Solar Powered Cellular Networks: Issues, Challenges and Solutions Biplab Sikdar 1 Overview • Green cellular networks • Motivating factors • System description • Case study and ongoing deployment • Challenges and proposed solutions 2 Need for Greener Base Stations • More than 6 billion mobile subscribers and increasing • 4 million base stations worldwide (~ 25 MWh/year) • Contribution to global energy requirement: 3 % • Contribution to global carbon emissions: 2 % Power consumption of a typical cellular network [2] [2] Hasan, Ziaul, Hamidreza Boostanimehr, and Vijay K. Bhargava. "Green cellular networks: A survey, some research issues and challenges." Communications Surveys & Tutorials, IEEE 13.4 (2011): 524-540. Factors Motivating Solar Powered BSs • Telecom markets shifting to developing countries (specifically in Africa and South Asia) • Challenges: • Poor grid connectivity • Of 400,000 BSs in India, more than 70% of the BSs face power cuts for more than 8 hours a day and many rural areas more than 20 hours a day • Currently around 250,000 off-grid and 550,000 poor-grid BSs in African and South Asian developing nations • High operating expenditure (OPEX) for off-grid sites with diesel generators • The cost of running a site with diesel is around 10 times as compared to using grid energy. Factors Motivating Solar Powered BSs Solar map of world [6] 5 Factors Motivating Solar Powered BSs • Low OPEX: only battery replacement (every 3-5 years). Typical lifetime of PV panels ~ 25-30 years • Less maintenance/ Reduced site visits • Government initiatives • Government regulations: TRAI regulation: 50% rural and 20% urban telecom towers to be powered by renewable energy by 2015 • Government subsidy: TRAI subsidy on solar deployments: North India ( 70%), other parts (30%) Factors Motivating Solar Powered BSs • Greater disaster resistance: • 2011 earthquake in Japan followed by a tsunami, more than 6,700 cellular BSs experienced outages • Solar powered BSs are immune to grid outages and can restore their services faster. • New base stations with low power consumption: • Macro BSs typically have high consumption, requiring large solar panel dimensions, thereby making solar powered solutions impractical. • Recent developments: macro BSs consuming 500-800 W and smaller BSs consuming 50-120 W Setup of Solar Powered BSs http://www.topsunenergy.net/solar-telecom-system.htm • Modeling parameters: • Base station power consumption • Solar energy harvested by PV panels • Battery (lifetime) PV Panels • Arrays of solar PV cells to convert solar energy to electricity • DC rating: power generated when the solar power available on panels is 1 kW/m2. • 1 kW PV panel is typically 5 m2 in area and panel lifetime is more than 25 years • Factors affecting the power produced by a PV panel: • DC rating • Geographical location • Tilt of the PV panel • DC-AC loss factor Cross Section of a PV Panel Batteries • Stores charge for night/ bad weather hours. • Lead Acid batteries a common storage option: cheap and time tested • Depth of discharge crucial in determining battery life. Number of cycles vs DOD for a deep discharge lead acid battery [9] Discharge-charge process and tolerable depth of discharge [10] Integrated Power Unit • Power requirements of a BS: transceiver equipment, cooling, miscellaneous loads (e.g. lights). • IPU: manages power supply to loads, and conversion and storage of the harvested solar energy • Power management unit: controls the charging of the batteries and the supply of power to the loads. • Battery charge regulator: monitors the battery state and disconnects them when level goes below a specified DOD (generally 50-80%). • The DC-DC converters: supply power to the transceiver and store power from the solar panels in the batteries • DC-AC converters: supply power to AC loads Current Deployment Scenario Solar Powered BS Deployment as of December 2014 Current Deployment Scenario • Bhutan Telecom Limited (BTL): • Solar powered BSs for use in rural areas, with ability to handle hundreds of users (in a range of few kms). • BSs require only between 50-150 W of power and have batteries designed for 3-7 day backup, aimed at providing autonomy during cloudy days. Current Deployment Scenario • Telkomsel Indonesia: • 234 solar powered BSs in 2012 • Example: BS in Sangatta • Average daily power consumption: 26 KW • Powered by 60 solar panels each with a DC rating of 205 W (giving a total rating of 12.3 kW). • 24 batteries each with rating 2000 Ah for autonomy of 4 days Case Study: Tigo Ghana • In 2012, 60% of the land area and 20% of the population (5 million people) of Ghana had no mobile coverage. • Primary reasons: (i) the lack of necessary infrastructure such as reliable grid power and (ii) too low average revenue per user (ARPU) to justify the deployment costs. • 2012: Tigo Ghana partnered with network solutions provider KNET and equipment manufacturer Altobridge to deploy 10 solar powered BSs • Compression techniques so that voice calls require rates of 4 kbps (compared to 14 kbps in conventional systems) • Cell site average power consumption of 90 W (compared to 130 W or more) Case Study: Tigo Ghana Case Study: Tigo Ghana • The BSs use satellites for backhaul, have a coverage range of 10 km, and capacity for up to 1500 subscribers. • Lowering of costs brought about by the design optimizations: return on investment for the operator in less than 24 months, assuming 600 subscribers with ARPU of $4 per month. • Currently there are has plans to expand to 300 additional sites, some of which have already been implemented. Challenges and Solutions • • • • Economic challenges Geographical limitations Resource provisioning and dimensioning Network management and resource allocation Economic Challenges • High CAPEX • CAPEX/TCO (total cost of ownership) ratio decreased by around 40% between 2009 and 2013 • Government initiatives such as subsidies • Market forces • Awareness of environmental issues • Government regulations • Large BSs • Powering a macro BS with power consumption of 3 kW would require an area of around 180 m2 for the PV panels. • However, larger BSs can still be cost effective, e.g. in the presence of government subsidies, though the payback period is still high (7-10 years). Geographical Limitations • Regions with Poor Solar Insolation • Solar power may be used in conjunction with the grid to power the BSs. • Urban Deployments: • PV panels should ideally be installed in open areas without shadows from obstructions due to buildings or trees • Difficult and expensive to procure such sites in urban areas. • Long Stretches of Bad Weather • Required size of the battery banks is very large • Increases the CAPEX and the possibility of outages during these periods. Resource Provisioning • The successful deployment of a solar powered BS requires meticulous dimensioning of the PV panels and backup batteries • Trade-off between CAPEX and outage • Dimensioning depends on • Solar irradiation profile • BS load profile • Desired outage probability System Resources • PV Panels • Number of PV panels: nPV • DC rating of each panel: Epanel • Overall DC rating: πππ€ = πππ πΈπππππ • Cost of PV panels: CPV ($/kW) • Batteries • Number of batteries: nb • Capacity of each battery: Ebat • Overall battery bank capacity: π΅πππ = ππ πΈπππ‘ • Battery lifetime: Lb • Cost of batteries: CBat ($/battery) BS Power Consumption • • • • • NTRX : Number of transceivers P0 : power consumption at no traffic. Δp : constant for a BS Pmax: power consumption at maximum traffic K: Normalized traffic 1400 BS power consumption (W) • LTE Base station ( ~ 0.6-1.4 kW) • Traffic dependent power consumption [7] 1200 1000 800 600 400 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Normalized traffic (K) 0.8 BS power consumption [7] [7] Auer, Gunther, et al. "Cellular energy efficiency evaluation framework."Vehicular Technology Conference (VTC Spring), 2011 IEEE 73rd. IEEE, 2011. 0.9 1 BS Traffic Calls are modelled as a Poisson process with rate dependent on time of the day and call duration exponentially distributed with mean of 2 minutes [8]. Normalized Traffic: πΎ π‘ = ππ‘ ππππ₯ 1300 0.9 Weekday 0.8 Base station power consumption (W) 0.7 Normalized traffic Saturday Sunday 1200 0.6 0.5 0.4 0.3 0.2 1100 1000 900 800 0.1 0 0 24 48 72 96 120 144 Hour Typical Normalized traffic for a sample week 168 700 0 2 4 6 8 10 12 14 16 18 20 22 Hour of the day Average BS power consumption on weekdays/weekend [8] P. De Melo, et al., “Surprising patterns for the call duration distribution of mobile phone users,” Machine Learning and Knowledge Discovery in Databases, pp. 354-369, Springer, 2010. Solar Data for 10 years Worst month for each year is selected Daily solar power generation is computed and days are sorted 1-α % Good weather days Find minimum and maximum values of power generated during a given hour Uniformly divide the region between minimum and maximum into 4 regions Calculate the average value of the solar power generated during each of hour for each of the regions from empirical data for good weather days For each hour calculate the probability of transition from each of the regions to each of the region in the next hour α% Bad weather days Modeling Solar Energy Resources • State of system • Daily transitions • Hourly transitions Battery Charge/discharge Dynamics • Battery charge level • Outage: Optimal System Dimensioning • Cost: • Number of batteries required over Trun years: • Optimization problem: Simulation Setup • LTE base station system • 10MHz Bandwidth • 2 × 2 Multi Input Multi Output (MIMO) • 3 sectors each with 2 transceivers (NT RX = 6) • Lead acid batteries. (12 V, 200 Ah) • Cost: $ 280/battery • PV panels • Cost: $ 1000/kW • Locations considered • Miami (USA) • Mumbai (India) • New Delhi (India) Results: Battery Lifetime vs Number of Batteries Number of batteries vs battery battery lifetime for the three locations for PV wattage of 14 kW Results: Battery Sizing Number of batteries required for a given outage for the three locations for PV wattage= 14 kW Results: PV Panel Sizing PV wattage vs number of batteries required for various outage probabilities for Mumbai. Deploying Small Cells • Small cells: • Reduced transmitter to mobile terminal distance • Higher data rates • Reduced transmit power requirement • Increasing network capacity and spectral efficiency Image: mwrf.com Deploying Small Cells • The main challenge: determine the number of BSs to deploy and their locations • Trade-off between outage probability and the number of BSs: • Preferable to have more small cell BSs with less energy harvesting (EH) resources rather than few BSs with larger EH resources [13]. • Approach: obtain the required number of small cell BSs keeping in mind only the desired outage probability, with other parameters(like the macro BSs and their location) kept as fixed. • The small cell locations are determined by factors such as the spatial distribution of traffic hotspots and solar insolation. Network Management and Resource Allocation • Resources available: • Energy harvested by the BSs • Transmission power level at which the BSs choose to operate • Spectrum available for transmission • Challenge: stochastic nature of the traffic intensity and solar insolation • Most widely explored problem: minimize the overall energy consumption of the network through a variety of mechanisms. • Existing methodologies for resource allocation and management consider both centralized and distributed mechanisms Load Balancing • Objective: preventing the BSs from running out of energy or being over-loaded • Constraints and factors: available energy, the expected harvested energy in the near future, and the traffic load at the BSs • Strategy: BSs cooperate by dynamically changing the area covered and traffic handled by each BS, in accordance to the energy available at each BS. • Two main techniques for load balancing among BSs: • Dynamic user association • BS beacon power control Energy and Delay Aware Downlink Power Control and User Association • Two key challenges while operating solar powered BSs: • avoiding energy outages • ensuing reliable quality of service (network latency). • Problem formulation: minimize network latency given the constrained energy availability at the BSs • System model: • Set of BSs B offering coverage to a geographical region R • User locations are denoted by π₯ ∈ π • Downlink transmit power vector of the BSs: P • Power level of BS j: π π ∈ 0, π, 2π, β― , ππππ₯ • Traffic arrival at location x: Poisson point process with rate π π₯ and an average file size of 1 π π₯ . Energy and Delay Aware Downlink Power Control and User Association • Assuming BS j is serving the users at location x, the rate offered by the BS to the users is ππ π₯ = π΅ππ log 2 1 + ππΌππ π (π₯) • BS load ππ (fraction of time BS j is busy) π π₯ ππ = π’π π₯ ππ₯ π π π₯ ππ (π₯) • Network latency indicator π·= π∈π΅ ππ 1 − ππ Energy and Delay Aware Downlink Power Control and User Association • Problem formulation 24 min πΈ,π,π π·π‘ π‘=1 subject to: π ∈ πΉ, ∀π 24 π‘=1 πΏπ‘ π ≤ πΊ π , ∀π ∈ π΅ • Approach: • Temporal energy allocation • BS downlink power control • User association reconfiguration Energy and Delay Aware Downlink Power Control and User Association • Temporal Energy Provisioning • Green energy is allocated to a given hour in proportion to the BS power consumption during that hour. • To avoid battery degradation, batteries are disconnected from the BS if the battery level goes below a certain threshold state of charge Energy and Delay Aware Downlink Power Control and User Association • Transmission power control • Non-convex optimization problem with respect to the BS power levels • Two factors affecting power levels for the BSs: • Avoid energy deficiency (i.e. π π > 1) • Avoid traffic overload at a BS (i.e. ππ > 1) • Strain index Ψ π = max 0, π π − 1 + max 0, ππ − 1 Energy and Delay Aware Downlink Power Control and User Association Energy and Delay Aware Downlink Power Control and User Association Energy and Delay Aware Downlink Power Control and User Association • User association • For any given set of BS power levels and green energy allocation find the optimal user association policy • BSs periodically measure their traffic loads and use it to determine their coalition factors which are advertised to MTs. • These coalition factors are used by the MTs to associate with the BSs so as to minimize the objective function. • The BSs and MTs update their association until convergence • Achieved by a transformation of the original problem min π πππ΅ subject to: π ∈ πΉ, π· ππ + π π ∀π ππ Energy and Delay Aware Downlink Power Control and User Association Energy and Delay Aware Downlink Power Control and User Association BS On/Off Strategies • Cellular networks are provisioned for peak-hour traffic: may be possible to turn off some BSs during off-peak hours • Strategies: • Determine the minimum number of BSs required to serve the area, with the desired quality of coverage as a constraint. • The switching decision may also take into account the energy availability of the BSs. • The problem of minimizing the overall energy consumption of a set of BSs, subject to a limit on the load on any BS, is known to be NP-complete. • Heuristics: greedily assigning MTs to BSs with higher loads so that the number of the BSs that have no associated MTs (and thus can be turned off) is maximized. Coordinated Multipoint (CoMP) • BSs cooperate to jointly serve MTs • Combat inter-cell interference in dense deployment scenarios • Enhancing network efficiency and overall QoS for users. • Implementation: • Form clusters of transmit points for CoMP transmissions and the allocation of resources to the transmit points. • Extent of cooperation and which BSs should cooperate to serve the MTs is decided based on the resources available at the BSs • Objective: maximize the system performance or to minimize the energy costs. • Cluster formation and resource allocation problems are tightly coupled and optimization problems to solve them jointly generally lead to non-convex formulations. Conclusions • Solar powered BSs are a viable solution for providing network coverage in areas without reliable grid power. • Solar powered BSs are expected to play a greater role in the future • Growing awareness of environmental issues and the push towards green engineering solutions • Technological advances in battery and PV panels • Open problems: • Network management • Resource allocation References [1]: Bogucka, Hanna, and Oliver Holland. ”Multi-Layer Approach to Future Green Mobile Communications.” Intelligent Transportation Systems Magazine, IEEE 5.4 (2013): 28-37. [2] Hasan, Ziaul, Hamidreza Boostanimehr, and Vijay K. Bhargava. "Green cellular networks: A survey, some research issues and challenges." Communications Surveys & Tutorials, IEEE 13.4 (2011): 524-540. [3]: http://www.trai.gov.in/ [4]: http://www.gsma.com [5]: http://solargis.info/doc/free-solar-radiation-maps [6] http://www.topsunenergy.net/solar-telecom-system.htm [7] Auer, Gunther, et al. "Cellular energy efficiency evaluation framework."Vehicular Technology Conference (VTC Spring), 2011 IEEE 73rd. IEEE, 2011. [8] P. De Melo, et al., “Surprising patterns for the call duration distribution of mobile phone users,” Machine learning and knowledge discovery in databases, pp. 354-369, Springer, 2010. [9] http://www.usbattery.com/ [10] http://quizlet.com/19225890/ree-554-ch-06-batteries-flash-cards/ [11] Vinay Chamola and Biplab Sikdar “Resource provisioning and Dimensioning for Solar powered Cellular Base Stations,” Proc. IEEE GLOBECOM, Austin 2014 Back up slides 51 National University of Singapore (NUS) Key Components Solar Panels -Mono/ Polycrystalline silicon panels -Efficiency: 16-18 % -Cost: $ 1000/ kW -Dimension: 1 kW: 5 m^2 -Life Duration: 25-30 years Fig 8: A PV 250 W PV panel (~.25 m^2) [12] [8] http://www.thesolarbiz.com/Yingli-245-Watt-Poly-Solar-Panel_2#gsc.tab=0 52 National University of Singapore (NUS) Factors effecting energy generated by DC Ratingpower Solar Geographical Location Fig 9: Solar map of world [6] Tilt DC-AC loss factor Fig 10: Tilted solar panel [9] [9] http://www.volker-quaschning.de/articles/fundamentals1/index_e.php 53 National University of Singapore (NUS)