Design, Modeling, and Capacity Planning for Micro-Solar Power

Sensor Networks

Jay Taneja, Jaein Jeong, and David Culler

IPSN/SPOTS 2008 – 4/23/2008

Applications and Science

Zebras in the savanna

People tracking

Bird Nest Monitoring

However, publications report that scientific results have been mostly underwhelming.

Why?

What Does It Mean To Be Systematic?

Need to develop models for planning of every subsystem

Must formalize application-driven constraints

Real science demands real engineering

If one thing fails – everything can fail

HydroWatch: Water in Motion

The Challenges

Capacity Planning for Microsolar

Mechanical Design - Weatherproof with Correctly Exposed Sensors

Incorporating off-the-shelf and custom-built pieces

Storage Charge-Discharge 72:1120:1240:11:1All Ideal Components 3:18 Hours of Sun Per Day 48:1Half Hour of Exposure Per Day

Components of a Solar Circuit

Regulator Efficiencies

E in : E out

66%

2%

60%

50%

Application LoadStarting point for capacity planning

Most time is spent sleeping (~20 uA) with short active periods (~20 mA)

Color

Device Average Current

Sensors 9 uA (550 uA at 1.67% DC)

Radio 0.206 mA (20.6 mA at 1% DC)

MCU 9.6 uA (2.4 mA at 0.4% DC)

Quiescent

15 uA

Total 0.24 mATotal Current

Energy Storage

Type Lead Acid

NiCad NiMH Li-ion Supercap

Operating Voltage Range

5.0-6.1V 0.8-1.35V 0.9-1.4V 3.0-4.2V 2.2-3.0V*

Volume Energy Density

67 Wh/L 102 Wh/L 282 Wh/L 389 Wh/L 5.73 Wh/L

Charge/Discharge Efficiency

70-92% 70-90% 66% 99.9% 97-98%

Charging Method

Trickle Trickle/Pulse

Trickle/Pulse

Pulse Pulse

Est. Lifetime (79.2

mWh/day)

98.5 days

33.3 days (2)

75.8 days (2)

35.4 days

3.8 days* estimated

Solar Panel

Cells in series and parallel

Important parametersIV and PV CurvesPhysical Dimensions

MPP: 3.11 Volts

Regulators

Regulators are “glue” matching primary components

50-70% efficiency for typical sensornet load range

Input regulatorRegulates voltage from solar panel to batteryCan be obviated by matching panel directly to storage

Output RegulatorRegulates mote voltageProvides stability for sensor readings

Model estimates that load requires 28 minutes of sunlight

Mechanical Considerations

Enclosure design is often application-driven

Sensor ExposureWaterproofingEase-of-DeploymentInternal Mechanicals

Temp / RH Sensor TSR, PAR Sensors

Network Architecture

Used Arch Rock Primer Pack for multi-hop network stack, database for stored readings, and web-based network health diagnosis

The Urban Neighborhood

20 Nodes for 5 Days

Mounted on house, around trees, and on roof

Meant to emulate forest floor conditions

Important for systematic approach -- provided validation of model

Urban Neighborhood Energy Harvested

Every node received enough sunlight

Three Nodes, Three Solar Inputs

The Forest Watershed19 Nodes for over a Month

Mounted on 4-ft stakes throughout the transsect

Forest Watershed Site

Forest Watershed Energy Harvested

Watershed

Most nodes struggle to

harvest sunlight

Three Nodes at the Watershed

Reflected Light

Though only minimally, a cloudy day helps a sun-starved node harvest solar energy.

Sunny

Overcast

Overcast

Sunny

Results and Conclusions

Open issues not captured in modelBattery sizing for days or for seasons?Solar spotting effectEnclosures with holes?

Systematic approach resulted in 97% collection of an unprecedented spatiotemporal data set

New view of watershed microclimates

Questions?

Outline

Microsolar Model

Sizing the Pieces

Network and Node Architecture

Deployment Results

Daily Solar Model

• Discharge: No incoming solar energy, battery discharging• Transition: Some incoming solar energy, battery still discharging• Recharge: Incoming solar energy for charging battery• Saturation: Incoming solar energy shunted, battery full

Application Load

Most time is spent sleeping (~20 uA) with short active periods (~20 mA)

Variation in hardware results in RMS currents from 0.25 mA-0.6 mA. Empirically measured 0.53 mA.

Device Average Current

Sensors 9 uA (550 uA for 5s in 5m)

Radio 0.206 mA (20.6 mA at 1% DC)

MCU 9.6 uA (2.4 mA at 0.4% DC)

Quiescent 15 uA

Total 0.2396 mA

Node Voltages

HydroWatch Node

A Tale of Two Deployments

Urban Neighborhood Deployment20 Nodes for 5 DaysMounted on house, around trees, and on roof to emulate conditions on forest floorProvided validation of model

Forest Watershed Deployment19 Nodes for over 1 MonthMounted throughout forest on 4-ft stakesRainy season

Network Diagram

Conclusions and Future Work

Good Engineering Needed to Undertake Good Science

Microsolar has subtle differences from mesosolar

Enclosure design requires careful engineering

Experimenting with flexible panels and application duty cycling

Solar Voltages

Solar energy is available