How can I handle regenerative braking when the lithium-ion battery is full?

Question

Long version

We are using third party motor driver with regenerative braking and a lithium-ion battery. Apparently, this combination is well known in e-bike enthusiasts world as having a huge problem of BMS disconnecting the load while braking with full battery. The driver manufacturer simply pointed out the statement in datasheet that lithium-ion batteries are not supported and we are basically on our own.

While browsing the web I've found many industrial solutions, mostly in VFD context. Nothing small enough to fit into wheelchair. There are also references to chopper circuits for dynamic braking resistors. These are sometimes included into motor drivers with IGBT for external resistor. This seems to be not very good idea unless very high chopping frequency is used to prevent BMS from activation. Another manufacturer suggested a diode with braking resistor. This will, of course, waste the energy and substantially reduce the range.

Short version

Is it possible to create a relatively simple circuit that will limit regeneration voltage by directing part or all current via BJT to dissipate heat on the transistor itself or transistor and resistor combination? The goal is to have it activate only when battery is fully charged and not waste regeneration otherwise.

I am thinking something like a combination of Zener diode, OPAmp, diode, BJT and resistor can do the trick. Not sure if I can come up with a complete circuit on my own, so any advice will be appreciated.

Update

I forgot one important requirement. Because the battery (and hypothetical circuit in discussion) cannot tell the difference between regeneration and charging, the circuit should have additional inhibit input that would disable it while charging. Normally this is done by shorting inhibit pin to the ground, but we do not have an access even to that pin. We might be able to get a signal "charger connected" from the motor controller.

Also here is some additional info asked about in the comments and answers that might help. The battery is 24V 7 cell Li-NCM. We do have an access to BMS configuration, like maximum current, charge and discharge voltage limits etc. Motors are brushed DC, 20A continuous, 40A peak. They have built-in parking brakes disengaged electrically by the controller when power is applied.

Finally, here are some examples of the commercial devices performing this function: from Roboteq and UniDrive. They are expensive, not designed specifically for our application and seem to be permanently out of stock. I was hoping that we can make much simpler device fine-tuned for the job.

Update 2

Many thanks to all participants in the discussion. I'll try to summarize the suggestions below and maybe we can pinpoint the best approach. It seems that all workable solutions can be grouped into the following options:

Option A. Chopper circuit with IGBT or MOSFET controlled by PWM from MCU
I believe, this is how the commercial shunt regulators are made.
Pro: With right gate driver and transistor most of the heat will be dissipated on a brake resistor.
Con: Potentially excessive noise; requires MCU supply; software development time; cost.

Option B. MOSFET controlled by Comparator
While very different from A schematically, I suspect this will behave quite similar to it, with voltage oscillating around threshold voltage, producing chopping effect.
Pro: Simple; heat dissipation on the resistor; cheap.
Con: Could be even noisier than A.

Option C. Linear regulation of the brake resistor current, controlled by over-voltage
This can be done using OPAmp or BJT pair in Darlington configuration, as suggested by Jasen.
Pro: Reduced EMI; relatively simple;
Con: OPAmp requires either dual supply or biasing; Substantial heat dissipation on BJT requires heatsink; Jasen's circuit might be tricky to adjust for threshold voltage.

At this point I am favoring either B or Darlington. What do you think?

Update 3

This question already has an answer, but I just wanted to give a little update to anyone interested.

Winter is here and we've stumbled upon unforeseen problem of BMS switching off charging FET when battery temperature drops under 0 deg. as per cell operational conditions.

So now this circuit is essential not only as a dump for excess regeneration power when battery is fully charged, but as actual braking resistor in cold weather.

Answer 1

this seems entirely possible, when the battery is getting kind of full (as determined by voltage) turn on an extra load until the battery is less full.

if it triggers above the maximum cell voltage it will need to be a fast acting circuit, perhaps something like this:

^{simulate this circuit – Schematic created using CircuitLab}

Here C1 will activate the transistors during rapid voltage increases and the zener diode will activate during over voltage.

Size the resistor R1 to handle the peak current output of the regenerative braking system, the bottom transistor Q2 will need heatsinking sufficient for 1/4 of the resistors peak output. the trigger voltage will be about 1.3V the cener voltage of zener diode D1

Answer 2

You have hit upon a well-known problem in Li-ion traction batteries. It the battery is full, it can't receive any more charge, so it disables charging.

The correct solution is to provide both regen and mechanical brakes in the vehicle. If the battery is full, the mechanical brakes are used instead. Also, the mechanical brakes are used at low speed, when regen is less effective (regen cannot hold a vehicle stopped on a hill, for example).

However, I know that some vehicle manufacturer skimp on the mechanical brakes, which is quite dangerous. I have seen it in golf carts, forklifts, and the "GolfBoard" vehicle. The better golfcart manufacturers include a big-ass resistor to absorb the regen energy when the Li-ion battery is full, in addition to mechanical brakes.

---

To your question: how to build a circuit that absorbs the regen energy when the Li-ion battery is full.

If the BMS has an output signal that is active when charging is disabled, then you use that signal to engage a contactor or MOSFET, which connects the big resistor across the motor controller.

Otherwise, you need to detect that the voltage at the battery input of the controller is excessive, and use that to turn on the contactor. (Detecting the direction of the current would be effective if there were any current, but there isn't because there's nowhere for the current to go). As a detector, you can use a comparator IC (not an Op-amp: that's not what op-amps are for), a voltage reference IC, a resistor divider, either a contactor or a MOSFET, and a high power resistor. For an e-bike, I would use a 500 W resistor: not cheap. I have used toasters as cheap high-power resistors in the past. The problem is that there is a time lag between when the voltage is excessive and the contactor or MOSFET turns on. During that time, the high voltage from regen may destroy the motor controller. To handle that, you can add a high-power transient voltage absorber/suppressor (TVS) that only has to work between when the voltage spikes and the contactor/MOSFET comes on.

Answer 3

Well, let's take something braking down in two seconds from 10m/s of speed with a mass of 150kg (which includes passenger weight, so it would be a rather light wheelchair). That means we need to burn off 3.75kW average over two seconds. You don't burn that amount of power in a semiconductor; you burn it in a device intended for burning off power (or storing it, like a capacitor or a flywheel). The most compact device are brake pads and/or disks. Those can work while sacrificing material and/or keep working while red hot. Resistors tend to be less effective in those categories, but a lot more resilient than semiconductors at similar scale.

Answer 4

You can also simply limit the maximum battery charge at the charging station to 90-95% of full charge, in order to have 5-10% as a reserve for braking. Not to mention that avoiding 100% charge is good for battery life as well.

For instance, a 1kWh battery charged to 95% will have 50Wh reserve capacity, enough to absorb the energy of 180kg body going 100m down. That's 1km of constant braking on a 10% downhill slope with no losses.

Answer 5

Yes, you can make a circuit that, when the battery is full, redirects power to a brake resistor. You can derive functionality from VFD's where a brake chopper is used to regulate the DC Link voltage, eg: the voltage on positive side of the motor mosfets.
You probably can't do it exactly like that because your battery is your DC-link voltage.

For this you will need two power mosfets, a comparator, and a signal that the battery is full.

One mosfet is in series with the battery (source) to BMS (drain), with the diode pointing to the bms.
When full it disconnects and prevents charging of the battery.
The second mosfet can be a simple n-channel as brake chopper.

The comparator controls the brake chopper and turns it on when voltage on the Battery-BMS mosfet (Vds) becomes positive (eg: bms > battery).

Challenges will be finding a mosfet capable of the current between the battery and BMS, it will also introduce additional losses. There will also be some hardening required to filter the noise generated by the chopper. And depending on the amount of energy you need to take on, it may need cooling on the brake chopper.

---

The battery switch will also need some boundaries to operate within safe limits, you can't turn it back on before at least discharing some 10 percentage again. In short the BMS vendor is right, this is difficult and requires significant additional effort to make it safe.

Answer 6

So, without worrying about the possible circuits as they have been well explained, you could use an eddy current brake.

Basically an electro-magnet on s spinning disc so that regen current is used to brake the vehicle. Many applications use them so lots of examples exist.

Answer 7

This complements (and overlaps) other answers:

Energy will be wasted at any time that it is dissipated rather than used for regenerative charging. This applies whether braking is mechanical, or dissipated in an external resistor or in the motors.

@User107063's and Dmitry's answers (and others) of reducing battery state of charge to allow burst of regenerative braking, PLUS using a PWM modulated resistive load to bleed off battery capacity to provide a reserve for regenerative energy absorbtion, will do what you want.

The PWM on/off ratio can be very easily varied to provide a variable factor of N x the resistor used. eg a 10 Ohm resistor with 1:3 PWM (25% on) is the equivalent of a 40 Ohm resistor. The switch (probably a MOSFET can be hard on or fully off and dissipate almost zero power. The battery state of charge must be kept below the point where it can absorb a maximum braking burst. The dissipating resistor needs to be able to dissipate ALL the average braking energy over a period of operation.

Shorting the motors will dissipate the energy IN the motors.

PWM shorting the motors will vary the braking power (and force).

Driving the motors actively against the direction of actual or potential motion will provide low speed braking and static holding.

Overall, not having any mechanical brake is not a marvellous idea.

The battery needs to be capable of absorbing regenerative energy at a rate that does not exceed it's specified charge current. eg a 120 kg total load descending Dmitry's suggested 10% slope at say 2 m/s (7 km/h) will dissipate mgv x 10% = 240 j/s (ie 240 watts power.) That's probably well within typical LiPo battery spec.
A rapid stop will be similar.

A 500 watt resistor is not overly physically large and may prove an attractive alternative. I used to test exercise bikes controllwers that I designed, with a maximum required dissipation of 500 Watts continuous. A fully aircooled (no fan needed) air-wound Nichrome wire coil did the job well,

Answer 8

(First time using the schematic editor, so disregard all values; I just left the defaults.)

You can usually use the motor itself to absorb the energy - the stator could handle short-term loading while shorted. You might have to check with your motor supplier. I've used this technique in the past, but I had access to the internal workings of the output bridge (inverter of the drive) and could use the IGBTs to short or "chop" the current in firmware. I suggest using a contactor if you don't have that luxury.

As a compromise, you can use power resistors to burn the excess energy. The value depends on how much heating your motor can handle during brake mode - lower value resistors = more aggressive braking. The limit, of course, would be a complete short.

^{simulate this circuit – Schematic created using CircuitLab}

Answer 9

There's an interesting electric-hybrid truck company building a prototype, and documenting it on youtube.

Edison Motors in Canada describes it in

https://www.youtube.com/shorts/EJJj_aTvg7g

In this short they mentioned using an air-cooled resistor bank for electric braking, and then realising their generator can also work as an electric motor, so backfeeding to the diesel means they can use a "jake brake" or against engine compression.

Elsewhere he has bemoaned that they couldn't use a high voltage Jacob's Ladder electric arc to discard the extra energy to the sky - that could be spectacular.

Another place to "dump" excess energy is into water via a heating element, but water is heavy and when that boils you can't keep adding heat.

Answer 10

Is it possible to create a relatively simple circuit that will limit regeneration voltage by directing part or all current via BJT to dissipate heat on the transistor itself or transistor and resistor combination? The goal is to have it activate only when battery is fully charged and not waste regeneration otherwise.

Let me try again.

You have not provided a schematic of how the system works at present. Nor do we know the model of the motor controller, the Battery Mangagment System, or the battery (which might or might not have it's own integrated electronics). As a result, answers will involve guesses, assumptions, and speculation. Reticence is not the way to get the best answers.

The EMF generated by a permanent magnet DC motor depends upon the rotational speed of the motor. If the EMF generated by a motor is used to charge a battery, either there must be some mechanism to boost the voltage from the motor to that of the battery, or the rotor speed of the motor must be high enough to generate that voltage directly.

The OP has stated in comments, that, as far as they know, there is no hidden boost converter, and that the battery is recharged directly with the voltage supplied by the motor.

From this we conclude that regerative braking in this vehicle occurs only when the motor rotor speed, and consequently the vehicle itself are going fast enough. That is, there is a vehicle speed, below which regenerative braking does not work. Furthermore, that speed is not insignificant. It is generally not attainable with an open "throttle" on level ground. That is a consequence of the EMF generated by a motor at a given speed will be less than the applied voltage (i.e. the battery) needed to reach that speed, (on level ground). But, to charge the battery, the EMF needs to be at least the battery voltage.

So, if the wheelchair does not brake regeneratively, pushing current into the battery at "lower" or "normal" speeds, how does it brake?. The OP informs us that there are no mechanical brakes other than a "parking" brake, which causes an undesirably sudden stop if engaged while the vehicle is in motion.

It appears that the wheelchair does not yet have a resistive brake. That is, a brake that uses the EMF of the motor to drive a current through a resistor or some similar dissipative load.

So how does the wheelchair brake normally? Unlike bicycles, most personal mobility vehicles can be powered in reverse. For that purpose, they generally have an H-bridge motor drive. If the actual speed is less than the commanded speed, voltage is applied to the motor one way. If the actual speed is greater than the commanded speed, voltage is applied to the motor the other way.

Slowing down a motor (aka braking) by applying power to the motor which creates a torque in the opposite direction that the rotor is turning is called plug braking. That is probably the method the motor controller uses when the motor's rotor is spinning below the speed necessary to charge the battery.

Plug braking causes high currents to flow through both the battery and the motor. Lead acid batteries are generally fine with very high currents. However a BMS for a Lithium ion battery may well decide the battery is drawing too much current and disconnect the battery. That is one possibility that must be addressed when substituting lithium batteries for lead acid.

The current drawn from the battery during plug braking may not be continuous. It may be pulsed in the H-bridge. However, even if the average current drawn might be within the safe range for a Lithium ion battery, the BMS may still decide that the peak currents are not.

Does the wheelchair brake "normally" when it's velocity is below regenerative braking speed? Does it disconnect the battery if the speed command is brought too quickly to zero? We don't know, but it would be good if the OP could inform us.

That brings us to an interesting question. If the motor controller uses plug braking at "normal" speeds, and regenerative braking at high speeds, it there an a way to disable such a transition? If so, we could use plug braking when the the battery is full or the velocity low, and regenerative braking when the battery is not full, and the velocity is sufficient for regenerative braking. But, alas, we know nothing about the motor controller, so we don't know if implementing such a scheme is feasible.

In a previous version of this answer, I proposed using chopper controlled resistive braking using a resistor rated for high power and a MOSFET. However, if the hypothesis that your motor controller uses plug braking at speeds below those sufficient for regenerative braking is true, we must ask whether the motor controller and the resistive braking system would play nicely together.

If the OP would kindly provide more information about the motor controller, couch as its manufacturer, its model, and datasheet if available, and likewise for the BMS and battery, we can perhaps resolve some of these issues.