7.3 Class BatteryProfileHelper
7.3.1 Class BatteryProfile
This class oers a generic container for battery parameters in terms of performance and chemical compositions:
RechargeCyles: the number of maximum recharge cycles of the battery which is >1 for Li-Ion batteries.
Weight: the weight in grams of the battery which is used as a reference to estimate the average amounts of dierent chemical substances in the battery.
Cost: the price of the battery,which usually varies in relation to durability (i.e. recharge cycles), or energy eciency (i.e. amperes per mg).
InstallationCost: the cost of installing/re-charging the battery initially and between recharge cycles.
EnergyCapacity: the watt-hour capacity of one recharge cycle of the battery.
Cost_per_wh: is a computed parameter as a function of the previous parameters.
<Chemicals%>: is a vector containing the percentages of dierent chemicals in the batteries:
[Lead, Cobalt, Lithium, Nickel, Thallium, and Copper]. Percentages are used for rechargeable Li-Ion batteries as obtained the research in [18].
This class implements the following methods:
Compute_wh_capacity: which computes Cost_per_wh depending on battery parameters.
Compute_human_toxicity, Compute_terrestrial_toxicity, and Compute_water_toxicity:
which computes human toxicity, terrestrial toxicity and water toxicity factor as described in [18]. An advance development of this section is if the waste recycling region is known (in terms of area and location), then it can allow estimation of the expected toxicity density in the waste destination area.
Figure 27: Simulation Framework Outline
Figure 28: Simulation Framework Class Diagram
8 Results
Cost of reference architecture is outlined in table 2. In gure 29 we can see the estimated minimum required back haul throughput on each GW for the backbone network. Varying GW density proved to contribute to network OpEx, as expected, by allowing use of less expensive link congurations to account for radio path loss. Deployment of nine GWs instead of ve introduced comprehensive enhancement of link OpEx as illustrated in CDF plot in gure 31. It also improved network general performance on all nancial, environmental, and technical metrics as illustrated in gure 30 of normalized metrics. Precisely, this coverage enhancement created saving near
¿12K of OpEx,27 kWh of energy,46years of ToA,1300years of battery life cycle duration, and 300grams of chemical waste. However, network became more saturated at nine GWs density as performance is not increased signicantly with deployment of thirteen GWs, except of increased ToA and decreased OpEx Waste.
OpEx Net 388,301.20 ¿
OpEx Sensing 30,186.38¿
OpEx Waste -48.67567531 ¿ User Trac OpEx 328, 271.9109 ¿ Management Trac OpEx 60,029.29 ¿
Energy 167.902 kWh
ToA 80.5 years
Battery Age 9737.5 years Chemical Waste 1707.3 gms Table 2: Cost of reference architecture for 365 Days
Figure 29: Estimated Minimum Required Back-haul Throughput on Each Gateway
Figure 31: CDF of link OpEx in the network for dierent coverage densities
Figure 30: Impact of coverage density on network budget
In the second set of experiments, ED conguration showed to contribute signicantly to network OpEx as illustrated in link OpEx CDF in gure 33 and to general network performance as shown in gure 32. We can observe that increasing packet inter-arrival time to 120 seconds instead of 60 seconds introduced the highest impact of improvement to the network on all metrics such as savings of ¿179K in OpEx,27kWh of energy, more than40years of ToA, and nearly600 gms of chemical waste . Similarly, cutting sensing sampling rate to be every two hours instead of one introduced signicant energy saving of ¿15K in OpEx. Reducing packet size by 10 bytes by omitting the timestamp, for example, created OpEx saving of almost ¿30 K.
Figure 32: Impact of ED conguration on network budget
Figure 33: CDF of link OpEx in the network for dierent ED congurations
Figure 34: Impact of battery quality on network budget
In the third set of experiments, we observe signicant patterns in the impact of battery characteristics on network metrics. Lowest durability batteries showed to reduce network OpEx as in gure 35 which saved OpEx of nearly ¿ 34 K. However, it leaves higher chemical waste and much shorter battery life-cycle age of the network as in gure 34. On the other hand, high durability battery shows better performance, as expected, in battery age in years and in reduced chemical waste, which comes at the expense of increased OpEx for the entire network, OpEx of sensing activities, and OpEx waste.
Figure 35: CDF of link OpEx in the network for dierent battery eciency levels
In the last experiment, we show network metrics compared to NB-IoT deployment based on estimated theoretical EPB and bit ToA for NB-IoT [24, 28]. The impact of throughput on total network ToA is clearly observed compared to LoRa in gure 36. It is also clear that low robustness of NB-IoT modulation compared to LoRa results in much higher OpEx waste due to packet loss (without considering re-transmits).
Employment of ILP optimization for minimizing network OpEx proved to play a signicant role in minimizing network OpEx while inducing sometimes unpredictable side-eects. For in-stance, using homogeneous battery deployments (whether high or low durability) resulted in additional ToA cost of 2.22 years in the network, and using homogeneous NB-IoT deployment resulted in lower energy consumption yet total lower battery life time duration in the network.
This is because using heterogeneous battery deployments or PHY congurations creates a gap between OpEx of the nodes in the network which causes network links of low durable batteries
Figure 36: Impact of LoRa vs. NB-IoT
to be much cheaper than that of high durable batteries even at higher spreading factor. In this case, ILP assignment model gives priority of optimal link assignment to EDs with more expensive OpEx to ensure they are connected to closest GWs while other EDs can still have cheaper links even if they are assigned to further GWs, as illustrated in gure 37, resulting in dierent link ToAs or battery ages from homogeneous deployments where all nodes have the same link OpEx per Joul. In this particular application, the savings of ToA were more than the additional ToAs in the basic conguration, resulting in lower network ToA in total.
All experiments showed interesting savings patterns from the default basic architecture as listed in table 3. Detailed estimations are available at Appendix 9. Improved signicant energy consumption savings were achieved at all experiments, reaching up to 56 kWh by just cutting packet rate to be every two minutes instead of one. Similarly,31 kWh of savings were achieved by enhancing coverage without changing anything of the network or the battery congurations.
Deployment\SavingsOpEx(¿)ToA(years)BatteryAge(years)ChemicalWaste(grams)Energy(W.h) LowDurabilityBattery33,771.342.223725.18-315.981239.73 HighDurabilityBattery(35,167.70)2.22-3878.70367.931239.73 HalfSensingFrequency15,093.190.001670.44284.8228278.45 HalfPacketRate179,057.4140.253198.33568.8655672.80 NoTimestamp29,842.906.71533.0694.819278.80 DenseCoverage11,835.6246.771311.03308.9227743.51 ExtraDenseCoverage14,184.3453.571570.89347.0831743.48 NB-IoT(theoretical)19,267.1367.202,133.12419.5739,741.84 Table3:Savingsofeachexperimentalarchitecturecomparedtobasiccongurationdeployment
Figure 37: Impact of ILP model ED assignment in heterogeneous battery deployment
9 Discussion and Conclusion
From a more comprehensive sustainability point of view, we describe how our proposed frame-work contributes to organizational sustainability in a more holistic perspective. Based on the framework proposed in [29], we show proposed sustainability analysis of our integrated budget and environmental models in gure 38. From sustainability point of view, we can highlight in-teresting insights from based on results in table 3. From economic sustainability perspective, a simple change such as removing timestamp saves nearly ¿30K which is equivalent of providing two minimum wage jobs in France at ¿1498,47/month [30]. Similarly, cutting packet rate into half (by transmitting every two minutes instead of one minute) saves nearly ¿180K which is equivalent of providing ten jobs. Also, cutting sensing frequency to half, by sampling every two hours instead of every one hour saves nearly ¿15K which is equivalent of providing one job.
From environmental sustainability perspective, we can see signicant energy savings reaching up to 56 kwhs by cutting packet rate and nearly 30 kwh by adapting some minor network features such as adding additional four GWs or cutting sensing sampling frequency to half. This is a signicant observation since it shows that network radio activity is not the only signicant area of optimization for LPWAN development.
Therefore, we achieved the following aforementioned contributions:
We formalized an OpEx model and environmental model for LPWAN architectures that estimates total network costs and solid waste footprint considering any possible underlying architecture setup or technology, thus allowing objective evaluation of LPWAN architec-tures. We demonstrate veried performance of the model by experimenting on large-scale simulation of a realistic setup.
We propose an Integer Linear Programming model that is proven to nd optimal ED to GW link assignment solution with global minimal OpEx in the network, regardless of network size or ED heterogeneity.
We show that algorithmic complexity's impact on IoT node processing time can be poten-tially negligible compared to input size. We show that in a certain general case of program complexity, time and energy cost per input element approaches a constant value as in-put size increases unless modular programming is heavily used. Therefore, there is strong
Figure 38: Sustainability Analysis of Integrated Model
potential for shifting computations to EDs but with careful use of modular programming paradigm such as object orientation.
We demonstrate that in how signicant QoS improvement as well as budget and envi-ronmental savings can be achieved without changing transmission conguration. Also, signicant budget and environmental savings can be achieved through minor network con-guration such as removing a timestamp, adding few GWs, or cutting down sensor sampling frequency.
Interesting future work based on this thesis can be driven from the need to estimate the budget from operators perspective which would include the estimation of the OpEx allocated for back-haul infrastructure. Such OpEx might include parameters such as the energy cost, number of subscribers, or service capacity used for IoT services in base stations. Furthermore, such
framework can be used to estimate the economic eciency of network MAC negotiation schemes or development of economy-aware network topology design processes.
In conclusion, the research presented in this paper veries and validate a budget model for LPWAN architectures which allows to quantify the real nancial, environmental, energy, and time costs of a dense LPWAN deployment. According to the principle of separation we are able to benchmark, experimentally, dierent components of a sensor system to understand its behavior in energy space and time space. After translating our benchmarks into an OpEx model, we are able to observe that signicant OpEx savings reaching thousands of euros and energy savings reaching tens of kWhs can be achieved through simple network updates such removing a timestamp from the packet payload, slightly reducing sensing sampling or packet rate, or just by introducing few GWs for better coverage.
References
[1] J. Porras, A. Seah, E. Rondeau, K. Andersson, and A. Klimova, PERCCOM : A Mas-ter Program in Pervasive Computing and COMmunications for Sustainable Development PERCCOM : A Master Program in Pervasive Computing and COMmunications for, in 29th IEEE Conference on Software Engineering Education and Training, no. APRIL, Dal-las, United States. 2016., 2016.
[2] G. Pasolini, C. Buratti, L. Feltrin, R. Verdone, and O. Andrisano, Smart City Pilot Projects Using LoRa and IEEE802.15.4 Technologies, pp. 117, 2018.
[3] J. Petäjäjärvi, K. Mikhaylov, A. Roivainen, T. Hänninen, and M. Pettissalo, On the coverage of LPWANs: Range evaluation and channel attenuation model for LoRa technology, 2015 14th International Conference on ITS Telecommunications, ITST 2015, pp. 5559, 2016.
[4] T. Petri¢, M. Goessens, L. Nuaymi, L. Toutain, and A. Pelov, Measurements, performance and analysis of LoRa FABIAN, a real-world implementation of LPWAN, IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, PIMRC, pp. 17, 2016.
[5] M. Bor, J. Vidler, and U. Roedig, LoRa for the Internet of Things, in Proceedings of the 2016 International Conference on Embedded Wireless Systems and Networks, ser. EWSN '16.
[6] K. E. Nolan and M. Kelly, An Evaluation Of Low Power Wide Area Network Technologies For The Internet Of Things, pp. 439444, 2016.
[7] A. Augustin, J. Yi, T. Clausen, and W. Townsley, A Study of LoRa: Long Range & Low Power Networks for the Internet of Things, Sensors, vol. 16, no. 9, 2016.
[8] L. Gregora, L. Vojtech, and M. Neruda, Indoor Signal Propagation of LoRa Technology, 2016 17th International Conference on Mechatronics - Mechatronika (ME), pp. 1316, 2016.
[9] S. Hosseinzadeh, H. Larijani, K. Curtis, A. Wixted, and A. Amini, Empirical propagation performance evaluation of LoRa for indoor environment, Proceedings - 2017 IEEE 15th International Conference on Industrial Informatics, INDIN 2017, pp. 2631, 2017.
[10] P. Neumann, J. Montavont, and T. Noel, Indoor deployment of low-power wide area net-works (LPWAN): A LoRaWAN case study, International Conference on Wireless and Mo-bile Computing, Networking and Communications, 2016.
[11] L. H. Trinh and V. X. Bui, Signal Propagation of LoRa Technology Using for Smart Building Applications, pp. 381384, 2017.
[12] K. Mikhaylov, J. Petäjäjärvi, M. Hämäläinen, and J. Iinatti, Evaluation of LoRa LPWAN technology for remote health and wellbeing monitoring, 2016 10th International Symposium on Medical Information and Communication Technology (ISMICT).
[13] D. M. Hernandez, G. Peralta, L. Manero, R. Gomez, J. Bilbao, and C. Zubia, Energy and coverage study of LPWAN schemes for Industry 4.0, Proceedings of the 2017 IEEE International Workshop of Electronics, Control, Measurement, Signals and their Application to Mechatronics, ECMSM 2017, 2017.
[14] F. Adelantado, X. Vilajosana, P. Tuset-Peiro, B. Martinez, J. Melia, J. Melia-Segui, T. Wat-teyne, and J. Melia, Understanding the limits of LoRaWAN, pp. 812, 2016.
[15] D. Magrin, Network level performances of a LoRa system, Universit{à} degli Studi di Padova, Tech. Rep., 2016.
[16] M. Al., LP-WAN Problem statement. [Online]. Available: https://tools.ietf.org/html/
draft-minaburo-lp-wan-gap-analysis-00
[17] L. Li, Product Specication 18650 3.7v 6600mAh, Hunan Sounddon New Energy Co., China, Tech. Rep.
[18] M. Chen, Potential Environmental and Human Health Impacts of Rechargeable Lithium Batteries in Electronic Waste, no. MAY, 2013.
[19] Semtech, LoRa Modem Design Guide, no. July, pp. 19, 2013.
[20] K. Mekki, E. Bajic, F. Chaxel, and F. Meyer, A comparative study of LPWAN technologies for large-scale IoT deployment, ICT Express.
[21] Atmel Corporation, Atmel V-1280 /V-2560, Tech. Rep., 2014.
[22] Semtech Corporation, Sx1272/73 Datasheet, Tech. Rep. March, 2017.
[23] 3rd Generation Partnership Project (3GPP), Standardization of NB-IOT completed, 2016.
[Online]. Available: http://www.3gpp.org/news-events/3gpp-news/1785-nb%7B_%7Diot%
7B_%7Dcomplete
[24] R. Ratasuk, N. Mangalvedhe, Y. Zhang, M. Robert, and J.-p. Koskinen, Overview of Nar-rowband IoT in LTE Rel-13, 2016.
[25] Adafruit Industries, Micro SD Card Breakout Board Tutorial, 2017.
[26] Île-De-France, Qualité de service de la téléphonie mobile sur les lieux de vie. [Online].
Available: https://data.iledefrance.fr
[27] C. Phillips, D. Sicker, and D. C. Grunwald, The Stability of the Longley-Rice Irregular Terrain Model for Typical Problems;CU-CS-1086-11, Tech. Rep., 2011. [Online]. Available:
https://scholar.colorado.edu/csci{_}techreports
[28] R. Ratasuk, B. Vejlgaard, N. Mangalvedhe, and A. Ghosh, NB-IoT System for M2M Com-munication, no. Workshop on Device to Device communications for 5G NETWORKS (, pp.
26, 2016.
[29] C. Becker, Requirements : The Key to Sustainability.
[30] Ministry of Labour/Employment/Vocational Training/ and Social Dialogue, Interprofes-sional minimum wage (SMIC) in 2018 Data from 1980 to 2018, Tech. Rep., 2018. [Online].
Available: https://www.insee.fr/en/statistiques/2417640
Appendix I: Detailed Architecture Budget Estimations
OpExNetOpExSensingOpExWasteOpExUser BasicCongurationDeployment388301.197130186.38521-48.67567531328271.9109 LowDurabilityBatteryDeployment354529.854218394.84371-29.95569634308123.7597 HighDurabilityBatteryDeployment423468.899641789.26204-68.02344084349873.0011 HalfSensingFrequencyDeployment373208.004515093.1926-48.67567531328271.9109 HalfPacketRateDeployment209243.791230186.38521-24.33783766164135.9555 NoTimestampDeployment358458.296130186.38521-44.61408976298429.0099 DenseCoverage(9GWs)376465.57730186.38521-27.50924626317422.5925 ExtraDenseCoverage(13GWs)374116.856330186.38521-22.85487844315269.5985 BasicCongurationNBIoT368821.174830186.38521-36.23682165310415.2238 ILPMinChemicalWaste390778.074530186.38521-51.74400589330542.3819 ILPMaxQoS391389.890730186.38521-44.93609871331103.2134 NBIoT368821.174830186.38521-10.54598468310415.2238 Table4:FullOpExBudgetMetricsofAllArchitectures
Energy DuringOpex Period
TimeOnAir
Battery Exp ected AgeYears
Batteries DuringOpex Period BasicCongurationDeployment167902.495425388897299737.543053207461.2175 LowDurabilityBatteryDeployment166662.763824688382286012.365218267961.7545 HighDurabilityBatteryDeployment166662.7638246883822813616.23888155685.2299 HalfSensingFrequencyDeployment139624.047425388897298067.10232257195.6732 HalfPacketRateDeployment112229.695712694448656539.212259301343.602 NoTimestampDeployment158623.695423273155859204.487921218729.2906 DenseCoverage(9GWs)140158.983410640815148426.515222244131.3108 ExtraDenseCoverage(13GWs)136159.0152849436078.48166.650824248642.4342 BasicCongurationNBIoT128336.27475005393.927580.972645254848.9419 ILPMinChemicalWaste166709.2678247244474910010.49908213453.5714 ILPMaxQoS170092.8336264160188210078.66161208543.0419 NBIoT128336.27475005393.927580.972645254848.9419 Table5:EnergyandRadioMetricsofAllArchitectures
Human ToWater xicityToxicity Terrestrial TCobaltCopperAluminum oxicity Chemical W
aste BasicCongurationDeployment353.41399.86404.27428.18364.49805.171707.36 LowDurabilityBatteryDeployment418.81473.86479.08507.43431.94954.182023.33 HighDurabilityBatteryDeployment277.25313.69317.15335.91285.94631.661339.43 HalfSensingFrequencyDeployment294.45333.15336.83356.76303.68670.851422.54 HalfPacketRateDeployment235.66266.63269.57285.52243.05536.901138.50 NoTimestampDeployment333.78377.65381.82404.41344.25760.461612.55 DenseCoverage(9GWs)289.47327.51331.12350.71298.54659.491398.44 ExtraDenseCoverage(13GWs)281.57318.57322.08341.14290.39641.491360.28 BasicCongurationNBIoT267.56302.72306.06324.17275.94609.571292.60 ILPMinChemicalWaste344.53389.81394.11417.42355.33784.931664.45 ILPMaxQoS354.03400.57404.98428.94365.13806.591710.38 NBIoT267.56302.72306.06324.17275.94609.571292.60 Table6:FullEnvironmentalWasteMetricsofAllArchitectures