PMSM control strategy 3

[jawerbuch]

Hi guys,

I have a question regarding the control of a PMSM via an inverter.
The idea is to control the motor's power according to the gaspedal-angle. So let's say it's a 85kW motor, pushing the gaspedal 50% would mean a power demand of 42.5 kW.
I've been looking at control strategies and I've bought the book "Permanent Magnet Synchronous and Brushless DC Motor Drives" by R. Krishnan. It discusses quite a variety of control strategies, like Vector Control (same as Field Oriented Control?) and Space Vector Modulation but I'm kind of lost on which one is the one I need.

I have also noticed that most practical examples are about controlling the speed of the motor, where I need to control the power of the motor. Can I instead control the speed of the motor, and calculate the necessary speed by using the motor's torque/rpm characteristic to determine what torque corresponds to a certain speed, multiplying those to give me the power? And what is the best control strategy to use in this case?

 

[jraef]

So just to clear this up...

Can both Vector Control (FOC) ánd Space Vector Modulation be used to control the inverter?

Kind of a problem with this question. Those two things are not really an "either and/or" scenario.

PWM is just about firing DC in pulses to change some sort of property in the DC to attain a desired output result. PWM can apply to AC or DC outcomes.

SVM is the switching method of the output transistors to make a circuit that the motor thinks is AC. It's the think that makes a VFD an "Inverter" drive.

Once you have SVM, then the pseudo AC output can be tweaked to get different levels of performance from the motor. That "tweaking" is referred to as the "control algorithm", which can be Scalar (aka V/hz) where there is no feedback from the motor as to how it is performing against the demand, or it can be "Vector Control" where the motor performance is monitored via a feedback loop so that errors can be corrected.

Vector control can then be further broken down into different levels of performance. SVC (Sensorless Vector Control) is simple but not as accurate, FVC (Flux Vector Control) is more complex and more accurate, then FOC (Field oriented Control) is generally considered the most accurate form of FVC (although the ABB crowd will want to argue for DTC (Direct Torque Control) as being equal or better). But since I do not claim to understand why DTC is supposed to be so much better (I have never seen it out perform FOC), I'll leave that to others. FOC can then itself be broken down into needing an encoder feedback, which is where you can attain the holy grail of AC motor control, Full Torque at Zero Speed; or "Sensorless" FVC, where you can get ALMOST there. Since it would be highly unlikely that you would need FT@ZS on an EV, let's assume you will use the Sensorless version.

With Sensorless FVC, the mP in the drive, using highly accurate current sensors, is able to fully separate the output current signature into the current that produces flux in the motor (what makes the motor a motor at all), from the current that produces torque in the motor (the current that does the work). They actually occur at slightly different times in the sine wave, so by using a very high speed powerful mP, you are slicing that "pie" of current into very fine pieces and tweaking the vectors of those two elements faster (compared to other algorithms) to affect a greater degree of accuracy and at the same time decrease the response time to a step-change in load, so that the motor corrects any error faster. WITHIN that FOC algorithm, there is a velocity loop AND a torque loop so that you can essentially control either or BOTH things virtually simultaneously. So you can do torque control WITHIN a velocity control loop, or velocity control WITHIN a torque control loop.

So how that applies to your question is, this gives you the best of both worlds. In a vehicle, you are deciding the speed you want to go, which is essentially a velocity loop between your sense perception and motor skills on the gas pedal. While you are performing that velocity loop, the FOC algorithm is going to AUTOMATICALLY deliver whatever torque is necessary (within the limits of the motor) to maintain that velocity.

As mentioned, that process is essentially "canned" in an off-the-shelf VFD (of any decent quality). Creating it from whole cloth on your own is likely going to be a daunting task. People spend entire careers developing these things (or lots of people with short pieces of their careers), funded by deep pockets of companies expecting to sell bajillions of them. If you can find one already made, it's likely the best thing for a small project.



"Will work for (the memory of) salami"

 

[iop995]

PMSM controll strategy (algorithm) is one issue, which may be made by various controllers like jaref said and control of EV speed is another issue, is the big loop closed by driver. Optimal control for PMSM may be FOC but is expensive and not indeed needed in EV applications (in EV must limit acceleration for security reasons, wheel slip, etc.), an open loop, simpler algorithm may be enough, depending of mechanicaly specs meeded or imposed.

 

[mikekilroy]

PMSM controll strategy (algorithm) is one issue, which may be made by various controllers like jaref said and control of EV speed is another issue, is the big loop closed by driver.

I have no clue what the subject of this sentence is; and I am curious.

WHAT is the big loop closed by driver?

http://www.KilroyWasHere<dot>com

 

[iop995]

Big loop is car speed loop. Inside it there are various loops that contoll/look-up for currents (motor / battery / brake - charging, etc.), anti-slip, torque-limit, inverter control, etc. depending of complexity / requirments of project. As motor choosed change, one or more of such internal loops are changes accordingly motor type / converter used, but big loop remain almost the same for a given car project.

 

[mikekilroy]

From my experience, ALL EVs I have studied, including my own, have only ONE loop: a current loop. Now I can't speak for any EVs I have not seen personally, but there are major reasons all are current loop drives only. I will try to explain this.

But first let me say there is but one control loop on drives for these PMSM motors. All the other things mentioned are just SWITCHES. current limits, regen braking, etc. OK,OK, chargers, antislip etc are loops, but not to control locomation so I don't count them in the drive control of the motor while rotating. Even the commutation is not a loop or any kind - it is just logic in typically an FPGA IC.

So why EVs do not use velocity loops.... again, as I mentioned previously, it is done due to gain and how people drive cars.

When you go down a hill in an ICE mobile, if it were velocity control your speed would not increase; ditto going up a hill you would not slow down. But instead you apply more or less gas pedal pressure to modulate the current (torque) going into the motor. Simple. It works. Same in an EV.

Now back to gain again. The car weighs 3-4-5-6,000#. This weight is reflected back to the direct driven wheel motor and it is 1,000's of times more than the motor inertia itself. You CAN put gearing in to reduce the reflected inertia, but the available ratios are from 1 to about 8 max (with normal diam tires and 8:1 gearbox, motor will be spinning at about 8,000rpm @ 80mph! So forget reducing the inertia down to within 1-5x the motor inertia. Since the inertia will be VERY much higher than the motor, if there is ANY compliance (spring) or backlash (there is always) then you cannot tune the velocity loop for the required bandwidth - or it becomes very very unstable when the compliance acts or backlash is seen. I will go over what bandwidth is required later, just suffice it here to say it can't be high.

Then there is the over 100 years of perception how a car operates. If it WERE possible, then people would have to relearn how to drive - going up hill would not require pushing more, going down not backing off the pedal. We tried a new scheme on my car to see if folks could get used to something similar in order to try to not have to modify the stock brake pedal, which requires more certifications and rules: we tried making the pedal apply more torque when pushed as normal, but then automatically regen brake when released from whatever the present torque was. So running down the road, if one let go of the pedal, the car tried to stop fast! If one shook their foot some, the car would jerk forward and back! This experiment did not last to day 2! SAME thing would happen with a velocity loop controlled by the pedal.

Now to bandwidth requirements.... when you step on the pedal, you expect action. So you will need to be able to saturate the VL output to command full torque in a reasonable time. What is that?

It is about 100 mseconds or 10hz. So you better have a 20-30hz BW in your VL. I guarantee with springs on the car between the wheels and 5000 load, THIS WILL NOT HAPPEN WITH STABILITY. Guaranteed.

You can probably get stable control around 1hz. So who is going to be willing to floor the pedal and wait 2-3 seconds for it to go? I am not standing in line to buy that.

But cars have cruise controls and they are stable you say! Sure! What is their BW? they respond VERY SLOW - in order to be stable. It can do this low BW since it only turns on at a given speed, and only has to control +/- a couple mph variance. THAT is the response you would get if you try to close a velocity loop around the drive's current loop.

http://www.KilroyWasHere<dot>com

 

[waross]

I prefer induction motors over synchronous type motors, but that said, If I had a large PMSM on hand I would be trying to use it rather than scrapping it and looking for a similar sized induction. So this is not a criticism of the OPs choice but a general question;
Mike, what do you see as the emerging choice in EV motors? Induction or Synchronous? Should I be changing my preferences?

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[mikekilroy]

I agree, Ac induction motor would have some really good advantages, main one I see is allowing use of sensorless vector or even v/hz vfd drive. I do not see anyone wanting to use sensorless control on a PMSM motor due to the weight (1#), size (2"dia x.5"L), and cost ($100) saving for questionable performance from and to a stop.

Unfortunately I think if used, the marketing used against it combined with the loss of mileage would reduce the potential sales by a factor of 10.

There is something called 'range anxiety' that all EV drivers have: it means we are to the point of paranoid watching the charge gauge; not having the option to pop into a gas station for a range extention, we are at the mercy of our batteries..... I have actually SWEATED in the past a few times wondering if my pushing it would leave me stranded. One time I could feel the motor shake due to dipping the available voltage below the required BEMF of the motor and the flat topped sine waves to the motor caused vibration if I stepped on it - in the last mile coming home.

With this need for more range, ANY loss is devastating. I'll make a guess that if I had the best designed induction motor possible I would loose about 10 of my 80 avg mile range.

My PMSM reaches 94% efficiency max; so we know an ac induction motor can match that, so it is not an efficiency issue.

It is a size & weight issue I think. There is no way to make the induction motor as small and lightweight as the equiv size PMSM motor with Neodymium magnets. In the past when we have compared both for specific applications where these two items were important, IIRC we could do the PMSM motor in less than 1/2 the size and weight of the induction. If that gets the driver with his range anxiety another 10 miles range, it is well worth it I think.

www.KilroyWasHere<dot>com

 

[waross]

Thanks Mike. lps
There is no substitute for information by someone who both designs and drives an EV.
That's the Eng-Tips advantage!
I'll work on changing my prejudices Re: induction Vs PMSM.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[iop995]

@Mike
Electric car is a system that need a driver, it's not autonomus. Driver impose a speed (desired/reference) and driver also sense car speed and adjust pedal to reach desired speed. That is big loop and don't treat it as a standard velocity loop. No need to tune this loop, it's indirectly "tuned" by driver (and internal loops/logicals that avoid unsafe situations). Inside it there are more or less loops that cover various internal blocks.
This big loop need low bandwidth (10-15Hz, or maybe faster for high-class Formula 1 driver ...) as you tell.
Some info about EV control and motors -

http://www.intechopen.com/download/get/type/pdfs/id/12061

 

[mikekilroy]

iop955, I really like your big loop description!

http://www.KilroyWasHere<dot>com

 

[cswilson]

Mike,

One of the key reasons that AC induction motors are popular for pure electric vehicles is that you can vary their rotor field strength widely and dynamically, thereby changing the torque/speed tradeoff, and eliminating the need for a mechanical transmission. The prototype Teslas had a 2-speed mechanical transmission that was failing quickly, and they got rid of that in favor of more aggressive field control.

Interior permanent magnet synchronous motors permit some additional variation of rotor field strength compared to surface magnet motors, but not as much as induction motors. A proper comparison of induction to synchronous motors must include the size and efficiency of the gear box/transmission.

 

[mikekilroy]

agreed curt - another good potential reason for ind machine.

but know folks are now taking the ipm's reluctance field to new extremes!

ck out the speed torque curve i attached a few posts back. i asked the delphi design engineer of that motor how he extended our typical +30% 'field weakening' ipm to induction motor levels but he did not answer. that 230v 3ph 2500rpm motor reaches 230v @ 2500 yet continues upto 11,000rpm max!

i just had to tell a good customer that, although we could run their siemens 1FE5093 (100 or so hp iirc) pmsm spindle motor rated 480v @ 2400rpm with top speed 7000rpm on our supplied test stand, we did not supply the required output crowbar voltage clamp so if they hit e-stop at max speed the motor would self destruct due to the Kb generated 1300v rms!

http://www.KilroyWasHere<dot>com

 

[cswilson]

Mike,

You bring up a good point. I don't know if I would want to sell or buy a car with that kind of failure mechanism, even with a good crowbar voltage clamp!

 

[sreid]

I think EVs will end up using Switched Reluctance Motors. This view is based on cost.

 

[DHambley]

jawerbuch,

You are heading down a good path if you're looking at SVM. Keep in mind that most all textbooks and papers describe SVM as a voltage-mode (VM) controller, i.e. the switch sequence results in an integral of V*t over each sample cycle equal to the voltage vector which you want to place onto the windings of the PMSM. An average current mode control (ACMC) algorithm is typically used to satisfy the torque command (remember torque is proportional to current). As you probably know, current (torque) lags voltage due to the series inductance seen by your inverter. The ACMC loop would basically attempt to follow the "gas pedal position" as a torque, as mikekilroy explained. A properly designed ACMC loop tells the SVM to put the voltage vector onto the motor which leads the back EMF voltage vector by an amount which will keep the current vector at the same angle as the field, for high power-factor and efficiency.




Darrell Hambley P.E.
SENTEK Engineering, LLC

 

[mikekilroy]

Sreid: I agree. They & their stepper type hi power drives just don't seem ready for low cost production yet. I have tried to use switch reluctance motors, due to some of their unique advantages, from the big players in the industry for a couple applications this last year and their price has always come in 3-5x the cost of induction or even PMSM - when the FANCY, special, and unique drive is added to the final cost.

http://www.KilroyWasHere<dot>com

 

[cswilson]

sreid, Mike:

I've been following SR motors for 30 years now, and there are definitely some intriguing aspects to them. As Mike notes, there are some "chicken and egg" issues in getting them and their drives into volume production compared to induction and PMSM motors.

They do have a key technical drawback compared to these other motors. Because they operate on a "vernier" principle - with the stator field counter-rotating with respect to the rotor and at a several times higher frequency, the iron losses in the motor, which are proportional to the square of this frequency are far higher than for the more traditional motors. The higher frequency also puts additional demands on the drive.

 

[Brad1979]

Just thought I'd link to this article about some of the reasoning behind why Tesla uses an induction motor.