Off-Grid Supplemental Solar Air Conditioner Project Build Log

AntaresAdroit

New member
Joined
Mar 13, 2021
Messages
2
Hello there. I've been lurking on this site and working on a small off-grid solar power system with battery storage for about the last month, so I thought I should make and account and start a build log. The goal of this project is to to run a small 550-800W, 8000 BTU window unit air-conditioner. Generally, I want to keep the bedroom of my house a bit cooler than the rest of my house. I installed a new central split air-conditioner a couple of years ago and for efficiency, I sized it so that on the hottest days here in Las Vegas it can keep a comfortable temperature if it runs most of the day. Unfortunately, that means there's little spare cooling capacity to shut off in the afternoon/evenings which is also peak consumption time on the grid. My goal for this project is to shift enough of the cooling load to this supplemental unit so that I can shut down the primary central A/C during peak-time and still have a comfortably cool bedroom by bed-time. I'm hopeful this will also make switching to time-of-use billing with my power company practical.

In general, I'm disassembling some 18650 LiPo e-bike packs for the storage in the system. I have about 280 3.3Ah cells, so I think I'm sitting at about 3kWh, though I'll probably limit the voltage range I use to keep from stressing the cells. I'm experimenting with some 3D printed cell holders with leaf-spring battery contacts for the pack structure. In general, I want the modules to be serviceable so that finding and replacing bad cells doesn't require reworking dozens of spot-welded or soldered connections.

For solar power, I have 4x 325W (~9a@32V) panels. I'll probably run these in 2s2p configuration, though I probably should see if a 4s1p config extends useful power generation in lower light conditions. I haven't selected a solar charge controller yet either.

I have a fair bit of experience with Arduino microcontrollers and logging things using MQTT and ESP8266 modules, so I'll be trying to do a lot of my system control and monitoring using those.

I'll be posting updates here as I work on things.
 
Last edited:

AntaresAdroit

New member
Joined
Mar 13, 2021
Messages
2
Thanks for the welcome! I'm working on a discharge cell testing station. I have the hardware done (well.... maybe I could add a few more sensors but I think my project scope is creeping enough already).

1616362620503.png


I'm using an ATMega 32U4 based, Arduino compatible microcontroller board from Pololu as the brains. The control board has a boost-buck converter so I can run it off of a single lipo cell and a micro sd slot built in so I can create discharge log files for the cells. Also, there are some extra I/O pins broken out and holes ready for soldering extra ground and power bus pins that I took advantage of.


I have 6 cell discharge stations, each with a fixed resistance load switched by a MOSFET. Using a fixed resistance value means the discharge current will drop as the cell voltage drops, but it simplifies the data acquisition and control scheme. All I need is a single analog microcontroller pin to monitor the cell voltage and I can calculate the discharge current from that (...and then the power and energy consumption). The MOSFETs each require a single pin to switch on and off. In theory, I could probably control the current by switching the MOSFETs with PWM, but for me, there's enough software to write already.

The cell and resistor holder part for each station is 3D printed from PETG.

1616362717600.png


I also made some 3D printed jigs for bending and soldering the 14awg wire that connects and supports the resistor network.

1616362790403.png
1616362837281.png

1616362887675.png


I'm using five 5W rated 15ohm ceramic resistors connected in parallel so the effective load is about 3 ohms. At 4.2V, the current is about 1.4A and the total wattage is about 6W, so each resistor is handling about a quarter of its power rating. I tested a bank at 4.2v for an hour or so and the resistors get pretty warm, but you can touch them for a few seconds without burning yourself, so I'm pretty happy with that. Also, the resistance of resistors generally varies with temperature bit, but I took some measurements and it seems like that is only causing a few milliohms of difference, which I think is fine for me. The load stations are designed so that the resistors are suspended from the 14 gauge bus wires and have room for air circulation on all sides.

The MOSFETs are ST STP55NF06L N-channel MOSFETS in the TO-220 form factor. The drain to source voltage (vds) rating for these is 60V, so way overkill for switching these 4.2V cells and the on resistance (rds-on) of these is around 10 to 20 milliohms when using a 5V control voltage, which seemed fine. In general, I didn't want to add any external heatsinking or to need to factor in the FET resistance in my load calculations, so I wanted a rds that was pretty low. I found these on Aliexpress for reasonably cheap (0.14usd/each).

The power supply for the microcontroller is a single 18650 lipo cell with a cheap USB charging and BMS board like this one. I sized the holder for my pack cells, but the cheap cell I'm using for the supply has an extra nub on the positive terminal, so it's a tight fit. The boost-buck regulator built into the Pololu A-Star microcontroller board steps the voltage from the cell (down to 2.5V) up to a stable 5V for the microcontroller and peripherals. I expect the control system to draw something like 200-300mA, so I'll probably mostly keep it connected to USB power, but it's nice to have the backup.

I'm using a small 128x64 pixel OLED screen and rotary encoder knob with a built in push button for the user interface. I also threw a real-time clock on there. I plan to use the knob to set an ID number for each cell (which will be 0 to ~300) and then I can pull a date and time from the RTC to handle naming the log file on the SD card something unique. I haven't really started programming yet, so I'm not really sure how it all will work, but I'm hopeful there will be enough memory to use all of the associated libraries and other code I need to make all of this stuff function.

The whole setup is mounted on a wood patterned ceramic floor tile so that if anything goes wrong, the surface is fire resistant. The control and monitoring connections are done with 30awg wire wrap wire and a wire wrap tool.
 
Last edited:
Top