Trickle charger efficiency measured ... it's not good

Today I charged for 20 hours from 20% SoC to 63% at an average of 1.59 kW, meaning an average of 7 amps at 230VAC.
I was hoping for a slightly higher rate but I discovered that the Ratio Electric EVSE drops down to 6 amps in hot conditions and has to be restarted.
I measured the AC energy using a TP-Link Kasa plug which I believe is very accurate from some comparison tests I've done with another energy measuring adapter. However, the Kasa adapter gets very warm and I think caused the EVSE power plug to trigger an overtemp.

The bottom line is that I needed 31.8 kWh to add 22.1 kWh to the battery, a shocking 69.5% efficiency, much lower than I had thought. With these losses and my overall electricity rate of 33 cents/kWh it would be slightly less expensive and a whole lot faster to simply visit the nearby DC fast charger at 40 cents/kWh.

Combined with the real-life 15.3 kWh/100km I'm measuring, the costing works out like this:

15.3/(0.978 x 0.695) = 22.51 kWh(AC@7A)/100km

In my case $0.33 x 22.51 = $7.43/100km, or for the 290 km trip 2.9 x 7.43 = $21.54

The 0.978 value is the battery cycle efficiency which comes into play between charging and discharging.

To be realistic, gasoline locally is $2.75 / litre for the highest octane. My ICE over the same trip would use 2.9 x 10L/100km x $2.75 = $79.75

So, I'm still 3.7 times better off. However, in just over 2 years EVs have to pay a road user charge which at the current rate of $76/1000 km will add 2.9 x 76/10 = $22.04 to the same trip.

So, at 21.34 + 22.04 = $43.38 it seems I'm still better off cost-wise with an EV.

 

Both the Kasa energy measuring adapter and EVSE got warm. I think the main issue was that I have everything in a closed cabinet and that was in the sun, and it was a 33°C day. Resetting the EVSE and opening the cabinet door seemed to resolve the issue. There's nothing wrong with any of this equipment, I just was misusing it.

In Australia and New Zealand portable EVSEs often default to 6 amps but can be set to 8 amps if configured before plugging into the EV. Some EVSEs can also be set at 10 amps (our maximum for a standard household socket) but must have temperature sensing in the plug. It's all about safety when using old, shagged out sockets - very common here.

In the end I have to pay in local currency and the exchange rate doesn't really come into the equation because electricity is not a global commodity. Salaries here are typically numerically-similar to the US, perhaps similar to those in Canada as well, so the value of a "dollar" is much the same. We have an oddball semi-unregulated electricity market here and as usual it's the consumer that suffers. But at least it's 90% green.

 

Yes, of course the losses I'm focusing on are at the OBC. The EVSE warmth was just a distraction that lowered my average rate. I should have been more clear with the title of my thread, such as "Charging losses high at trickle-charging power levels". I don't always think of all the ways things can be misinterpreted.

 

A further low-power charging test with better accuracy and more information. The TL;DR version is in the graph itself while a detailed explanation follows.



I've now completed a second test of low-power charging, taking advantage of the lessons learned from the DC tests. The main difference between this test and the last is that (a) I now have believable data for the gross DC power provided to the HV system from the OBC, and (b) I've discovered that Torque Pro gives up logging after about four hours and I've had to shorten the test interval to suit, including factoring the AC energy measurement since it carried on charging for twice that time.

My goal is to make these tests as accurate as practical, so I'm trying to maximise confidence in all sources of data. Values need to have three numerical places of resolution where possible to avoid errors building up. For example the lowest resolution is found in the battery charge energy meter "CEC" (0.1 kWh) so I should add a charge amount of least 100 times that increment to make the contribution to the overall error less than 1%.

The AC energy is derived from a TP-Link Kasa smart plug which I've verified as matching the reading from an Elto EMA-1 power meter. If there was a discrepancy I wouldn't know which one was correct but since they read the same I can fairly safely assume they're both correct. It helps that the power factor of the OBC is about 1.0 and the Elto tells me that.

Gross DC OBC ("on-board charger") output energy, which is powering everything on the HV bus is derived from integrating (over time) the PIDs named "000_Battery Voltage" and "000_Battery Current". So, at every 5 second interval a slice of energy to the tune of VxA/(1000x720) is added to the prior entry so that it accumulates. "720" is 3600 seconds/hour divided by the 5 second logging interval. On the DC bus I'm assuming we have only the traction battery and the LDC (low-voltage DC power supply). The LDC powers all the Kona's electronics. There are no other plausible loads to the HV system when slow charging at 17°C pack temp with no climate systems working.

The Kona's CEC meter records only what enters the battery pack. As before, I'll reiterate that this test does not involve the battery's efficiency. That's only the main load; it could just as well be a huge light bulb and the results would be the same. The SoC readings are not relevant for this reason.

I think the graph above is self-explanatory. All the curves are cumulative energy values over time. Generally, during AC charging all the rates are steady so the lines track fairly straight. But variations can be found at the start (as evident in the DC charging tests) because the 12V battery gets a charge for the first 30 minutes. That doesn't mean it draws power for that long, it simply means that it can if it needs to. So, to minimise that impact on my test I placed the Kona in Utility Mode for 30 minutes beforehand to get the 12V battery fully charged.

Note that for the red jagged line you only read the peaks of each step. The value (CEC) is truncated so that it only increments when it arrives at each new value.

The purple top line in the graph is AC input based on the average rate for the test period. I can't log from the Kasa meter in real time so I have to make it a straight line.

The next line down in blue is the gross OBC DC output. New to this test, the OBC efficiency can be derived by the ratios of the curve slopes, in this case 86.4%. As a note, Hyundai quotes an efficiency of 91% which undoubtedly applies at full power rather than 7.2 A. As a second note, we will assume the OBC electronics is powered off the 12V system and so that loading is not included in the losses.

The jagged red line is, as mentioned, the energy accepted by the battery pack, expected by logical deduction to be the gross OBC output minus the LDC load. We have PIDs for the 12V system voltage and the (unverified*) LDC output current but I don't know how much power the LDC uses from the HV bus. However, since I can calculate the LDC output power I can make a starting guess at the conversion efficiency to get the related input power. The gross OBC DC output minus the LDC input power results in the net OBC out, that's what is left to charge the battery pack. *Any error with the LDC output current PID will zero out in the final overall charging efficiency.

So, the net OBC output is the black line and that slope is affected by the OBC efficiency, until now simply guessed at. So what I can do is iterate the guessed value so that the black line coincides as closely as possible to the CEC tracking (at the peaks). When completed, that gives me the LDC efficiency of 92%, an entirely plausible (and laudable) value. There is a slight nonlinearity between the black and red lines, perhaps due to the efficiency varying with load during and after the 12V battery charging interval.

Multiplying the two efficiencies together gives me an entirely believable overall charging efficiency of 79.5%.

 

It's true that the internal resistance increases, however the measurements taken here are not affected by the nature of the load even though that does seem intuitive.

 

I'll just add that the Control Pilot signal from this particular EVSE allows for a 7.8 A draw so the Kona is to blame for only taking 7.2 A. I actually took this photo a few years ago but never did the calcs. The first pic shows the 1kHz pulse in charging mode, the second a detailed view of the PWM part of it that advertises the allowable current to the EV.