The potential of successful PV mesh networks is based on five variables: system design and sizing, solar energy availability, economic analysis, and integrating geospatial information to optimize deployment locations. An example system was designed and built to assess the economic feasibility and overall system design. Utilizing PVSyst (refs) and ArcGIS (refs) technology, energy availability and demand was calculated for the system. Further analysis was done on the economic potential of the system compared to comparable technologies using previously published works (refs) and geographical referencing.

Locations of the study were chosen for their differences in solar energy availability and costs associated with Internet infrastructure implementation. Additionally, public or educational access to GIS data needed to be available. Therefore, the three locations chosen were in North America and compared townships in California, United States; Kentucky, United States; and Ontario, Canada. Each area has unique demographics, solar and infrastructure availability.

Design Methodology[edit | edit source]

The following design methodology was created, employed and demonstrated in a prototype system. The system initially includes one unit connected directly to a broadband source to demonstrate the feasibility of using ultracapacitors as the only energy storage devices. In the future, two other units will be constructed to desmonstrate thier functionality acting as relays to the initial unit. It was decided that the system use only ultracapacitors for energy storage to demonstrate their potential in such an application and the feasibility of implementing such a system. The sizing of the electronic components was determined using PVSyst. The associated converters and controllers were selected to optimize efficiency while minimizing cost and necessary components.

Mechanical components including the structural elements, mounting hardware, and fasteners were selected based upon the following criteria: cost, strength, component life, weight, transportability, ease of use, and maintenance. The structural support system was designed to hold the solar array at the optimum angles of 30ͦ and 60ͦ for the latitude at which the system will be demonstrated. The layout of the mechanical components was designed using a rapid prototyping approach. Using this method, parts were kept to a minimum while the focus was placed on function and simplicity. Ultimately, the design shown below prevailed and can easily switch between the optimum summer angle of 30ͦ and the optimum winter angle of 60ͦ. The entire support structure was constructed for roughly $30.

Sizing of system parts[edit | edit source]

Equipment list inclues: PV panel, ultracapacitor bank, router, converters, batteries, hardware - this will be based on either PVSyst or GRASS/rsun in ArcGIS.

A micro-controller was chosen to control the converters and was selected to be XXX. Using PVSyst, it was determined that the system required a 13Ah capacity battery and a 190 Wp capacity solar panel. This would provide enough power and store sufficient energy for nights and low light situations. An ultracapacitor bank of 10 3000F is being used as the storage. The size of Ultra cap bank is 16.84 Ah and each Ultra capacitor is 2.25 Ah. The results of sizing from PVSyst obtained are as 13Ah capacity battery, 92Wp capacity solar panel, Generalizing the design, a 10 ultracapacitor bank of 3000F each is being selected as the storage, and a 190Wp solar panel is used.

Going with an autonomy of 1 day and a maximum Loss of Load of 5% we can arrive at the size of the system as follows: Ultra cap size = Total Wattage * number of hours *allowed autonomyDepth of Discharge*Efficiency*Operating voltage The router chosen is a Cisco Linksys WRT54-- 6W. The Cisco router has specifications of: DOD=0.98, Efficiency=0.95, and an operating voltage of 9V.

The frame is constructed of 2x4s for the initial prototype with a field quality prototype being constructed from 1in (1/16th in) aluminum angles. A simple plastic storage container functioned as a waterproof housing to protect all necessary electronic components. This was done using a simple container and waterproofing it with silicone glue on the seams. The chosen container is a Sterilite 25-quart modular latch box. A rack to hold the ultracapcitors in the waterproof housing was designed using UniGraphics NX 7.5 and printed using a reprap machine. The profile for the part can be seen below. The final product was printed to a thickness of 1/2in.

Reprap part.jpg

The.stl file can be found here – Maxwell 3,000 Farad ultracapcitor holder.

Each mechanical component was modeled using UniGraphics NX 7.5 and is shown in the image below.

Material properties for each mechanical component were assessed to ensure all components could withstand the loading conditions including wind loads and snow loads (PV Systems Engineering 2006). The wind load was calculated to be 55psf using the following equations.

Velocity pressure (q) 	= 0.00256*Kz*Kzt*Kd*V2*I

Where Kz = celocity pressure exposure coefficient at height z Kzt = topographical factor Kd = wind direction factor V = basic wind speed in mph I = importance factor

Design Wind Pressure (p) = q *G*Cf						

Where G = gust effect factor = 0.85 Cf = force coefficient = 0.7

The snow load was assumed to be less than 8 psf as is recommended in Photovoltaic Systems Engineering by Roger Messenger and Jerry Ventre. This yields a combined load force of 71 psf. Each mechanical component was then sized using a factor of safety of 2 as is recommended by the American Society of Civil Engineers. The factor of safety was determined using the following equation.

Factor of Safety = Factor of Safety = Material Strength / Design Load

Using 6061 aluminum angles with a yield strength of 34,000 and a max stress of 16,600 psi under the design loading conditions results in a factor of safety of 2.

A finite element analysis was also performed on the frame members using Abaqus 6.11. The analysis confirmed that 1in aluminum angles to be used on the field testing prototype will withstand the loading conditions while accounting for the safety factor of 2. A tetrahedral mesh was used and seeded every 20 mm along the edges as can be seen in the picture below.


The boundary conditions set secured the bottom of the frame to the roof preventing and displacement or rotation. The applied load was placed on the frame of the module which is directly attached to the support structure itself. The applied load of 71 psf was distributed around the module frame. The boundary and loading conditions for each position can be seen below.

Boundary and loading conditions at the 30ͦ tilt angle.

FEA - 30 - bc and loading.jpg

Boundary and loading conditions at the 60ͦ tilt angle.

FEA - 30 - bc and loading.jpg

The stress and deflection distribution for the 30ͦ tilt angle can be seen in the pictures below.

FEA - 30 - stress analysis.jpg

FEA - 30 - deflection analysis.jpg

The stress and deflection distribution for the 60ͦ tilt angle can be seen in the pictures below.

FEA - 60 - stress analysis.jpg

FEA - 60 - deflection analysis.jpg

The follow are a list of all mechanical components:

  • Wood – 16 feet of 2x4 and 8 feet of 2x2 for frame
  • Fasteners
  • (x10) M6 40mm in length with correstponding nuts and washers to mount PV module

Part no. 91280A152 -

  • (x24) 1/2in hex nuts to secure aluminum ultracapacitor connectors to ultracapacitors

Part no. 93827A245 -

  • (x9) 1 inch aluminum bar, 18th in thick by 51/2in in length to connect ultracapacitors

Part no. 8975K17 –

  • Electronics housing – Sterilite 25-Quart Modular Latch Box

Hardware Implementation[edit | edit source]

The implementation of the electronic hardware is going to be made using DC/DC converters as shown in the schematic shown below.


The firing of the MOSFET's is controlled by the Arduino Uno and the control code is available online. The incremental inductance algorithm is used for MPPT control as it is It is easy to implement, robust in structure, and has efficiencies >97% in operation

The MPPT algorithm is implemented as shown below in the flow chart.


The Arduino Uno micro-controller was chosen to control the converters. Using PVSyst, it was determined that the system required a 24Ah capacity battery and a 190 Wp capacity photovoltaic module to provide enough power during operation and store sufficient energy for nights and low light situations. An ultracapacitor bank of 12, 3000F of Maxwell make (BCAP3000-P270-K04) ultracapacitors with 2.25 Ah each, is being used as the storage and provides 27Ah total. The router chosen is a Cisco Linksys WRT54, because it is among the least power consuming routers, drawing only 6W.

We can also calculate the sizing using simple calculations as shown below (Equation 1) the total Ultra capacitors required is around 11, but to balance the voltage across the series connected capacitors we are taking 12 with 4 ultracapacitors in series and three such series connected arrays in parallel. The ultracapacitor has specifications of DOD=0.97, Efficiency=0.95[24], and an operating voltage of 9V. The router chosen is a Cisco Linksys WRT54 drawing 6W. Now considering the losses in the converters a total load of 8.5W (as the boost converter efficiency is usually in the range of 70-80%) can be considered. Going with autonomy of 1 day and a maximum Loss of Load of 5% we can arrive at the size of the system as follows:

The electric system was simulated in MATLAB/Simulink (, taking into consideration the real time parameters of the PV. The Design of Ultracapacitor bank is done as follows:


A resistor is placed in series with the capacitor bank in order to limit the current flowing into the bank initially as it acts as short-circuit. During the time of discharge the resistance gives rise to voltage drop and there is considerable power loss across it so a Schottky diode is placed in parallel to it which has a forward voltage drop of 0.25V only and acts as a short circuit during discharge and open circuit while charging. The 24 hour simulation has been done on a simplified model of the system where the components are modeled mathematically. The modeling of ultracapacitor bank in Simulink is done as follows:


The voltage across the capacitor bank and there by the SOC for 24 hour period is as fol-lows:


Economic viability[edit | edit source]

The present cost of bringing broadband Internet access to rural areas was compared to the implementation of PV mesh networks. A Comparison of router configuration was also done in order to optimize energy demand, with system configuration and antenna spacing. Finally, a comparison was made between the use of standalone PV systems and connecting areas to electric grid in distance from current availability. In order to perform the economic analysis the cost of stacking standalone systems to a rural area from a WLAN hub, a comparison to the cost of extending electric and broadband infrastructure was made. The total cost of the system was calculated to include solar panels, inverter, batter bank, router, mounting, installation, and travel to rural areas (McLaughlin et al. 2010). Additionally, connection costs and WLAN hub was also included but varies with distance from land lines.

Lots more details needed here -- think equations for the comparison.

Geographical potential using GIS[edit | edit source]

In this section we will use Geographical Information Systems (GIS) to identify optimal areas for the PV mesh network deployment. Information on obtaining a copy of ArcGIS can be found at [1]. In order to assess variables influencing PV system implementation, a geographical representation of deployment potential is useful in identifying optimal locations. Local level spatial disparities are necessary to locate places lacking infrastructure and technology for broadband access (Grubesic and Murray 2001).

Due to the potential of emerging technology and because of its relatively high solar irradiance availability, California was used as an initial case study. Among the geo-processing capabilities of GIS perhaps the most useful is the overlay function. GIS has shown its usefulness in rural and urban planning, sustainable development, and identifying site specific locations. To do this we will mainly look at population density and current broadband access.

Getting Started[edit | edit source]

First, base layers are needed for creating a quality map. Many free shape files can be found online. For our map we used California Basemap [2] and a general basemap for the United States and Pacific Ocean [3]. After unzipping the files, open ArcMap and 'add data'. Select the shapefile you would like to use. If one of the files has labels, for example state names, you can double click on the file and select the 'Display' tab and select the field that contains names next to the 'Field' drop down window. The window should look like this.

Basemap 1.jpg

Next we need to download and add to the map county boundaries, which were found here: [4] These files are in the same geographical format as census data which will be helpful in the next step. After uploading the map will look like this.

Boundaries 2 resize.jpg

Now we need to find census data in order to map population density. These files can be found at the U.S. Census Bureau's FactFinder page. [5] Download the data that is the most appropriate for your application. Download the data in txt format in order to transfer to excel. This will allow us to link the data with our shapefiles in ArcMap. Open the file in excel. If you downloaded population data, calculate the population density with this equation.

Population Density=Population/(Land Area*2,589,988)

Now the file can be linked with a shapefile in our map. To do this. Click on the file you downloaded from and select 'Join and Relate' and then 'Join'. This will pop up this window to select the file and table to connect. Note: The files can only be linked if they share at least one attribute.

Join 3 resize.jpg

After identifying the symbology to be displayed and number of categories, you should get a map that looks something like this.

Pop density 4 resize.jpg

Next, lets try to find some utilities data for California. Specifically, we are interested in broadband providers. Luckily the California Public Utilities Commission has already compiled much of this data and keeps ongoing and up to date files of various service providers. Maps with broadband service are found here: [6] After adding this layer, the combined maps will look like this.

Pop density and broadband access 5 resize.jpg

An "underserved" area is an area where broadband is available, but no wireline or wireless facilities-based provider offers service at advertised speeds of at least 6 mbps download and 1.5 mbps upload. "Served" areas exceed these advertised speeds. Let's take a closer look to see how the layers overlap in a closeup view of the map.

Close up 6 resize.jpg

Refining the data to look at certain demographics or population densities is easy with "Select by Attributes". This allows you to define specific parameters in a table or shapefile and export it as a new database or map.

Defining variables 8 resize.jpg

In this case we wanted to isolate the higher population density areas.

Select by attrubutes 7 resize.jpg

In the next map we will combine the high population density file (in purple) with the "underserved" areas to see if there are areas with relatively high population density but no broadband access. This can be done with the "Select by Location" function in GIS.

Select by location 9 resize.jpg

Overlaying this file results in two optimal locations, roughly 2x3 miles, with high population density and no access to broadband that would make ideal locations for mesh network pilot projects. The two areas are difficult to see in the below map.

Optimized locations 10 resize.jpg

But by zooming in we can see the two areas.

Sonoma 11 resize.jpg
Ventura 12 resize.jpg

Make sure that the area you choose is not water by downloading a river and lake shapefile.

Proper ground validation is important with GIS to ensure that the area does indeed provide the attributes you were looking for.

Potential next steps include using high resolution satellite imagery for object recognition analysis. By identifying individual houses a more accurate map can be made of high density housing clusters. Additionally this will allow for optimal routing algorithms to be applied for technology dissemination.

Testing methodology[edit | edit source]

The example system was tested using the following methodology. The system, comprised of three stand alone units, was placed outdoors with each branch unit roughly 500 meters from the main unit. The range and capacity of the mesh system was then measured. Finally, current and voltage readings for each unit were collected.


Since the units were placed on rooftops, it was assumed that shading played a limited role and that systems were placed in an optimized location above shaded areas. Therefore, calculations regarding orientation and sizing due to shading parameters were ignored.

Start getting permissions for this now.

See also[edit | edit source]

Discussion[View | Edit]

Cookies help us deliver our services. By using our services, you agree to our use of cookies.