MorningStar TriStar-45 PWM and Raspberry Pi 3 data project

full_mosey wrote:
I am working my way through the MS MODBUS Specification specs to capture data onto the RPi3.

The plan is to avoid fixed control panels, and instead, run a web server on the RPi3 and then view info via smartphone, tablet, laptop, etc clients. A mifi will provide LAN connections.

As of last Friday, I have been capturing the 'RAM' data fields for each minute. The fields are listed on page 4.

I have no plans to use spreadsheets. I store the data in an sqlite3 database.

Here is a sample output from an sql query run from the sqlite3 CLI:

day cs cstate avg_bvolts min_btemp max_btemp minutes
---------- -- ------ ---------- ---------- ---------- ----------
2017-04-23 3 Night 12.93 9 25 620
2017-04-23 5 Bulk 13.35 9 26 178
2017-04-23 6 PWM 14.86 10 22 328
2017-04-23 7 Float 13.36 22 27 314

The query displays a summary of the minutes for each control state for the previous day.

The bank is 270ah AGM which should explain the high Night battery Volts. I do run two 12V compressor fridges 24/7.

Looking for Comments, Suggestions, Questions.

HTH;
John

Comment Number 1: COOL!!!!!!!!!! :)

Number 2: Sounds like you're a RPi expert but just my two cents: invest in a good microSD card if you're hammering the card with IO, like sqlite will do. I have 6 RPis in the house, and when I buy cheap SD cards they last about 8 months. I bought a couple of Samsung SD cards and they've gone three years.

Web Server - Not sure if you have a favorite, but if not I've been running lighttpd on my home Pi and like it. Small. Solid.



I've got an RPi3 with a SensorHat in our coach collecting environment and OBD2 data. I'll follow your lead if/when we add solar!

Very good to hear I'm not alone in my madness :)

I don't have solar, but do have a RPI2 collecting data from my Magnum and several other custom built WiFi sensors and switches around the trailer... it all ties together in a web based "dashboard" that I can bring up on any tablet or smartphone.

Not familiar with the RPI3, but if it's using one of those tiny microsd cards for the database, you'll want to move the db over to a USB stick or external USB hard drive. The SQL server on my RPI2 burned through two microsd cards already, the constant writes killed the cards dead.

RPi3 is an RPi2 with built in WiFi, Bluetooth (BLE) and a higher clock speed (1.2 GHz). Add a heatsink and folks are overclocking to 1.4ish. The RPi3 really needs a good 2.5A or better power supply. Mine was flaky with a 2A feed.

Uno mas - Western Digital has a line of external drives designed especially for the Raspberry Pi.

pconroy328 wrote:
Uno mas - Western Digital has a line of external drives designed especially for the Raspberry Pi.

I put an old laptop hard drive in a usb box and plugged it into the Pi. Cost about $7 for the box. More space then I'll ever need for databases and source code.

Thanks all for the feedback.

The rpi3 is installed in the van. You could look at my profile for details.

While here at home, the rpi3 is connected to the home wifi. When traveling, it will be connected to a mifi in the van.

The rpi3 will be headless. Right now I have a wireless USB keyboard and an HDMI to monitor cable attached until I can automate reconnection to the mifi(LAN) during reboot.

The rpi3 uses between 0.28A to 0.34A per the DROK USB power meter. That translates to ~8Ah/day.

I have booted the rpi3 with a 1A rated USB charger. It took several tries before I found a good enough USB cable to carry the power. Get good cables! I run it in the van with a 2.4A Anker 12V car-plug charger.

I want to keep the power low, so no external drive.

For backup, I copy the data and programs from the rpi3 with my laptop's file manager which connects to the rpi3 via ssh.

If write-wear on the micro-sd card becomes an issue, I could reduce sqlite3 writes to once per day by keeping an intraday database in memory.

A cron is run each minute to capture the data.

The programming language is PHP5. Installing PHP automatically includes the apache2 web server. PHP facilitates dynamic html.

I could under-clock if I have thermal issues!

Per 'top', my PHP code uses between 2% and 4% CPU while extracting from the controller and saving into the sqlite3 database.

I have had too much pi today!

Can I interest anyone with analysis of the log data? After all, the goal of the project is to understand what happens.

HTH;
John

Data analysis is where I'm lacking right now. I'm collecting reams of it, but not really doing anything useful. I only putsy with the programming when camping and it's either unpleasant outside or insomnia hits me... so it takes a long time to get anything accomplished.

Below is output from an sql query.

The query did an hourly summary of yesterday's one minute interval data.

hour avg_cc avg_tv avg_cs avg_pwm%
---------- ---------- ---------- ---------- ----------
00 0.2 0.0 3.0 0.0
01 0.2 0.0 3.0 0.0
02 0.2 0.0 3.0 0.0
03 0.2 0.0 3.0 0.0
04 0.2 0.0 3.0 0.0
05 0.2 0.0 3.0 0.0
06 0.23 4.98 3.67 32.48
07 1.43 14.96 5.0 99.89
08 5.41 14.94 5.0 99.97
09 4.89 14.9 5.83 62.86
10 3.38 14.85 6.0 32.63
11 3.57 14.78 6.0 28.73
12 3.29 14.7 6.0 24.33
13 3.53 14.63 6.0 25.72
14 3.34 14.58 6.0 24.65
15 4.05 14.53 6.0 36.16
16 2.12 14.15 6.3 19.88
17 3.3 13.29 7.0 60.46
18 1.77 13.31 7.0 72.84
19 0.84 14.11 5.7 99.62
20 0.2 3.88 3.53 26.9
21 0.2 0.0 3.0 0.0
22 0.2 0.0 3.0 0.0
23 0.2 0.0 3.0 0.0

I have bolded three values of average control state. The TS reported 4 states; 3=Night, 5=Bulk, 6=PWM aka Absorption and 7=Float. By averaging these numbers we can see the TS transitions between states during the hour.

The 5.83 at hour 09 indicates a transition from Bulk to PWM states.

The 6.3 at hour 16 is PWM to Float. This is confirmed by the lower average target Volts. Changes to target Volts are also affected by battery temperature.

The 5.7 at hour 19 is more complicated because the 14.11 Target Voltage is either Bulk or PWM. However, the 99.62 PWM% indicates a rapid return to Bulk.

This data beats the c.r.a.p out of the Green, Yellow, Red LEDs.

HTH;
John

You might want to look into RRDtool for storing data. It's meant for handling time-series data. You can easily set it up to, for example, keep 1 minute stats for the previous day, 5 minute stats for the previous week, and 1 hour stats for the previous year. It can handle min/max/avg, even defined functions, so would be well suited for tracking voltages and charge/discharge data. And, it makes nifty graphs for you.

I have a graph of the Volts from today's(May 3, 2017) data.



There are three data items;
tv - target Volts average, temp comp
bv - battery Volts; maximum and minimum

Data is collected at 1 minute intervals and then summarized by the hour.

All done on the RPi with PHP language.

HTH;
John