Solar powered home data center

[madsmao]

A little while ago, I came up with the crazy idea to build a little solar powered data center in my home. I work professionally with hosting solutions for the Thai market, and we use big traditional data centers to support our infrastructure, but I'm keen on experimenting with more sustainable solutions at home. Hopefully, over time, I will be able to come up with innovative solutions to help lower the overall power consumption of the equipment needed for smaller scale hosting, and find reliable ways to supply the power needed through renewable sources of energy. That's my dream.

I am by no means an expert on DC circuits and solar power, but I do know a thing or two about computers. I have already done a ton of research on DC circuits and solar power, but I'm hoping that I can get some feedback here for my initial setup. It would be nice to validate my ideas before I get started, and hopefully avoid spending too much money on equipment that won't get the job done. Here's my current idea for a basic setup:

Solar Panel -> Charge Controller -> Battery Bank -> Small server with 12V PSU

Obviously, this is very simplistic, but let me go into a bit more detail about the individual parts of the setup. Starting with the load:


Small server
The goal is to have a server with 4 cores, which will support running virtualised applications inside Linux Containers (I'm happy to expand on this for those interested). Obviously, power consumption needs to be as low as possible, and I hope to get below 100W (or even lower) under full load, but I wish to plan a setup which is able to support a steady consumption of 150W to make sure I have enough power. I have already found a very neat little 12V DC power supply that I plan to use, and I'm also thinking about equipping the server with a built-in UPS as an added failover measure.

I have a few questions regarding the server:

• The PSU accepts a maximum input of 12v/8A, but is supposed to automatically compensate for current fluctuations. Would it still be wise to put some kind of fuse box in front of the PSU?
• How big a battery would I need for the built-in UPS to allow the server to run for 1 hour?

Battery Bank
I would like to scale the battery bank to allow the server to run on battery power alone for up to 48 hours. However, I'm quite uncertain about this particular part of the setup.

These are the questions currently on my mind:

• How do I go about calculating the capacity needed for 48 hour non-stop operation?
• Can I use 12V batteries, considering that I also want to charge the 12V internal battery in the server hooked up to the UPS?
• How do I wire up the batteries? Serial, parallel, a combination?
• How will fluctuations in ambient temperature affect the batteries?
• Any recommendations regarding brands/battery types?

Charge Controller
I understand the purpose of the charge controller, but I would love recommendations on certain brands/types, that would work well with the setup I have in mind. Obviously, it should be a controller that has the capacity to support my overall setup, and it would be great if it's big enough to support future expansion.

Solar Panel
I expect to have minimum 8 hours of direct sunlight on any given day. Based on that assumption, here are my questions:

• Given the right size battery bank, as outlined above, how many solars panels — and what capacity — would I need, to be able to fully charge my battery bank within 8 hours — and power my load at the same time?
• How do I need to wire up the panels? Serial, parallel, a combination?

Power/performance measuring
The whole experiment is rather pointless if I cannot measure the efficiency of the setup. Here are the main things I would like to measure:

• Power generated by the solar panels
• Power consumed by the load

Ideally, I would like to be able to collect data over time and download them to my computer for analysis. However, I have had a very hard time locating equipment that will allow me to do that, so I'm very interested in hearing about equipment that will fit my setup, and help me accomplish my goal.

Wiring
I will not be pulling cables over long distances (maximum 15-20 meters), but, as I understand it, there can be significant power loss if using the wrong cables. I would greatly appreciate some input on how to pick the right cables for various parts of the system.


This is a pretty big project for me, but I'm super excited about getting started. I realise that I'm asking a lot of questions, but any kind of input is greatly appreciated.
I will obviously be sharing my experiences with you as I build this, and I plan to document everything in video, photos and text. It would be great if others can benefit from my experiments if trying to build a similar setup.

[SeanB]

Fuses would be good, and as it is 12V you can use regular automotive blade fuses, 15A would work for this application between the battery and the PSU, so that a fault does not cause a fire from melting wiring. Another on the input to the charge controller, and a 60A car audio style fuse right by the battery for protection from wiring faults. As to batteries you really want to size so the battery does not discharge below around 70% during the non daylight time, and as well the panel will have to supply an extra 150W to run the computer as well, so you really need a minimum of 1000W of panel, so that you can both charge the battery bank and run the roughly constant load of the server. Here you have to take the charge efficiency of lead acid into account, so larger panel size would be required. Panel type and connection depends on the charge controller, as to what it accepts as input for MPPT input, and you need one which can charge a 12V battery bank. Battery is sized on current draw, so you have roughly a constant draw of 8A, so you need to size so the 20Ah drawn is under 30% of full capacity for the normal 16 hour night, so you need over 80Ah of battery, which is doable, though you probably want more to do your 48 hour outage without too much battery damage.

As to monitoring you can use a charge controller with built in monitoring, or by a cheap LCD display wattmeter off eBay to put in the feed to the server and another to measure the power coming in from the panel. Batteries would be likely best as lead crystal or deep cycle marine ones, and if buying deep cycle marine ones get flooded cell types with caps so you can measure electrolyte level and top up with distilled water when required.

As to wiring, thicker cable is better, lower voltage drop from the panel to controller, so you will be looking at 4mm copper wire for up to 20m or so, and 6mm for up to 50m, as a minimum. Up to 10m you can use regular 2.5mm cable. Wiring from controller to battery, and to the server, is likely to be shorter, so use 4mm wire for it all. Ensure the solar panel mounting frame is bonded to an earth stake or series of stakes driven deep into the ground ( 2m long 12mm steel rebar driven into ground, at least 2 of them, and connected with 6mm copper wire to the support of the panel) for lightning protection, though you would want a proper lightning rod system higher up as well, or be in the shadow of one, so that storms do not blow up everything.

[madsmao]

SeanB, thank you so much for taking the time to reply to my post with so much useful information. I will take some time to digest the information you have provided and apply it in my research. I will no doubt return with slightly less noobish questions in a not too distant future ;-)

[madsmao]

I have been doing a bit of research tonight on charge controllers. I understand that I have to make a choice between PWM (cheap, but less advanced/efficient) and MPPT (more efficient/advanced, but also more expensive). I have found many sources saying that MPPT is a superior technology, and that makes total sense based on the description of how the technology works. However, I have also found a lot of sources saying that the benefit of MPPT over PWM are marginal in a sub-tropical climate. Since I live in a sub-tropical climate, I'm now a bit uncertain about what to buy.

For reference, I'm currently planning to start with an array of 18V/100W panels (probably 3-6 for starters) to charge 12V batteries. Am I likely to see any huge benefit from using an MPPT charge controller rather than PWM, or are my sources right about PWM and MPPT being equally good in very warm conditions?

[PoBoySolar]

i'd say go with PWM unless you are building your own controller.  The reason is why 18V panels are used in the first place. At high panel temperatures, the panel power point is about 15-16V.  With IR loss, diode loss, converter loss there just isn't much to gain with a 12V array. Place the panels in series and it becomes a different story.  With active use a system like you proposed probably never really gets beyond bulk charge mode (like a car).  I build my own controllers and operate them at a fixed power point.  Power point is only dependent on temperature if you ignore IR losses.  Any SMPS can be made to operate as a power point controller.  You just need to monitor the input voltage and fool the output sense voltage circuit that the output is to high.  Given the availability SMPS that are rated to operate at 100V this should be easy when you have about 6 panels.  I have found a lot of wall wart supplies will work down at 50V with reduced power.

I was doing some testing on a single 100W panel last spring in a water heating test at power point in a northern location.  This was to get typical real world results for my location.  What is interesting is the highest peak power readings are often on days with poor total day results.  This results from overcast clouds keeping the panel cool.  When the clouds break for a short period of time, maximum power is produced from a cool panel (till it heats up).
PEAK READINGS AND DAILY TOTAL

3.73A  64.1W     255WH.      partly cloudy day
3.68A  63.2W     126WH       VERY OVERCAST
4.22A  72.6W     191WH       VERY OVERCAST
4.08A  70.7W     155WH       OVERCAST & RAIN
2.87A  50.4W     285WH       WISPY CLOUDS
2.92A  51.2W     307WH       HAZY, PTLY CLOUDY
5.45A  88.6W     162WH       OVERCAST RAIN
3.57A  57.7W     413WH       
3.95A  68.4W     301WH       PT CLOUDY

3.13A  55.1W     374WH       SUNNY

3.18A  56.3W     252WH       SUNNY

[madsmao]

Very useful input, PoBoySolar. The real world test results are a big help, and I guess I can expect to see pretty descent results from the 18V panels with a PWM charge controller, as the sun is almost always shining here. Very overcast days are extremely rare, and happen almost exclusively during the rainy season. I have some cunning plans I'm working on for producing power from other sources during the rainy season, but that is something I will get back to at a later point.

So, given that I'm most likely gonna go with a PWM controller (I'm not quite ready to start building my own controller just yet), a few question still remain:

• What kind of input current should the controller be able to handle, if I start out with 4x18V/100W panels? I'm not quite sure if the panels need to be connected in series or parallel for my purpose, which is why I'm not sure of the controller specs.
• Does anybody know of a specific PWM controller that will allow me to collect usage statistics, and make some fancy graphs for analysis? Also, I really want to be able to share real-time data with anybody who is interested in the project, so this is a key feature that the controller must support — unless there is a better way?

[SeanB]

PWM means they will be in parallel, so your controller must handle the peak current of the 4 panels, or around 10A per panel. At this current you will need a blocking diode per panel, probably best to use a old PC power supply to give you 2 high current dual diodes on a heatsink, using both halves in parallel so they only drop around 0.5V, so you have low voltage drop on the diodes. You will need 2 PSU units, so you get the 4 diode packs needed, keeping them on the heatsinks so you can handle the 5W per diode easily.  You can reuse the cases as housings for the diodes, so they can cool easily. You can even use the fan, feeding from the incoming side via another diode ( a small 1A one) to actively cool the diodes. You can fit the entire controler into the case easily with a little care.

PWM control is often implemented using a simple control loop that shorts the input of each diode block with a high power MOSFET driven into saturation to shunt the panel as needed, so the panel voltage drops to zero. That mosfet can be mounted on the heatsink as well, there should be enough holes to do so. Choose a mosfet with low on resistance and one that will saturate with 10V on the gate, so it does not run in a linear mode as it will fry. All 4 can be driven using a simple comparator ( with delay of around 0.5 second) such that they are switched on at 14.4V on the battery bank, and hysteresis so they turn off again when the voltage drops below around 14.2V as the pack recovers to float voltage.

Cabling to the controller will need to have low resistance, so the 4mm cable will do up to around 20m, giving a panel voltage with a full battery of around 15V before the mosfet switches on, so not max power point but usable, and better at low charge state where it will naturally deliver more current into the battery. Put the monitoring after the diodes and before the battery, as otherwise you will only see current pulses or a wildly changing voltage.

[madsmao]

SeanB, I really appreciate your input, and I do get the overall idea, but you did have me Googling quite a bit to try and understand everything you just said — and I'm still left with little to no idea of how to achieve what you are suggesting :-)

I look forward to the day where I have the necessary insight to implement your suggestions, but I'm just not there yet. I'm a software guy, and the detailed understanding of the inner workings of hardware is something that I'm only slowly starting to grasp.

So, in short: Is there an off-the-shelf solution that will fit my needs, or will I be forced to speed up the self-education process to make this happen? Or perhaps the MPPT controller is the better choice for a noob like me?

[SeanB]

Ebay listing for solar charge controller.

Random listing is

http://www.ebay.com/itm/3A-5A-10A-15A-40A-45A-MPPT-Solar-Charge-Controller-Home-Battery-Regulator-12-24V-/400400454545?var=&hash=item5d39ba1391:m:mHg_RaoHxA7muwosZAMUitQ

http://www.ebay.com/itm/30A-PWM-Solar-Panel-Battery-Regulator-Charge-Controller-12V-24V-Auto-Switch-BA-/390909701375?hash=item5b0408acff:g:6isAAOSwVFlT7GkI

Read the description, and you can see the panels and powers it supports.  You probably want the 30 or 45A one as a start, sizing to the panel power.

[madsmao]

Thanks, SeanB. That looks more appropriate for my current level. Still doesn't look like any of those charge controllers directly support a network linked setup allowing me to collect usage data in real time though, but I will keep looking to see what I can find.

[SeanB]

Arduino or Rasberry pi will do that, though the Pi does have built in ethernet, and the new mini pi is low power as well. Plus the onboard Debian allows you to control things as well, like controlling power or shutting down the servers gracefully at low battery before the power is shut off,

[madsmao]

I like the idea of the Rasberry Pi for that purpose. Not that I have a clue how to hook it up to achieve what I want, but I did find some interesting info during a quick Google search. I will have to dig a bit deeper and see what I can find.

Again, thank you so much for your input. This is all a pretty steep learning curve for me, but I'm super excited about making this happen.