Some months ago I started toying with the stock pack management a lot using the OBDIIC&C SoC reset function and generally just by paying attention to IMA info and acting on it. This was in preparation for trying to implement a lithium LTO pack at some point, without a BMS. I've kept track of various OBDIIC&C IMA parameters for years, for every trip. I log nominal state of charge, pack temp, etc., and reset amp-hour counting at the beginning of each trip and log these at the end of each trip. So I have a massive spreadsheet of all these data...
I've also calculated a few things for each trip and, over the years, have honed-in on a couple simple metrics to keep track of what's going on, what the true state of charge of my pack is or should be, etc. For example, I add net amp-hours from trip to trip, so I have a running total of how many amp-hours have coursed through my pack...
After I started doing the 'toying', I noticed something interesting when I graphed these various data. If I let the car do its own thing, efficiency looks bad. If I deliberately do some things, deliberately thwart stock pack management, efficiency is better. That's what's illustrated in the graph below... Letting the car do its own thing is the left half, about 3 1/2 months, 'toying' with stuff is the right half, also about 3 1/2 months...
Over the first half the blue cumulative count SoC curve has a steep slope, while the black nominal SoC curve is relatively flat. This demonstrates that the number of amp-hours needed to maintain the pack state of charge is continually more than 100%. For instance, if the pack were 100% efficient, the blue curve would be flat and would track the black curve - I'd put in say 500mAh, I'd pull out 500mAh, and state of charge would be at exactly the same place where it started. Less than 100% efficient ('Coulombic efficiency') is normal and expected. But by how much?The steepness of the slope of the blue curve reflects the degree to which inefficiency exists...
Inefficiency has two main parts: self discharge and throughput efficiency. The green curve attempts to capture the throughput (in)efficiency part. When we calculate state of charge we adjust it down slightly because we know that not all the current contributes to charge or discharge energy...
In a nut shell - since this is taking me longer than I had hoped - after I started messing with stuff, the blue curve stopped rising as fast and as much, and the green curve started going down. Both of these reflect better efficiency: I'm using fewer amp-hours to maintain roughly the same states of charge. Since the green curve is supposed to account for throughput efficiency, most of the gain in efficiency is probably coming from less self discharge. That's more of a guess at this point, but probably the case...
black curve - the nominal state of charge as seen on the OBDIIC&C SoC parameter
blue curve - state of charge based on net amp-hours cumulative count. For example, cumulative count amp-hours might be 8500mAh after a few week's worth of trips, whereas the nominal capacity is 6500 - so the cumulative count SoC would be 8500/6500 X 100 = 131%.
green curve - state of charge based on net amp-hours cumulative count but adjusted for throughput (in)efficiency. Throughput inefficiency is the fraction of current that gets lost on the way, the current that doesn't end up actually charging the pack or contributing to output energy. I've calculated this a few times based on longer drives and on the car's own state of charge determination. The loss is about 2% - so for every 100mAh that go through the pack, only about 98 of them do something useful...
I also graphed average pack temp - I log pack temp at the beginning and end of each trip and the white curve reflects an average of these two values. I added this curve recently, because I was wondering if maybe higher temps over the first half and cooler temps over the second were responsible for the difference. But it doesn't look like that's the case. Average pack temps start to fall at about the middle of the left half of the graph, but we don't see a change in the slopes of the blue and green curves...
The red triangles show when positive recals happened. These are junctures at which the nominal state of charge curve accurately reflects what the car thinks the state of charge is. Since one of my toying methods involves resetting the nominal state of charge with the OBDIIC&C, the black curve doesn't always reflect this 'true' nominal. For example, one of my methods involves resetting SoC high when it's really low, typically by 10 percentage points. At those junctures, the nominal state of charge will read say 75% but it's really 65%. I don't go too far in this direction, though, so the black nominal curve generally reflects the 'true nominal'... The red triangles show when indeed we're looking at true nominal values...
Each data point represents a single day. Usually there's just one trip per day, but sometimes there's no trips and a few times there's multiple trips. I've condensed the multiple trip logs into one log per day, and I've added days where there were no trips. This makes the x-axis consistent with respect to time. The labels are divided into weeks, so about a week's time per vertical grid line... The whole graph is about 7 months. The first half, up until about the end of October, reflects a period when I was letting the car do its own thing. The second half reflects a period when I was toying with the pack management...
I've also calculated a few things for each trip and, over the years, have honed-in on a couple simple metrics to keep track of what's going on, what the true state of charge of my pack is or should be, etc. For example, I add net amp-hours from trip to trip, so I have a running total of how many amp-hours have coursed through my pack...
After I started doing the 'toying', I noticed something interesting when I graphed these various data. If I let the car do its own thing, efficiency looks bad. If I deliberately do some things, deliberately thwart stock pack management, efficiency is better. That's what's illustrated in the graph below... Letting the car do its own thing is the left half, about 3 1/2 months, 'toying' with stuff is the right half, also about 3 1/2 months...
Over the first half the blue cumulative count SoC curve has a steep slope, while the black nominal SoC curve is relatively flat. This demonstrates that the number of amp-hours needed to maintain the pack state of charge is continually more than 100%. For instance, if the pack were 100% efficient, the blue curve would be flat and would track the black curve - I'd put in say 500mAh, I'd pull out 500mAh, and state of charge would be at exactly the same place where it started. Less than 100% efficient ('Coulombic efficiency') is normal and expected. But by how much?The steepness of the slope of the blue curve reflects the degree to which inefficiency exists...
Inefficiency has two main parts: self discharge and throughput efficiency. The green curve attempts to capture the throughput (in)efficiency part. When we calculate state of charge we adjust it down slightly because we know that not all the current contributes to charge or discharge energy...
In a nut shell - since this is taking me longer than I had hoped - after I started messing with stuff, the blue curve stopped rising as fast and as much, and the green curve started going down. Both of these reflect better efficiency: I'm using fewer amp-hours to maintain roughly the same states of charge. Since the green curve is supposed to account for throughput efficiency, most of the gain in efficiency is probably coming from less self discharge. That's more of a guess at this point, but probably the case...
black curve - the nominal state of charge as seen on the OBDIIC&C SoC parameter
blue curve - state of charge based on net amp-hours cumulative count. For example, cumulative count amp-hours might be 8500mAh after a few week's worth of trips, whereas the nominal capacity is 6500 - so the cumulative count SoC would be 8500/6500 X 100 = 131%.
green curve - state of charge based on net amp-hours cumulative count but adjusted for throughput (in)efficiency. Throughput inefficiency is the fraction of current that gets lost on the way, the current that doesn't end up actually charging the pack or contributing to output energy. I've calculated this a few times based on longer drives and on the car's own state of charge determination. The loss is about 2% - so for every 100mAh that go through the pack, only about 98 of them do something useful...
I also graphed average pack temp - I log pack temp at the beginning and end of each trip and the white curve reflects an average of these two values. I added this curve recently, because I was wondering if maybe higher temps over the first half and cooler temps over the second were responsible for the difference. But it doesn't look like that's the case. Average pack temps start to fall at about the middle of the left half of the graph, but we don't see a change in the slopes of the blue and green curves...
The red triangles show when positive recals happened. These are junctures at which the nominal state of charge curve accurately reflects what the car thinks the state of charge is. Since one of my toying methods involves resetting the nominal state of charge with the OBDIIC&C, the black curve doesn't always reflect this 'true' nominal. For example, one of my methods involves resetting SoC high when it's really low, typically by 10 percentage points. At those junctures, the nominal state of charge will read say 75% but it's really 65%. I don't go too far in this direction, though, so the black nominal curve generally reflects the 'true nominal'... The red triangles show when indeed we're looking at true nominal values...
Each data point represents a single day. Usually there's just one trip per day, but sometimes there's no trips and a few times there's multiple trips. I've condensed the multiple trip logs into one log per day, and I've added days where there were no trips. This makes the x-axis consistent with respect to time. The labels are divided into weeks, so about a week's time per vertical grid line... The whole graph is about 7 months. The first half, up until about the end of October, reflects a period when I was letting the car do its own thing. The second half reflects a period when I was toying with the pack management...