic programming for activation and procurement, then uses a lower-complexity near-optimal purchasing policy to address DP's scaling limitation.Source: §IX · p. 19, A. Operational Cost Minimization · p. 19

B. Utility Maximization

  • Utility-oriented planning jointly considers base-station sleep states, electricity procurement, network demand, retailer capacity, and CO2 emissions.The reviewed work compares sum, max-min, and proportional-fair utility criteria, with decentralized optimization offering faster convergence and lower complexity at a performance cost.Source: B. Utility Maximization · p. 19

C. Game Theory for Energy Procurement

The survey reviews game-theoretic electricity procurement and cooperation among operators, followed by two-way energy trading and coordinated wireless-resource management. It also describes distributed beamforming designs for operational-cost reduction.

C. Game Theory for Energy Procurement

  • A two-level Stackelberg game models base stations choosing retailers and purchase amounts while retailers set prices; the proposed strategies reduce energy expenditure and CO2 emissions.The cited work proves existence and uniqueness of equilibrium for one formulation and uses a low-complexity iterative algorithm for another.Source: C. Game Theory for Energy Procurement · pp. 19–20
  • Extensions model retailer, macro-cell, and femtocell decisions in a three-level game, while renewable suppliers and network operators also negotiate pricing and storage policies.For the mmWave backhaul case, the distributed policy is reported to yield higher operator profit than the centralized scheme.Source: C. Game Theory for Energy Procurement · p. 20
  • Two competitive mobile operators can use repeated Nash bargaining for day-ahead and real-time group purchases and load sharing, achieving Pareto optimality and reducing both operators’ expenditures.The two stages allocate purchase commitments first and then coordinate real-time purchases, traffic, and inter-operator payments.Source: C. Game Theory for Energy Procurement · p. 20

A. Energy-efficient Schemes

  • Energy collaboration can trade surplus power through the grid or share renewable energy through an aggregator, while communication collaboration shifts traffic, spectrum, or coordinated transmission.Traffic offloading, spectrum sharing, and CoMP operate at different timescales; in energy-harvesting small-cell networks, SNR-based association outperforms distance-based association but still overlooks cross-block coupling.Source: A. Energy-efficient Schemes · p. 21

B. Minimization of Operational Cost

  • Distributed stochastic ADMM beamforming uses conditional-value-at-risk to limit exposure to high costs, with SDR guaranteed tight when channel covariance rank is one or the base-station count is at most two.Simulations verify SDR tightness in general cases; the offline method uses historical realizations without prior knowledge of harvesting amounts or prices.Source: B. Minimization of Operational Cost · p. 21

C. Utility Maximization

The survey reviews utility-focused energy management and optimization, then surveys applications of energy harvesting and smart-grid-powered communications across emerging 5G/B5G systems. The application examples span mobile edge computing, non-orthogonal multiple access, and ultra-reliable low-latency communications.

C. Utility Maximization

  • For micro-grid-powered base stations, cost efficiency is maximized by jointly scheduling generation and transmit power; Dinkelbach optimization and a decentralized ADMM method converge to the global optimum.Auxiliary variables separate coupled constraints across two variable subsets, enabling simple updates at each step.Source: C. Utility Maximization · p. 21

D. Lyapunov-based Online Optimization

  • Lyapunov control supports multi-timescale energy planning and two-way trading without advance knowledge of future CSI, transaction prices, or harvested energy.A harvest-use-trade strategy and Lyapunov drift-plus-penalty reformulation also support long-term cost minimization, with suboptimal SABF and ZFBF algorithms for an NP-complete problem.Source: D. Lyapunov-based Online Optimization · p. 22

XI. APPLICATIONS OF ENERGY HARVESTING AND SMART GRID-POWERED WIRELESS COMMUNICATIONS TO 5G/B5G

  • Energy harvesting and smart-grid-powered communication are relevant to emerging 5G/B5G applications including mobile edge computing, machine learning, NOMA, and URLLC.The survey also discusses applications in subsequent subsections.Source: §XI · p. 22

A. Mobile Edge Computing (MEC)

  • In mobile edge computing, energy-harvesting studies cover servers and mobile users, with online offloading and CPU-frequency decisions informed by congestion and renewable-energy states.Proposed strategies jointly optimize offloading, channel allocation, and computing resources; Lyapunov optimization and deep learning address stochastic environments.Source: A. Mobile Edge Computing (MEC) · p. 22

B. Deep (Reinforcement) Learning

  • Deep learning has been investigated as an optimization method for energy-harvesting wireless communication systems.The survey notes that neural networks can address sophisticated non-convex problems without explicit mathematical formulations.Source: B. Deep (Reinforcement) Learning · p. 22

C. Non-Orthogonal Multiple Access (NOMA)

  • For NOMA, surveyed energy-harvesting and SWIPT work examines relaying, resource allocation, outage, and sum-rate optimization; SWIPT is reported to facilitate user cooperation and alleviate signal-leakage self-interference.In collaborative SWIPT NOMA, opportunistic user-location selection yields lower outage and better throughput than random selection, without compromising diversity gain relative to conventional NOMA.Source: C. Non-Orthogonal Multiple Access (NOMA) · pp. 22–23

D. Ultra-Reliable Low Latency Communications (URLLC)

  • For URLLC, the survey covers an SWIPT-enabled network that maximizes system credibility by selecting parameters under power-splitting and time-switching protocols.Other reviewed work addresses resource allocation, payload allocation, and risk-aware traffic management.Source: D. Ultra-Reliable Low Latency… · p. 23

XII. LESSONS LEARNT AND FUTURE RESEARCH DIRECTIONS

The survey identifies practical open problems in energy-harvesting wireless networks, including imperfectly known system states, multi-user coordination, and computational complexity. It highlights opportunities across point-to-point, multipoint-to-point, multipoint-to-multipoint, and multi-hop systems.

A. Point-to-Point Links; 1) Learn-and-Adapt Algorithms for EH Systems:

2) Multi-User Networks and Energy Cooperation:

  • In multi-user networks, coupled schedules and difficult-to-obtain causal information make optimal decisions based only on current information an open research need.Even a two-hop source schedule affects relay data arrivals, and scheduling complexity generally grows with user count.Source: 2) Multi-User Networks and Energy… · p. 23

B. Multipoint-to-Point Systems; 1) Power Allocation for EH-powered NOMA:

2) Channel Estimation and EH Prediction:

  • Multi-access networks need improved channel estimation and energy-harvesting prediction because many studies assume perfect CSI and non-causal energy knowledge, while more users can worsen interference and estimation errors.Few studies address energy-harvesting prediction in these systems.Source: 2) Channel Estimation and EH Prediction · p. 23

C. Multipoint-to-Multipoint Systems; 1) Implementation of Uplink CoMP Techniques:

  • For multipoint-to-multipoint systems, convex offline problems can attain optimality and online optimizations can be asymptotically optimal; uplink CoMP research should also account for user energy and CoUR implementation challenges.CoUR challenges include low latency, imperfect CSI, signaling overhead, user-side energy harvesting and coordination, and spectral efficiency, energy efficiency, and computational complexity.Source: C. Multipoint-to-Multipoint Systems · pp. 23–24, 1) Implementation of Uplink CoMP… · p. 24

2) Dynamic CoMP Clustering:

  • Dynamic CoMP clustering can adapt cluster size and reduce inter-cluster interference, but signaling overhead, scheduling, and precoding complexity challenge scalable low-cost operation.Machine learning and big-data methods could support online allocation, energy sharing, traffic scheduling, and clustering using locations, traffic, latency, CSI, harvested energy, and prices.Source: 2) Dynamic CoMP Clustering · p. 24

D. Multi-Hop Wireless Links; 1) Opportunistic Routing:

  • Opportunistic routing can deliver packets reliably in disconnected, sparse multi-hop networks by selecting relay nodes opportunistically, but energy-harvesting applications remain limited.The approach uses radio broadcast and can adapt to changing channels, bursty arrivals, and intermittent energy harvesting.Source: D. Multi-Hop Wireless Links · p. 24, 1) Opportunistic Routing · p. 24

2) Impact of Mobility on EH-Based Multi-Hop Routing:

The survey's future directions address unresolved design challenges in energy-harvesting routing, SWIPT, energy trading and planning, end-user energy harvesting, and machine learning. Its conclusion reports that redistributing and trading redundant energy can reduce providers' electricity bills and brown-energy consumption.

2) Impact of Mobility on EH-Based Multi-Hop Routing:

  • Mobility-aware resource allocation for energy-harvesting multi-hop routing remains missing, and the effect of mobility on routing decisions needs investigation.Existing mobility models capture node position, direction, and velocity, but cited studies concern one-hop links.Source: 2) Impact of Mobility on EH-Based… · p. 24

3) Network Coding-Aware Routing:

  • Integrating renewable energy and energy harvesting into network-coding-aware routing remains an open problem for multi-hop networks.The routing approach uses omnidirectional antennas and wireless broadcast; cited work reports benefits including capacity, reliability, delay, and energy consumption.Source: 3) Network Coding-Aware Routing · p. 24

E. SWIPT Systems; 1) Security of SWIPT:

  • In SWIPT, greater transmit power can increase desired transferred power while also escalating eavesdropping risk, motivating designs that strengthen legitimate reception while limiting eavesdropper signal intensity.The survey identifies security as a research direction for SWIPT systems.Source: E. SWIPT Systems · p. 24, 1) Security of SWIPT · p. 24

2) SWIPT for CoMP Systems:

  • SWIPT with CoMP may reduce total transmit power through cooperation, but involving all base stations and users in energy transfer and information sharing requires enormous backhaul capacity and interference management.Coordinated beamforming is preferred over joint transmission from a practicality perspective because it requires much less signal communication overhead.Source: 2) SWIPT for CoMP Systems · pp. 24–25

3) Robust Designs of SWIPT Systems:

  • Dynamic information transmission and energy harvesting make accurate advance CSI difficult to obtain, while imprecise-CSI offline policies may cause high outage; robust beamforming remains a challenge.The passage frames robust designs as necessary for dynamic SWIPT systems.Source: 3) Robust Designs of SWIPT Systems · p. 25

F. Energy Trading and Planning; 1) Energy Harvesting at End Users:; XIII. CONCLUSION

  • The survey identifies end-user energy harvesting and distributed optimization as research needs, and concludes that energy redistribution and trading can reduce provider electricity bills and brown-energy consumption.End users consume energy for reception, decoding, and uplink communication; user grouping introduces energy sharing, allocation, interference, QoS, and complexity considerations.Source: 1) Energy Harvesting at End Users · p. 25, §XIII · p. 25

2) Application of Machine Learning:

  • Machine learning could support user spectral-efficiency and QoS learning, smart-grid base-station power allocation, and online load sharing; applications of learning techniques in energy-harvesting network design remain largely open.The survey discusses supervised, unsupervised, and reinforcement learning for prediction, clustering, and decision-making applications.Source: 2) Application of Machine Learning · p. 25
Loading 1912.13203v1…