Have a look at this. Some info on running those things on a battery. http://blog.hekkers.net/2015/04/06/esp8266-good-enough-for-a-battery-powered-sensor/

My sensors have been running for 5 months and still going strong.

My problem is that I'm reporting the temperature every minute, which I won't be changing, so I'll never get a year off just a battery.

Some Lithium battery charging boards arrived yesterday for me to charge some 18650s pulled from a laptop battery. I also discovered a new version the same day I ordered which also have discharge protection for unprotected cells, so I'm waiting for them to arrive then I'll have a setup which will be...
 

Solar Panel
     |
Charging/Protection Board -> ESP8266
     |
18650 cell

"I'm reporting the temperature every minute"

Why?

The "system" controls my central heating so I like the "thermostat" readings to be fairly regular so it turns on/off when it needs to.

This doesn't cause it to turn on/off much more frequently, increasing wear and tear?

You could check it every minute but only send the temp if it's changed from the last check.

Definitely possible but I've had a look at other remote control type heating system and I'm sure they check the temp at least every minute.

I'd imagine the older type bimetallic rotary dials could flick on/off fairly regularly, unless they naturally have some sort of hysteresis curve.

In reality it very rarely clicks on for just a minute. It takes a few minutes for the radiators to get warm and the room heat up enough to hit the target temperature. Even if it does only need to increase by 0.5C

That's definitely possible. Only reason I'm not doing that is because it would leave gaps on my chart. I could fix that by running a script every minute or so and if a value hasn't been sent just repeat the last value.

I'm working on a system that has to sample an analogue feed at about 50 kHz. I'm using an fpga to do it which supposedly is capable of sampling at 40 MHz, but I'm having a bitch of a time getting the timing right (as in recording the time of the sample, needs to be consistent). I hate it. Wish I could sample once a minute instead :((

50 KHz data rate - that is impressive.   Sometimes portions of the code are running much faster than others and you have to add "wait for data to be ready" code.

I have not written code in a long time, but I am working with an embedded code writer on a project and he runs into weird stuff like that all of the time.

I assume that you are using real time coding to pull that off?

edit

The company that I work for has struggled to use an FPGA sampling for analog sampling even at 10 ms sample rates and feed it into a controller error free.  There was a lot of noise in their incoming analog signal, more than they expected. 

The only boards that I have built were for power LED + LED driver on one board.  For prototypes, I use Iconic PCB in Australia.  I know it sounds crazy, but they do a great job by cnc machining the boards and traces.

The main advantage for me was that I needed 2 oz copper traces (single layer) and a really healthy heat spreader + ground plane. 

They bonded together a 1.5mm copper sheet + thin FR4 + 2 oz copper stack for the project.  

- The 1.5mm copper was the ground plane and a great heat spreader.  Access to it was by machining openings down to it from the top, and filling in the opening with solder paste.  This was especially useful for the control chip heat spreader connection on the back of the package.  I also was able to use it for the LED heat spreading.

- FR4 layer - normal function

- 2 oz copper was for all non ground traces

- I used 0 ohm resistors to "jump" over traces as needed, rather than add another board layer.

- Typical part size was 1206, mostly because I have poor dexterity, but also it was handy to have larger parts for heat dissipation.

With this setup, I was able to duplicate an applications note circuit done in 4 layers with just the ground plane and 2 oz copper layer. 

I don't have the knowledge to design the circuit, so I hired a local EE to do the drawings.

The Iconics people are great, really a family business.


 

Decided it wasn't possible in the end, at least not with my skill set and this hardware. I don't have the background or knowledge to pull it off at the level of accuracy asked for - 2 analogue channels being sampled, but the hardware muxes all analogue inputs through the same ADC so they physically cannot be read at the same time (which is what I was asked to do), even timing the sampling perfectly they samples end up being [sample time] + [switching input time] apart.

Now testing a recovered 18650 cell with a couple of solar panels to charge.

This one should hopefully last longer because the solar panels no longer have a voltage drop with a diode.

Once the lithium charging boards with extra protection arrive (they're taking ages) I'll make a board up with all the bits properly soldered.

I suppose a board like this could be used as a reasonable battery backup type system for plugged in sensors. Saying that, if the power goes it will probably take out the server, wifi etc so it would be a little pointless.

This setup is currently at 9 days on solar+battery.

ESP recorded voltage hasn't dropped below the maximum it can read - 4.09V. Today I noticed that the charging module was showing "fully charged" so it looks like the two panels are providing enough power during the day to top up the battery to full. Might actually last forever if I want it to.

Nice one. Be good if it will keep going without needing to charge/replace batteries :) How much sunlight are they getting?

I have two panels (6V - 0.6W and 1.1W). Attached to a south facing window, so getting as much as possible up north.

I could do with removing one of the panels to see if I can continue with just one.

(now unplugged the smaller one)

Small panel unplugged and the battery charger showed fully charged a moment a go.

Full charging with just the large panel, but there's been a lot of sun today so it's probably not a surprise. Time to try just running on the small panel.

Small panel looks like it isn't sufficient to keep the battery topped up. It has dropped below the 4.09V max voltage and over the last two nights has had a new low value so it's gradually getting lower over each night.

How much sun have you had over the lat two days?