Tag Archives: EV adoption

Testing Tony Seba’s EV Predictions 17 (More about Batteries)

In my previous post, I suggested that we are on the cusp of putting electric vehicle (EV) range issues behind us. Two distinct technologies are overcoming the range problem: the growth of fast charging networks and the rise in the energy capacity of EV batteries. In this post, we are going to drill further down into the energy capacity issue.

In the past, the battery constraint has been size and weight. Producing a battery that can deliver 400 or 500 miles of continuous driving is relatively easy: you just make the battery ever bigger. The problem has been delivering the required energy capacity within a sensibly sized and weighted unit. The Tesla Model 3 battery comes in at 478 kg, contains 75 kilowatt hours (kWh) of energy and can propel the car 310 miles between charges according to the US Department of Energy. In the Tesla Model 3, for every 1 kg of battery we get 157 watt hours (Wh). This is called the specific energy of the battery, or the gravimetric energy density, and is measured in watts-hours per kilogram.

I mentioned that the Model 3 has a 478 kg battery. We are really talking about the battery pack here, which incorporates a number of battery modules, which in turn incorporate a number of battery cells. As is the case of many things EV, we are frequently faced with the problem of comparing apples and pairs. That is, if we want to compare specific energy figures between vehicles, we need to compare like with like: battery pack with pack, module with module or cell with cell. The building of a battery, from components, to cells, to modules, to the pack can be seen in the illustration below (source: here):

ValueChainEVBatteries

The combination of the battery elements is a complex interlocking process involving a lot of different disciplines such as chemistry, electrical engineering and mechanical engineering. And it also involves trade-offs. Securing specific energy gains in one area can result in losses in another.

BatteryProcess

For example, the Tesla Model 3 uses state-of-the-art Panasonic ‘2170’ battery cells that are likely the highest specific energy battery cells deployed in mass production cars. (Note that the 2170 number represents the dimensions of the battery cell not the battery chemistry; 21 mm is the diameter and 70 mm the length.) But the battery chemistry employed in these cells is quite difficult and requires a sophisticated cell management and heat control system to prevent thermal runaway; i.e., the battery catching fire. Obviously, the more sophisticated and complex the cell management system, the more the overall battery pack is bulked up.

Of course, by definition, battery cells have a far greater specific energy (gravimetric energy density) than the battery pack since all the battery pack parts surrounding the cells have zero specific energy. In the Tesla Model 3 tear down that I referred to in a previous post, Jack Rickard extracted the four modules that go into the Tesla Model 3 battery pack. They are slightly uneven in size. Two of the modules contain 1,058 cells and two contain 1,150 cell, so the overall battery pack has 4,416 cells in total. Jack also weighed the modules: in total they came to 362 kg. So with usable energy of 75 kWh, the modules alone have a specific energy of 75 kWh divided by 362 kg, which translates to 207 Wh/kg (Jack blogged about this tear down here). From the top of the post, remember that the specific energy of the battery pack in its entirety was 157 Wh/kg.

We can go even further down to the specific energy of the individual cells. Before we do that, here is a short video of a Tesla Model 3 ‘2170’ cell being dissected:

Surprisingly, I’ve struggled to find an official weight for each individual battery cell. From a Tesla forum conversation, I have seen a figure of  66 grams quoted, but I can’t verify this. Until I get a definitive number, however, I will run with 66 grams as it likely to be only a few grams out. So if we have 4,416 cells each weighing 66 grams, that gives us a total weight for the cells alone of 291 kg. This time, let’s divide the total energy of the battery pack, 75 kWh, by this new figure. The results is that each cell has a specific energy density of 257 Wh/kg. Compared with the theoretical maximum specific energy density of around 400 Wh/kg, you can see that there is the potential for some future efficiency gains but not transformational ones.

BatteryMooresLaw

In the Model 3, Tesla has a car that can compete with ICE rivals such as Audi, BMW and Mercedes, but for Tesla to utterly dominate all its competitors it would be helpful if we could get its driving range even higher than 310 miles between charges. How easy is it for Tesla to do that within the existing battery chemistry limitations highlighted above?

First, let’s focus on the non-cell components in the battery pack. We already established that the battery cells weigh 291 kg in total. If we take that number away from the overall battery pack weight of 478 kg, then the non-cell components weigh 187 kg. Let’s say that through mechanical and electrical engineering incremental improvements, we reduce the non-cell weight of the battery by 12% per annum for three years; in other words by roughly one third. That will free up about 65 kg out of that 187 kg.

Next we allocate that 65 kg to install more cells. So the cell weight goes from 291 kg to 356 kg. That’s a 22% increase. If we hold the specific energy of the cells constant at 257 Wh/kg, we now have a 91.5 kWh cell pack that should give us a range of 378 miles.

Turning now to the battery chemistry, we recognise that specific energy improvements are harder to achieve in this area than improvements in electrical or mechanical engineering. So for the cells, let’s assume Tesla and Panasonic improve the specific energy by 6% per annum for the next 3 years. That will result in specific energy going from 257 Wh/kg to 306 Wh/kg, an improvement of 19%. With our 19% improvement, the battery now goes from 91.5 kWh to 109 kWh and range improves from 378 miles to 450 miles between charges. At a 450-mile range, I declared in my previous post that all worries over EV range would disappear.

In a perfect world, I would like to not only get up to a 450-mile range but also shrink the battery weight and size. But for that, we probably need to wait for a jump in battery technology that delivers specific energy north of 500 Wh/kg. There are a variety of advanced batteries in the pipeline that aim to do just that as can be seen below. Nonetheless, we have yet to see any that are close to commercial production.

FutureBatteryChemistry

The conclusion of this post, nonetheless, is that even with only incremental improvements in existing technology, EV powertrains (plus their batteries) are getting very close now to matching or exceeding ICE powertrains (plus their fuel tanks) in every single area of performance. Throughout this discussion, however, I have left out one critical parameter: cost. That will be the subject of my next post.

For those of you coming to this series of posts midway, here is a link to the beginning of the series.

 

 

Testing Tony Seba’s EV Predictions 14 (Deconstructing the Car)

In my last post, I set the conditions that determine whether the auto market tips from internal combustion engine (ICE) vehicles to electric vehicles (EVs). EVs need to either match or exceed ICE vehicles with respect to every car ‘attribute’ at an equivalent cost. Then it’s game over for ICE vehicles.

The attributes of a car give a consumer happiness. That utility comes from a) the mobility a car provides, b) the aesthetic of the car (the pleasure the owner gets in owning the car that is not related to other people) and c) status-signally through displaying ownership of the car to other people (such status signalling is not restricted to investment bankers and their Ferraris; it also covers hippies in their Citroen 2CVs and Green Party members in their Nissan Leafs).

The purchase decisions of consumers are based on their current budgets and future budgets. Current budgets determines how much they are able to pay for cars and future budgets determine how much they can afford to run their cars (fuel, maintenance, depreciation).

If EVs are better with respect to some of these aspects of the purchase decision but worse in others, then taking market share from ICE vehicles will be an uphill slog. That is what Exxon Mobile believes as illustrated in this chart from its latest “Outlook for Energy: 2018”:

ExxonMobileEV Forecast

If such a projection is correct, around 50 million EVs will be on the road (which includes pure battery and plug-in hybrids) in 2030. That compares with the Tony Seba S curve view of 130 million EV sales alone in that year.

To tease out who is likely to be right, let us think of the physical limits auto makers have to work with. Basically, a car is a three dimensional irregular cuboid shape constrained by such external factors as lane width and parking space size. Certain things are then put into this irregular cuboid shape to provide the mobility, aesthetic and status-signalling attributes we identified before.

Lots of car ‘stuff’ is not a function of whether it is an ICE vehicle or EV. For example, the headlights, wing mirrors, windscreen wipers and so on. We can exclude such items from our analysis since an EV can match the ICE vehicle in these domains. Moreover, there is no reason why an EV can’t match an ICE in terms of aesthetic or status-signalling should its design be good enough and its ability to fulfil the mobility criteria.

The main differentiator between an EV and ICE vehicle when it comes to mobility relates to the drivetrain. Taken, holistically, we can think of this as encompassing a store of energy and a means of converting that energy into motion. We can now compare EVs against ICE vehicles in respect of this broadly defined drivetrain across a series of factors, most importantly:

  • Weight
  • Volume
  • Efficiency
  • Flexibility

Given its position as the undoubted pacesetter in cutting edge EV design, performance and production numbers, Tesla’s Model 3 is a worthy champion for the EV camp. The standard Model 3 has a curb weight of 1,610 kg, while the extended range version is 1,730 kg. The crotchety Jack Rickard did a tear down of a wrecked extended-range Tesla Model 3 (warning: it is a long video) and extracted the battery, which weighed 478 kg. So that means the battery weighs roughly 28% of the car.

Let’s choose the BMW 330i Sedan as a typical ICE competitor for Tesla; its specifications taken from BMW’s USA site can be found here. This sedan comes in at 1,610 kg, so the Tesla is 7.5% heavier. Curb weight generally includes a full tank of fuel, which in the BMW’s case is 15.8 gallons or about 45 kg in weight (you can see here the extraordinary energy density of fossil fuels).

So that in a nutshell is the handicap of the battery as an energy source: more than 400 kg of extra weight. On the other side of the ledger is the fact that you wonder where the engine has gone in an EV. Here is the schematic for the Model 3:

Model3Schematic

First, you notice the radical shrinkage of the actual engine itself. An internal combustion engine is a system of controlled explosions that first translate into lateral movement of the pistons, which in turn has to be translated into circular movement to the wheels. That requires a complex multipart machine.

The video below compares and contrasts the Tesla drivetrain with a traditional ICE (but note it highlights the induction motor in the Model S; the Model 3 motor and electrical motors in other automakers EVs are somewhat different) and emphasises the fact that the electrical engine has radically fewer parts.

And here is a couple of minutes on the Model S engine showing its intrinsic simplicity:

The key differentiator, obviously, is the disappearance of a bunch of ICE components: transmission, exhaust system, fuel pump, fuel injection and spark plugs. An EV does need some kind of cooling system for both the motor and the battery, but this is relatively modest in both weight and volume.

Overall, if we take out the gas tank and the battery from the equation then we get this:

EV drivetrain weight <  ICE drivetrain weight

EV drivetrain volume < ICE drivetrain volume

But through adding the battery and gas tank back in, these inequalities reverse:

EV drivetrain weight >  ICE drivetrain weight

EV drivetrain volume > ICE drivetrain volume

Now, it’s very difficult to put numbers into these inequalities. But the interesting thing about Tesla’s Model 3 is that it incorporates a large battery in terms of kilowatt hours (kWh) but the car is still in the same ball park weight category as its ICE competitors. Moreover, we are currently going through a period of rapid battery cell shrinkage (weight and volume per kWh). Let’s say Tesla can shrink the 479kg battery that Jack Rickard extracted from the wrecked Model 3 by 25%; that would give a weight saving of 120kg. We are now getting into matching territory. And remember the conditions for tipping. EVs don’t need to exceed ICE vehicles for the market to tip: they just need to match in most areas and excel in a few.

Next we come to flexibility, which really relates to the configuration in our irregular cuboid. So yet again putting the battery to one side, the EV has an instant advantage. The drivetrain units can be arranged more flexibly as they are linked principally by wires, not by a complex transmission mechanism.

Even with the battery, the possibilities of dividing it up and distributing it around the car have yet to be explored. Safety and cooling issues not withstanding, the overall battery is cellular and is just composed of thousands of small batteries. We talk of form factors with mobile phones, and this is ultimately where we will move with cars. With an ICE, you have to design around the drivetrain, with an EV the drivetrain can become subservient to the design.

So then we move to efficiency, with respect to which the EV wins hands down. An electrical motor can deliver instant power and torque. In the US context with imperial units we have this equation.

PowerEquation

Which translates into this chart:

TorqueRPM

As a result of the dynamic in the above chart, Tesla is currently able to deliver supercar type performance at a fraction of the cost of the likes of Porsche, Bugatti or Ferrari (source: here). Note that Tesla’s new Roadster due to be released in 2020 will have a base model that delivers zero to 60 in 1.9 seconds; that will be the first production car ever to break two seconds.

CarAcceleration

Moreover, such high-end, halo EV performance profiles will trickle down. Ultimately, taking price to price comparisons, the EV will leave the ICE car in the dust when the stop light turns green. For those of a non-macho disposition, you may not care. But to repeat (again) if the EV is equal on all metrics but ahead on just one that you care about (all at an equivalent price), then you will buy the EV.

And yes we still have the constraint of range and price. And yet again this takes us back to the battery. Indeed, the EV battery is like the little Dutch boy Hans Brinker whose finger in the dyke is the only thing stopping the entire neighbourhood being flooded and his family and friends being drowned. But once the battery gets down to a price and efficiency point not far from now, that dike will go and the ICE industry with it. The battery is the subject of my next post.

HansBrinker

For those of you coming to this series of posts midway, here is a link to the beginning of the series.

 

 

 

 

 

Testing Tony Seba’s EV Predictions 2 (Setting Out the S Curve)

In my last post, I explained Tony Seba’s basic thesis as follows: he forecasts the complete transformation of the world’s entire transport and energy infrastructure by the year 2030. And while, Tony stopped there, I surmised that this disruption, if it takes place, will extend into every aspect of the social sciences. Indeed, for those like me who sometimes despair at the state of the planet, his forecasts could even prove a ray of hope with respect to the wicked problem of climate change. I don’t think it is hyperbole to say that such a transformation would be politically, socially and economically revolutionary.

At the heart of Tony’s thesis is the S curve: the idea that the adoption of technology follows an S curve consisting of three distinct phases: a gradual uptake, explosive growth and then a tapering off. His book “Clean Disruption” doesn’t really touch on the S Curve, but in his presentations this issue is front and centre. I highly recommend you watch the section of the video below from 7:50 through to 9:45 minutes.

This is the heart of his argument:

“No technology in history, successful technology, in history, that I know of have ever been adopted on a linear basis, ever. It gets adopted as an S curve.”

And Tony posits that S curves are getting steeper, with saturation points reached in years not decades as shown by the almost vertical lines for the most recent technology adoptions.

Adoption Rates

Therefore, if Tony is wrong, it will be with respect to whether EV adoption follows an S curve and what shape that S curve will take. OK, let’s start by fitting an S curve to Tony’s following prediction:

There may still be millions of older gasoline cars and trucks on the road. Ten- to twenty year old cars are still on the road today. We may even see niche markets like Cuba where 50-year old cars are the norm. But essentially no internal combustion engines will be produced after 2030.

Now an S curve has four parameters; that is, variables that control is shape. Bear with me: it is actually quite intuitive. More formally, an S curve is produced by a logistic function, which you can see examples of here).

From the chart below, we have the starting point ‘a’: in our case EV sales as a percentage of total global car sales, which in 2017 was 1.3% (I’ll come back to that number). We also have an ending point ‘d’. Now Tony says “essentially no internal combustion engines will be produced in 2030”. I have taken that to mean 95% of new sales in 2030 will be EV.

The inflection point ‘c’ relates to whether the growth will be front-end loaded into the beginning of the forecast period, or more back-end loaded into the end of the forecast period (or somewhere in the middle). Generally, it’s easier to ramp up production at the beginning, since you have fewer resource constraints.

Finally, we have ‘b’ the steepness of the curve. That really tells us whether all the growth is concentrated into a short burst; in the adaption curves at the top of the post, those curves which in effect go vertical, like that for digital cameras, have a high value for ‘b’.

SCurveParameters

Now because we can produce different curves to get from 1.3% penetration in 2017 to 95% penetration in 2030 it may take a little time to prove whether Tony is right or wrong in his projections. But by inspecting the shape of the curves, we can start to discern which of them are completely barking mad and which are mildly ambitious. So I will start with a curve that I have rustled up in Excel as the base-case scenario:

Seba Central Scenario

Under this curve, we start with a penetration rate of 1.3% in 2017 and end with one of 94% in 2030, with 50% penetration reached in 2025. Note that it takes 6 years to go from 20% penetration in 2022 to 80% penetration in 2028. Next, let’s increase parameter ‘b’ and get the curve to stand up.

SebaHyperGrowthSecenario

This is pretty damn aggressive. Tony is doing his victory lap in 2025 and the move from 20% to 80% penetration has taken all of four years. That is a lot of lithium, a lot of battery cells, a lot of battery units and a lot of EVs to bring on stream in short period of time. But note we could take that graph and shift it 5 years to the right. Under that scenario, Tony would still have bragging rights in 2030, but the curve would not go vertical until around 2025.

Now I am going to make the growth period a bit less manic in the middle, with a longer run-up by increasing the value of ‘a’.

SlowRampUpScenario

Now Tony gets to 95% one year late (I think we should be generous enough to give him that). Further, the EVs take over the world period (from 20% to 80%) now takes place between 2024 to 2029.

OK, time for some real numbers. Here are global EV sales and penetration rates from EVvolumes.com (as you can see, this is where I get my 1.3% starting penetration rate from in 2017).

EVVolumes.com

This adds a couple of new dimensions to our analysis: unit sales of EVs per year and year-on-year percentage growth rates. Keeping unit sales and growth rates in mind, we can take the theoretical underpinnings and parameters of Tony Seba’s EV S-curves, and attach just such real-world numbers onto the curves and see if they look sane. That will be the topic of my next post.

For those of you coming to this series of posts midway, here is a link to the beginning of the series.