How the Internet can Green the Electrical Grid

Uploaded by GoogleTechTalks on 16.04.2010

>>S. Keshav: Some of you have been out of university for some time. Casey was my student
two years ago. So I'm gonna start with a quiz because just to get you into the right state
of mind.
And I'm gonna show you seven pictures and I'm gonna ask you what's common to all seven.
And these are pretty diverse pictures. So here we go.
Here's the first one. So just look carefully at what it is. You don't need to give me an
answer right now.
This is the second one.
This is the third one.
This is the fourth one.
There's five.
There's six.
And seven.
Alright. So I'll do it again slowly bec, a little bit just because you guys look all
totally bewildered. You're supposed to be Google people; smart guys. So figure it out.
Here's number one.
Here's number two.
Number three.
Here's number four.
Here's five.
Here's six.
And here is seven.
So any [laughs] anybody wanna try, take a guess at what it is?
Sorry you missed the quiz.
You'll have to ask your friends later. [laughs]
>>voice in audience: [inaudible]
That's possible. [laughs]
If it's my quiz that's possible. Yeah.
So any, anybody wanna try? So I'll give you, I'll put names to it. So that's an, that's
a avalanche.
That's the Berlin Wall, you know that.
That's the Xerox Alto; the first PC.
This is Len Kleinrock with the INP which was the first node in the ARPANET in 1969.
This is a map of the Balkans in, just before World War I.
This is well, Das Kapital by Karl Marx.
And that's a boy plugging a dike in Holland. It's a sculpture of a boy plugging a dike
in Holl, in, in, in the Netherlands.
>>voice in audience: Oh, it's a sculpture.
>>S. Keshav: Yeah, it's a sculpture. It's a famous story about Hans Brinker this guy
who put a thumb in the dike to prevent it from collapsing.
So, okay. So--
>>voice in audience: [inaudible] it's just before a major tipping point.
S. Kesav: There you go. Now you know why he's the boss. [laughs]
So these are all things that happened just before a tipping point.
To go back to the beginning, this is just before the dike broke. This guy was kind of,
the, the, he put a finger in because if he let go the whole dike was gonna collapse.
This set off the Communist Revolution; enough said.
This is the powder keg of Europe; the Balkans before the World War I. One assassination
in Sarajevo 1914 caused seven years of misery to millions of people in Europe.
This is the beginning of the Internet. The first node, the Internet Message Processor
and somehow in 40 years you cannot build the entire Internet infrastructure.
This started the PC revolution. This was the Xerox Alto which had WYSIWYG. The WIMP interface
--you know-- windows, icon, mouse, pointers.
And this is the Berlin Wall; rearranged the map of Europe.
And this may rearrange your face if you get caught [laughs] in it.
So what's important about all of these? What's interesting is that this is what it looks
like. Basically this is accumulation of energy or potential inside the system. Okay? And
it reaches this crux and at some point a small push, the straw that broke the camel's back,
is going to cause an energy minimization with a large release of energy.
You can call this equal to nuclear fission is you want; [laughs] or all of these other
things. They all have this notion of ingrained, concentrated, large energy which is then being
pushed over. And my analysis is that there are actually three things going on.
There's this notion of Internal Contradictions. You know Marx talked about the the internal
contractions of capitalist society.
And similarly being a dike; the dike is keeping the water mass away. The Berlin Wall was,
was separating a nation which was artificially split. So in all these cases an internal contradiction.
Plus there're external pressures. There's some external pressure which comes from where
ever and adding to these two is a technological push. There is something that extra technology
that causes the whole thing to basically fall apart.
So what I am going to argue is the electrical grid has reached this. And I'm going to now
ask you to hold this thought in your mind for just a few minutes when I tell you what
the grid is and then I'll come back to this; explain to you why I think there's gonna be
major changes.
So let's start. What is the grid like? So this is like a idiot's guide to the electrical
grid. And what we have is basically on the left, on the, on my left, on the left hand
side we have sources of production. So we have coal; it's a dominance of production,
we have nuclear, and we have hydroelectric. These are dominant sources of production pretty
much in any country. You can also add in other things like natural gas and so on which I
haven't really shown.
The second component is called transmission. That's all the wires that you see over here,
all over the place; transmission. And it's a long, long distance.
And then finally we have distribution where the distribution corresponds to basically
is a substation that blocks over there and then the wires going out to each house; that's
And those are the three components of any electrical generation system; generation,
transmission and distribution.
Now the problems and the int, inherent contradictions and external pressures are apparent from this
beautiful figure which comes from the Lawrence Livermore National Lab in the U.S. and it
shows in 2008 how much energy was used in quadrillion BTU's, called quads. And you can
see that it's a sort of a spaghetti and I don't know if that's all visible, but I'm
just gonna walk you through it.
On the left hand side is generation and the right hand side is consumption, basically.
On the left hand side you can see the yellow box on top is solar, point, point 09 BTU's;
nuclear is 8.45; hydro's 2.45. So the boxes are not to scale but the lines are. Those
10 lines are to scale.
You'll see right away, right away it jumps out that the big black generation boxes are
coal and petroleum. Coal is 22 qu-quads and petroleum is 37 quads, and these are basically
Now most of the petroleum is actually going to transportation; that's your diesel and
your transportation, but ignoring that, most of the coal, in fact 90 plus percent of the
coal is going to electricity generation, that box on the, on the top over there it's using
39.97 quads in 2008.
So electricity generation is dirty. It's based on coal, at least in the U.S.
The second biggest component is natural gas, although it is closely followed by, sorry,
second is nuclear closely followed by natural gas. We have take coal and natural gas basically
are pumping carbon into the atmosphere like crazy. So electricity is not clean, it's very
That's the first big take away.
The second big take away is actually absolutely bizarre to me; it blew my mind away to see
this. All of the electricity generated which is 40 quads give or take, two-thirds or more
is wasted. Only 12.68 out of 40 quads that orange stuff; the gray stuff is just waste.
It's wasted in generation; it's wasted in transmission; and it's wasted in distribution.
Now transmission losses in the U.S. are not very high; about 7 percent, but transmission
loss in other countries are very high. India transmission losses are 25 percent, ranging
up to 43 percent in some areas.
Generation losses are because mechanical conversion of mechanical energy to, or nuclear energy
to electricity is inefficient. You can approach 40, 50, 70 percent, but it's still not 100
percent. Even nuclear energy which is kind of this energy of the future; all they're
really doing is boiling water. [laughs] I mean at the end of the day you do all this
crazy stuff and you're just boiling water. It's like a tea kettle on steroids. [chuckles]
So you're wasting a lot of energy; you're wasting huge amounts of energy is the gray
bar; so those are two take aways, ones from this picture. I would encourage you to actually
look at this picture and study it because you will get a tremendous understanding of
the energy system in the U.S., which is similar to other countries or about the same. For
example, in Japan and France, 70 percent comes from nuclear. But then nuclear anagram is
unclear what you do with the waste.
Here's the third fact: according to testimony at the, at, at a hearing on November 30th
last year by the Commissioner of the Massachusetts Department of Energy 15% of generating capacity
in Massachusetts. That's a lot of billions of dollars is needed fewer than 88 hours a
year. It's only pulled in at the absolute peak summer, less than one percent of the
So you're building this enormous capacity and you're just not using it. It's crazy.
This is what I mean by internal contradiction. This is what Marx was all upset about, and
so am I. [laughs]
So let's look at the structure of the system in terms of internal contradictions and external
pressures. So to begin with the technology's ossified. The last clever stuff done in the
grid was about a hundred years ago. After Tesla, Westinghouse and Edison, pretty much
anybody with brains moved on to somewhere else, like cars.
Interesting fact: Henry Ford used to work for Edison in Edison's electricity plant in
Detroit. And then he quit and started a car manufacturing firm. He was still around. [laughs]
So smart guys left; including Tesla who died penniless in 1943 which is terrible.
So rising energy prices; we are seeing a diminishing of coal and natural gas and cer, not coal
but ga, well oil and gas. Diminishing resource peak oil means rising energy prices. That's
an external pressure.
Energy security. If oil is sec, oil is coming from Middle Eastern countries where you're
propping up dictatorships in order to get access to their oil; invading them as the
case may be, it's not particularly secure. So that's an external pressure. Inefficiency,
enough said.
Global warming. Carbon footprint is gonna be a problem, is a problem, and once carbon
taxes go in which is absolutely unavoidable, it's a problem. So we are seeing tremendous
pressures on the grid from the outside and in-internally. Is that a question over there,
>>male in audience: I just had a one question. So most of those things are going up. Inefficiency,
is that not coming down?
>>S. Keshav: Yes, it is. Inefficiency is coming down but not as much as you'd like. So the
question was is inefficiency coming down? And I think it's, 'cause many of these inefficiencies
are built into the way you produce electricity. And built in the way to distribute things;
how many splices you have; what kind of transformers you have; how long your lines are; things
like that.
The systemic inefficiencies are very hard to actually get rid of. But yes, they are
coming down.
I'm just saying that inefficiency, you could live with inefficiency when the cost of the
inputs was low. When the cost of inputs goes up, you can't tolerate that anymore. So it's
just, tries to cut it out?
>>S. Keshav: So what's gonna bring it down? The structure of the energy goes up and then
something happens; it falls down. So what's gonna bring it down? I believe seven things
are gonna change.
So these are the technological push factors.
The first one is renewable energy. More than a trillion dollars of investment is going
into renewable energy around the world. Just you name it, solar, wind, geothermal. Google
is investing in basically [chuckles] putting in high explosives down a big pipe, shooting
down water, and pulling out steam; five kilometers down.
Anywhere in the world you can shoot energy, if, if, if you throw a bomb eight kilometers
down anywhere in the world, you'll get hot water. It's as simple as that. All you need
to do is dig a hole eight kilometers deep and you're done.
The same thing with tidal. People are building these snakes which kind of float and then
they go up and down and as they oscillate you just harvest energy from there. It's been
off the coast of Scotland, for example.
Wind of course is, is happening.
Communication. So this is what, sort of at the heart of what the rest of the talk is
gonna be about. But if you can overlay the infrastructure with communication you are
able to actually know where the eff, the inefficiencies are. If, if, if you can tell the home owner,
"Look if you don't turn off, if you turn off the AC right now even though it's hot, we're
gonna give you 200 dollars," you can save two billion dollars perhaps.
It's not necessary that everybody, so right now I'll just you an example of this. ASHRAE
is the American Society of Heating, Refrigeration, and Air-Conditioning Engineers and they have
Policy 55 which says: "The set point in a building should be between 28 degrees and
26 degrees centigrade and the middle point is 23 degrees. And every building in North
America if it's a triple A class office space it's set point at 23 degrees throughout the
year. Right now it's 23 degrees. Pretty, we measured it, I know. [laughs]
And in winter its 23 degrees and in summer its 23 degrees. Do you really care whether
it's freezing outside? Are you willing to wear a sweater? Because you wear a sweater
to come to work anyway, but ASHRAE will not permit you to put on a sweater in winter 'cause
you want to keep it at 23 degrees.
If you can communicate in a big sign saying, "Look, if you are guys are willing to wear
a sweater, come to work in the middle of February we'll give you a hundred dollar bonus" you
could save a of lot money. That's communication. Now there's a lot more to it I'm just giving
you some pragmatic examples over here.
Other things that, yeah.
>>male in audience: I, I pulled up that interesting energy usage slide-
>>S. Keshav: Yes.
>>male in audience: So just on the right hand side rejected energy and energy services.
>>S. Keshav: Yes.
>>male in audience: What's rejected energy?
>>S. Keshav: Rejected energy basically means energy that's not in use and once you heat
a car, car's engine by driving it where's the heat going? It's rejected.
>>male in audience: That's essentially the waste?
>>S. Keshav: That's waste. That's waste also. You could be putting a hamburger on it or
having a, having a burger when you drive home. I recommend to you the Radiator Grill Cookbook,
it's actually a real book. You just put your burger in aluminum foil, stick it in your
car, drive home, it's done. It's--
I'm not making this up. And there's even stuff to do with road kill which I won't go into.
Okay, so energy flows right now uni-directional. Basically everything is coming down into your
house. But if you had a solar cell you could push it back out again. So that's another
technology that's changing. Basically so looking at diodes you're looking at bidirectional
energy flow.
Storage is a big deal. Bill Gates in a talk recently said something absolutely amazing
'cause he'd gone and quantified it. He found that the total amount of energy storage available
in the world today is equal to 10 minutes of worldwide production. That's it. So imagine
that the amount of storage available for YouTube was 10 minutes of production. You can see
there somethings out of whack. So a lot of-of investment is going into storage in many different
Ultra capacitors, nano, nano, bat-batteries with nanomolecules and then these, rooms this
size absolutely vaccumed with air with this enormous big gyroscopes which are spinning
like crazy. And by day when you, when you have cheap energy, by night cheap and then
you spin it up and by day draw it down. People are doing that. Yeah.
>>male in audience: So what's the definition of storage? All your natural gas is in storage.
>>S. Keshav: Yeah, so storage here basically means for the most part storing electricity
during off peak time. Natural gas of course is storage but it's hard to create it, right?
So, so but whereas all of these other things are, are reversible reactions.
One of the cheapest ways is to pump water up and down. So it goes back up and it comes
down again, right, so people do that as well. Com-compressing air is another thing you can
do. So all these technologies for storage are coming and technological push factor.
I just think storage is coming.
Metering is coming in, smart metering. Almost all of you by now have now a smart meter in
your house, I do. What you may not know is that it forms a secure WiMAX mesh to communicate
with the local utility to tell you exact, to tell them exactly how much electricity
you're using. Every few minutes or every few seconds, I don't know what this metering interval
is. We don't have access to the data as homeowners, but this is being collected. So that's happening.
Plug-in hybrid electric vehicles. This is a big deal and the reason is very simple.
The amount of energy consumed in a day is approximately 20 kilowatt hour; for a typical
American home 20 kilowatt hours.
The Chevy Volt has a battery and the capacity of the battery is 20 kilowatt hours. If you
buy a Prius or one of these plug-in hybrid ones you'll get four kilowatts hours with
it. All electric cars basically have one day's worth of storage. Which means if some disaster
struck and you cut off your electricity for one day, you'd be fine as long as you could
use your car, plug it in.
But other interesting things open up over here which is what happens if you have two
neighbors and one has a lot of cars, back up electricity in the car and the other one
doesn't. Or it's a hot day and you're home for whatever reason you can use your car.
You can actually use this storage. PHEV is just another form of storage that you are
buying on behalf of the electricity company basically.
So you're, you're in debt to the car loan companies paying off the electricity bill.
So that's very interesting fact.
Another point that's was interesting is that you can actually carry energy in your car
and I can give you energy, I can say, "Oh, you're running short here, I'll give you one-tenth
of a battery" and you drop it off in the driveway and you can actually trade energy this way.
So that's changing. That's really a big deal.
Just to give you a sense of this, the amount of energy density in a lithium ion battery
in this laptop is greater than that of dynamite. You can look it up. Energy density, energy
per cubic centimeter in a lithium ion battery is greater than dynamite. Why they allow you
to take it on a plane is beyond me.
And finally, high voltage DC and super conduction. We are actually building transmission lines
between Philadelphia and New York which are superconductors. They have no resistance whatsoever.
They are cooled down to whatever it is, 240 Kelvin or whatever to make it work.
So these are things that are changing; changing the electrical grid. And I, I won't have time
to go through all of them and, and neither do you. But this is the bottom line. This
is what it's gonna look like in the end, at least my visualization. Instead of coal we're
gonna have wind farms and solar farms and hydro will stay there and it, at these substations
those orange boxes you'll find next to them I have shown a large battery. This could be
fuel cells. Ballard which is a Canadian company based in Vancouver has got these large fuel
cells. They're pretty big. And you can use them to produce electricity. Recently there's
this company out of Silicon Valley, I forget the name right now, who said they've been
powering EBay using a fuel cell as well.
So that's happening.
The nice thing is you can store stuff at night and release it so you can smooth out, modulate,
the energy so that if 15 percent of capacity that you were wasting you don't need to build
out anymore because your peak average ration has gone down essentially.
And also in each house expect to have solar and wind. Why is it? Well Ontario has some
most aggressive and most forward looking energy policy in the world, including Denmark and
Germany, it's actually quite amazing.
You, if you put a solar cell on your roof you'll get 82 cents per kilowatt hour, a feed
in tariff. No matter how much you consume. Every kilowatt hour produced about 82 cents.
If you work out the math you get a loan from the bank for 30,000 dollars and you put in
a solar cell in your roof, you will actually make 16 to 22 percent rate of return on your
money, which is way more than Scotia Bank is gonna give you.
And I'm not making this up. You can go to or, I don't remember,
and they will work with you to put a stuff on your, put a solar cell on your roof.
Or you can go to San Diego Farmer's Market on Saturday and next in the peddlers there
is a guy who's sitting there who will put up solar cell for you ten dollars a kilowatt
hour, sorry ten dollars a watt, not kilowatt, ten dollars a watt. So a kilowatt is 10,000
dollars and you can get money back. So you, you, I worked out the math. You basically
if you're gonna stay in your house more than six years you're gonna make money hand over
fist six years from now. You pay off in six years.
So this is gonna change what's happening here. You'll see full page ads in the paper very
soon. I'm predicting like next few weeks. You'll see ads from banks saying, "Come into
our solar program, you'll take, you'll put this on your roof and your gonna get, make
money." It's like having a renter except its cheaper and their quieter.
So this gonna change the way Ontario rooftops are gonna look like dramatically in the same
in way that if you ride, drive through rural Germany everywhere you find windmills. Why?
It's the same thing, it's cheaper, you make money on it.
>>male in audience: So I was just wondering what regulations are like here, right. One
thing one of my friends has said he is really mad with Palo Alto Power power goes out.
>>S. Keshav: Yes.
>>male in audience: 'Cause he had solar panels on his house.
>>S. Keshav: Right.
>>male in audience: He was fully able to self-sustain.
>>S. Keshav: Yes.
>>male in audience: Except the controller isn't allowed to run on a solar panel.
>>S. Keshav: Right.
>>male in audience: [inaudible]
>>S. Keshav: Right. So there are obviously issues with respect to this. Now there are
people who will put in a complete off grid system with local storage; lead-acid local
storage not lithium ion. Lithium is too expensive right now.
But regulations are changing. In fact the biggest probable regulatory hurdle we have
in our data right now is not the hydro company. It's insurance. Because this thing of parking
a BMW on your roof [laughs]
thirty thousand bucks. What if somebody steals it when you're gone on vacation? [laughs]
[chuckles] I'm thinking it's a non-trivial problem. So yeah some of these things have
to be, they will be crossed because there's tremendous energy stored up in the pan.
This is my prediction: the next decade will determine the structure of the grid in 2120.
So what the grid is gonna look like in 100 years from ten years from now, a hundred ten
years from now look, we will decide what's gonna happen, what's gonna happen.
what are the problems? I've, we've painted this vision but there are lots of problems.
I'm gonna just mention a few things.
One big problem is suddenly instead of 150 sources of energy or 200 sources of energy
which are the cold plants and nuclear plants. You have tens of millions of sources. Everybody
and their brother has got a solar cell on their roof and that's exactly what the Palo
Alto problem comes in because they wanna control it. But they don't know how to control ten
million sources. This is a big problem. 'Cause you wanna maintain your liability; you wanna
maintain what's called dispatch ability, that's a problem.
Second thing is the sources are not constant bit rate or constant energy there, they're
variable bit rate, means they go up and down. The sun goes away every day. Every night there's
no sun. Whoops. It just turned off your power plant so what do you do? If the cloud goes
over the face of the sun you can lose up to 95 percent of your energy in about five seconds.
If the wind stops blowing you lose hundred percent of your energy in about a minute.
So when you have these kinds of variable stochastic sources the power engineers are just kind
of wondering what to do because they, they've never been taught this stuff. They know how
to do transformers and winding patterns and stuff like that, but you show them the stochastic
source and they say, "we never learned the math for this one."
So this is all approaching like crazy. So the rule is for every watt of solar you need
to put in five watts. So just build five watts that's why it's so expensive, one of the reasons
it's so expensive.
Two-way flow is easy, but these are hard to get right and I-I won't go into that too much.
Non-utility nuclear is basically means you and I utilities, no. Just like they're putting
a Wi-Fi access point in your house you become an ISP. By putting a solar cell you become
a utility provider. But you're not playing by the same rules as the utility providers
play, which is whatever API's they use for controllability you're not really doing that.
How do you get reliability? There are issues of multiple time scales. This means your control
protection switching happens at 20 millisecond time scale. At the same time you need to plan
out transmission lines which happen on the time scale of 15 to 20 years. If wanna put
a line between here and Owen Sound for example to get wind energy from Owen Sound to here.
Just think about the regulatory hurdles to get land allocated to you along the way, along
Route 6. It's a pain in the neck. Plus you, so that happens in 15/20 years; its very expensive.
At the same capability, short term things. The cloud came over the face of the, cloud
cover moved into Ontario all of our solar went away, what do we do? We've gotta come,
somehow get stuff in from Quebec, for example. So that's a problem.
Incentivization. How do you get people to do the right thing? To wear sweaters in winter,
for example if they're expecting to do 23 degrees throughout the year.
Security. Well, [chuckles] it's the Internet.
Storage. When you put storage into any system it causes delays. When you have delays in
control it's a mess because you have exactly first order oscillator behavior; you put dampeners
you get flow surges. I mean its, its control with delays is a pain in the neck and storage
basically adds delays. So this is a well know problem in control, it's a difficult problem.
In, in addition to the supply changing over time the demand is also changing over time,
right? Because now what you're doing is you're putting in to like peak load pricing, whoops,
you're prices went up because it's 8 o'clock instead of 7:59, right? That's what peak load
pricing is coming into us this November I believe we're gonna have peak load pricing,
or when ever it is. So suddenly your demand is variable as well because people will all
turn off their dishwashers when the price goes up and so suddenly the demand sinks.
So the demand issues that have been predictable suddenly changing as well.
The other thing is resources are remote. What do I mean by that?
In order to replace one megawatt of electricity from solar you need 110 acres of land. [chuckles]
You're not gonna be able to do that in your backyard anytime soon.
So these renewable energy sources are primarily gonning be in the rural countryside. But that's
not where the supply, the demand is. The supply's in the countryside and the demand is in Waterloo
and Toronto. So how do you get it from here to there? It's not obvious. Maybe use cars
to drive it in.
And finally we have the legacy. We have 110 ten years of crud. Where, where ever it is,
you have to basically deal with that.
So these are all problems. [laughs] And if you go read the reports written by the Commissioners
in charge of electricity or maybe talk to Hydro One people and so on, this is what they
look like.
So what can we do? So now that we have a few minutes to talk about my [laughs].
So I think this is end of part one. Part one says, "Things are gonna change in a big way
and there's nothing that you and I can do other than to watch it." It's gonna happen;
I can predict it's gonna happen, I'm pretty confident of that.
Part two is we know something about the Internet. And guess what? What we know is the Internet
matches the grid. So I'm gonna go through some similarities and differences. And then
the end I'll sort of paint a research hypothesis. What can we do?
So what are the similarities? Interestingly enough the Internet and the grid historically
started out the same way. It was bottom up at first. People put up their own little LANs.
Or there was this WAN but most people had their own ethernet LANs and said, "So let's
hook it up." It kind of grew up from the bottom. In the same way people started building their
own electricity grids because there was, there was a windmill, sorry there's a water source
In 1881 the first generated, generation was put up in Niagara Falls; it's a, it's a DC
generator actually in 1881. This was before the grid, right? This was just some guy wanted
to run some machines and he got some DC machines from Edison so he put up a DC machine, a DC
generator in Niagara Falls. So it was bottom up. And then slowly they said, "If you diversify
sources you can actually get the power of diversity. So let's use a grid." So it was
bottom up.
And then top down it came we must have a grid. We must have a national grid. We must have
this. So in the same way the Internet grew up from the bottom and then became a top down
initiative once the politicians got into the picture.
They're both vast. I mean they're everywhere, right? Pretty much. Especially with cell phones
today Internet is everywhere because you can carry it in your pocket, you get 3G, or satellite
phone. They're heterogeneous, right?
Electricity grid already showed us different kind of sources, different kinds of users.
They're both critical to society and they're both ossified. The Internet, a quarter of
the Internet is stuck in 1973. It's never gonna get out of 1973. And the core of the
electricity grid is stuck in 1892.
I mean, if you see a picture of a transmission line or a generator or whatever from 1895
and you look at 1995, it's the same thing. You could guide somebody's who's dead a 100
years, bring him back to life, and say work on the electricity substation. He'd know exactly
what to do. Nothing has changed.
Incidentally if you get somebody who died in 1950 and you brought them to your kitchen,
they'd know exactly what to do. Other than the microwave nothing has changed. Kitchen
technology is also ossified, by the way.
So it's fascinating how many things have ossified. [laughs]
I could go into a whole riff on that, but I won't.
They are hierarchical. The, the transmission core is actually a mesh, for obvious reasons.
But the distribution is basically tree-like because we don't really care about redundancy;
whereas in the core we care about redundancy. And a mesh-like core uses its own technology.
We use high voltage transmission, all right. And the distribution is low voltage. It's
probably like in the thousands of volts rather than hundreds of thousands volts.
So in the same way in the Internet core we use MPLS, right? Nobody uses IP. If you want
to strip out all the headers and do something fast and smooth and sleek, but in the access
network we use IP. So we have tree-like access network. We have hubs and very little redundancy.
So it's example of the, this is the U.S. Transmission Grid. I couldn't find out one for Canada.
So this is the transmission grid as of 2007, I believe. Now you can see what it looks like.
And this is the Internet density of IP ad, IP addresses. This is roughly similar. It's
pretty much what you expect; where there are people there's electricity and there's information;
one flows energy; one flows information. They are very similar in that respect.
There are varying degrees of control. Now you think of the grid as being very much under
control, but actually what you plug into your grid at home is not controlled. You can plug
in anything you want as long as you obey the API which is two pins or three pins, that's
it. As long as you obey the API and you don't overdraw your current limit, you are fine;
you can plug whatever you want.
That's what allows innovation to proceed in the same way as IP's are narrow waves, the
three-prong plug is a narrow waste of the electrical grid. And, and, and the control
beyond that is very, very loose. In fact the Internet has more control on the edges than
the electrical grid, in, in some ways.
There's storage in both networks. The storage in the grid is far less than what you'd want.
And there's simple API to both of them, right? To write a app on the, on the electrical grid
you build a toaster. [chuckles] And that's an app. You can do whatever you want. And
well you know how to build apps on the Internet.
And they both use, they're basically solving the same problem. There's a bunch of distributed
demand sources, demands from people and they're distributed sources, generators and so on;
you wanna match the two. So, so resource matching problem it's a standard resource allocation
problem and you do whatever's necessary. You build a standard tree and you do whatever
you need to do to get things going.
And finally, or not finally, there's this balance of centralization and decentralization.
We wanna centralize some things and then decentralize it.
For example, and let's focus on just this one thing. In the Internet AT&T and Verizon
and these big guys provide long-haul transmission and then they attach the Tier Two's which
provide basically access, an aggregation. They don't really want to give service to
every end point.
In the same way, Hydro One connects hundreds of local distribution companies, like Waterloo
North Hydro. Waterloo North is not generating electricity, it is providing you access. They
are your ISP, your ESP if you want, electricity service provider and they are decoupled from
the back point; this is their transmission. So the same kind of truth tasting is going
There are differences. One big difference is electricity doesn't have headers. [laughs]
I mean it's kind of silly to state it, but it's really very, very profound because headers
carry two things that are important. One is type and one is destination. So, on the Internet
I can say this packet is of type X, whatever that type is. And it could be a TCP packet,
UDP packet, whatever I can put there in the headers. I know what to do with that.
In electricity, electron is electron. Similarly in the Internet I can say, "I'm gonna send
pic, I'm gonna send electricity, or a packet down this wire and I'm going to put the Mac
address of that access, of that host over there and it's gonna go to that one host."
I cannot address a single light bulb. So if somebody's studying in that corner, I have
to send electrons to all the lights over here, because they're not addressable. The only
way I can control them is to have one circuit per bulb, which is too expensive.
So this addressability fundamentally allows efficiency. If only communication in the Internet
was broadcast you'd be in trouble which is why Wi-Fi sucks. 'Cause that's, it's broadcast
and so everybody gets everything and so it's not very efficient; besides interference.
So that's one difference.
The second difference is as I mentioned earlier electricity generation is primarily one-way
and whereas Internet is more, more or less two-way. So we have a difference over there.
The time scale of controlling the Internet is in seconds. You, or within milliseconds
or seconds, depending on what kind of control you're talking about. About scheduling it's
a nanosecond, if it's roundtrip, flow control it's milliseconds and so on and so forth.
Whereas in the, things happen must more slowly in the grid. If you want to bring up a nuclear
reactor it takes two days to two weeks depending-- [chuckles] on what's going on.
You never have to wait two or three weeks to bring up a source on the Internet. You
can typically do it within seconds.
And then here's a very interesting difference. If you have a fiber optic link between here
and Owen Sound and you wanna go from 100 meg link to one gig link, you just change out
the lasers at both end and you start sending more lambdas, you're done. You can't do that
with the grid.
You, long-haul transmission is a constrained resource and you can't actually add more capacity
to it. Now with superconductors presumably you can do that, but without all the superconductors
you're basically stuck with whatever you have. So you need to put in more wires, more copper,
whatever it is.
The Internet is less predictable because you have flash crowds; whereas the grid is very
predictable. They can pretty much guess what's gonna happen.
And these are some kind of differences. So the bottom line is I think of the electrical
grid as being a content distribution network for a single video stream. It's a video stream
because its continuous generation going on and the they won't actually distribute this.
So you, so let's take the case of somebody trying to grab electricity under their car
and driving it to somewhere else. It's like downloading a file onto your memory stick
and going and giving it to your friend. That's the same exact thing as downloading electricity.
That's a single video stream because there are no types; it's all the same.
So which means that we can start applying the same tricks we use for CDN's for the electricity
grid. We'll come to that in just a minute.
So this sort of concludes the part two of my talk and I'll just pause and take any questions
at this point.
Okay, so let me move on.
So what I'm gonna do is to make this hypothesis. And the hypothesis is that the research and
technologies developed by Internet, by Internet researchers over the last 40 years can be
used to green the grid. So it's in green.
So it's a hypothesis because I haven't proved it. I'd like to spend the next several years
proving it, but I'm going to give you some taste of why I think it's true and then I'll
sort of conclude.
So I must state that this is not the same as these two other things. I'm not saying,
"Reduce electricity usage for Internet data centers." Now it's great to reduce electricity
usage anywhere including Internet data centers, but let's face it it's only using one and
half percent of all electricity produced. It's not a big deal.
Heating and cooling uses 40 percent of all energy; 20 percent for commercial; 20 percent
for domestic. So if you affect that, you're far better off than affecting the data center.
Now for Google Internet data center is a big deal, but for the rest of the world basically,
I don't care. [chuckles] You could burn all the electricity you want, you're really not
doing much harm, but every single building, every single restaurant that keeps its door
wide open in summer with the AC blasting, now they are a big deal. Because there's millions
of them and they're all completely messed up. So we need to figure out incentives to
make them do the right thing.
The second thing that people have talked about in green networking is to use the Inter-Internet
as a communication overlay. Now I mentioned there's already smart metering. And yes, that's
a great idea, but it's seems to me it doesn't go as deep as one would like. You can go far
deeper by actually reformulating the grid in terms of Internet research as I'll show
you the next few slides.
So one idea is local matching. So I said the grid is a CDN Content Distribution Network.
What is it that corresponds to delay? Now what we want to do in the CDN is to reduce
the delay, right? We wanna get the content down as fast as possible. Well the longer
the electricity travels on a wire the more transmission losses there are because the,
the, you have "i square r" is the heating of the wire. So each kilometer the electricity
travels on the wire you wasting en-en, wasting, wasting electricity; so minimizing delays
isomorphic to minimizing transmission loss. As I showed you earlier these losses are fa-fairly
So what you want to do is to perhaps use pier-to-pier cooperative caching to reduce losses. How
does this work?
So you'll take your Android phone and stick it in your car and it forms a wi-wireless
mesh network with all the other Android phones in everybody else's car and they talk to each
other and say, "Hey, how much electricity do you have? This is how much I have. You
wanna do a swap?" And you have this chit chat going on and at the end of the day they control
the appropriate diodes and so on, and electricity flows out from the cars, through the plug,
into the substation, and down to somebody else. And at the end of the month you a bill
that says, "You just made 200 bucks because your neighbors all wanted your electricity."
Not bad.
Just because you happen to work near the hydroelectric power plant and they have lower rates over
there you can bring electricity to your neighbors in your car. And you could do this. I mean
this is technologically feasible today.
So we have to just use the same ideas in cooperative caching that you all know about; which we
know about and the grid people are clueless about. [chuckles]
Tomography is the determination of the traffic matrix from a number of observations. You
have CAT scans. CAT is just computer aided tomography. So you take a bunch of x-ray images
and you figure out backwards, you do a sparse matrix inversion, you figure out where everything
is; in the same way the Internet tomography, you look at a bunch of aggregate points and
you figure out the aggregate flows or the traffic matrix flows. From some source as
to some destination D, how many bits were sent over some period of time.
So right now maybe we can do the same thing with grid usage. We can just measure substation's
use inserting smart metering for home. Is there a way for us to figure out in what the
sparse matrix and not have to meter everybody? Maybe we can do it through tomography. So
I'm just throwing this out. I have no idea; [chuckles] I mentioned this to somebody from
Hydro One and they were not immediately aghast. They said, "Oh, maybe we could do this." And
they actually are going to give me access to their data. So I'll know what you guys
are doing. [chuckles]
So as I mentioned earlier solar and wind are stochastic sources; they're variable bit rate
sources. Now one thing that the Internet people have done is to model stochastic sources like
crazy. I mean I can point to probably 200 papers on, on, or maybe more, 2,000 papers
on variable bit rate sources or to model them using any number of techniques: Markov-Modulated
processes; Autoregressive models, you name it, right? And the big question is the Internet
model is always the same. If I have 100 VBR sources going to the link what is the probability
they're going to overfllow the link? Should I accept them or not?
The exact analogy of this is this: you, you, you build a hotel; it has a hundred rooms
in it, with a hundred rooms in it; and the technician comes from the phone company and
says, "How many dial out phones do you want?" Now imagine this is before cell phones, right?
So you had to choose how many lines you have. If you put one line in for each guest room
outbound they can always call out from the hotel, right? That's great, but then it's
expensive; if you put in just one that's probably too low. What number do you use? Its workload
dependent, right? Its workload dependent and that's the same problem over here.
>>voice in audience: [uinintelligible]
>>S. Keshav: Oops. What did I do now?
Ah. Why did it go away? Or did I pull this other wire, it came out maybe?
Ah. It was a mechanical problem.
So under what combination, under what conditions is the sum of these VBR sources greater than
sum value? It's the probability the sum exceeds X should be greater than .9, five nines for
liability. So Hydro One operates on five nine's. So they wanna say that the baseload in the
province is whatever; 200, 200 megawatts, 250 megawatts.
So how many sources do we need to put in? How many solar, how much wind so that with
very high probability the sum of these sources exceeds the baseload? That's the answer they
want to get and it's workload dependent and it's stochastic and they don't know how to
do it and I think we can; we know how to do it.
Now remember this is very, very different from the kind of thinking you do to say, "Let's
turn off the routers and not be used." It's very different than saying, "Let's use the
Internet as a communication overlay." It's got nothing to do with communication whatsoever.
It's using Internet thinking rather than the Internet itself.
Delay tolerant networking basically says, "Look, I have this USB stick in my pocket,
right?" I used to have it, I don't know; I don't know what happened to it. Casey has
it, I guess. So let's just pretend, here we go. This is an Internet stick; it has 30 gig,
whatever, 8 gigabytes. I'll just make it up; maybe it's not. And it has eight hours of
movie; I put in my pocket, I walk home, I have 8 hours of movies.
Well, maybe you can carry energy like this. You, you take your laptop; take the battery;
stick it in your pocket; go home; and you can run your dishwasher. [chuckles]
So how do you do that?
So this is a very interesting question. Every time I click on a link, and I'm using Google,
assuming I'm using Google; you guys know what I did. You know what I clicked on and where
I went to. So you kind of computing the, the, the, the page rank of the, of of, the, sorry
not the page rank, but the you know. Basically you're computing a profile on me.
What happens when I turn a light switch on and off? That's a click as well, right? Now
imagine the following thing: let's say I can harvest the electricity clicks and somebody
is gonna turn their AC on and on a hot summer day when the peak load is very high. And everybody's,
newspaper and TV is broadcasting, "Please don't use your AC;" this idiot does it anyway.
What do we know about this person? They are rich. [laughs]
Sock, sock it to them; charge them a lot of money because they don't care. Or maybe their
teenage kid doesn't care, but whatever it is they are, they are indifferent to price.
So the demand elasticity they have in economic terms is very low. They are not elastic. They
have inelastic demand. If it's hot they're gonna turn the AC off no matter what. These
people also drive Hummers. [laughs]
Probably on the wrong side of the road as well.
But this kind of information is very valuable. Two companies who want to harvest information.
So imagine streams of electricity clicks coming in from hundred million end points into some
data source. You can imagine the data mining possibilities are out there. So people like
Ashraf who are data base people should probably view this as, "What can you do with this data?
What can you do?" Hopefully for the social good. I don't know about Google. [chuckles]
Now how do you get somebody to turn the light off. It's a game-theoretic problem. You have
to make it incentive compatible. You have to say, "Look if you turn the light out you
save money." So assuming you can do this, maybe there's a mechanism design problem.
Now we know how to try and get people to not download too many movies. You have a band
width limiter or some kind of thing from Sandvine that says, "If you use too much you're gonna
delay things." So this kind of thinking about the game-theoretic modeling of the negative
impacts; a tragedy of the commons is very important; it's a game-theoretic model. Again
the kind of ideas you developed on the Internet can be applied over here.
Any kind of blackout which means you lost, you, your demand was created in supply for
extended period of time; means more than five minutes; means you have a blackout or a brownout
where your phase goes down from 50, 60 kilohertz, you can go to 50 hertz or whatever. That's
the same as network congestion on the Internet.
So essentially all the kinds of distributed control algorithms that you have developed
on the Internet for network congestion can be used over here. To give us example of this,
what's the electricity grid's analog of TCP? TCP says if I see a packet loss it must be
me. It's the very, I don't know the psychoanalytic term for this, but [laughs] self-blaming view
of the world, "Whoops, must be me sorry". You know it's a very sorry behavior.
Analog for this electricity grid is if you overstep a limit your circuit breaker trips
off and you're not allowed to use anymore. But that's very coarse; it's very coarse.
It says, "If you draw more than 20 amps, oops, the fuse went off," and you go and do something
manually and you cannot do it. Because the equivalent of FTP is a short circuit. [laughs]
That's the equivalent of FTP. Give me everything right away; that's your short circuit. But
you don't wanna, wanna prevent that so just like we have rate limiters, we have circuit
You wanna make it a little bit more graceful. You want to probe and say, "Maybe the substation
has more energy I can use more right now to charge my car; maybe it doesn't and I'm gonna
back off." So if you want to have this modulated additive increase multiplicated decrease approach
to energy consumption, rather than just saying, "Well you wanna get five amps and I get, if
I can't get more than five amps, I'm gonna shut down." So there's some interesting analysis
that needs to be done over here.
And finally simulation. So we really, I talked to a bunch of people and said, "How do you
decide how much grid power you need? How much solar? How much wind? Depending on the stochastic
ra-radiations and so on a so forth." The answer is, "Well it's seat of the pants." There really
doesn't seem to exist any simulation. I've talked to a bunch of people about this including
the, the Green Energy, Energy czar at Google, Bill Wile, and that was awhile ago. But maybe
you guys have it now, but as far as I know there is no simulators.
One of things I'm tryin' to do is to build a continental level simulation of, let's say
Brit, North America. Simulate cloud cover, simulate wind movement, simulate 200 cities
and the variable demands. Simulate the movement of cars and transmission lines. We know exactly
where the transmission lines are, I already showed you a figure. That's a downloadable
data set; that you can download and run it.
So I have a student whose gonna work, working with me exactly on this. We're really trying
to simulate North America and see what happens. It can help us as a planning tool.
Again, this is using Internet style thinking. So I'm just about running out of time. So
I'm going to give you three thoughts.
First one: the decade 2010 to 2020 will decide the grid of 2120. My great-grandkids will,
will, will be using it and probably yours too.
The Internet approximately equal to the grid. It's not the same. We don't have headers,
we don't have a bunch of different things.
And I believe that 40 years of Internet research could, should, may help. At least I'm gonna
try doing it for the next some years so with that I'll end and I'll have any questions.
Anymore questions that come up.
[unintelligible speaking in background.]
>>male in audience: So I was thinking about these incentives in Ontario. You get some
rate for solar and some rate for wind and I, I noticed that the, the rate for wind was
much lower than solar. And it seems like that's, that's just an incentive that makes it economical
to build X, right?
>>S. Keshav: Um-hum, um-hum.
>>male in audience: Whereas in Ontario maybe Y is, is the choice because it, you get more
wind electricity per, per dollar invested--
>>S. Keshav: Yeah, yeah.
>>male in audience: than solar.
>>S. Keshav: Um-hum.
>>male in audience: It seems kind of, I don't know, a bit wasteful to, to use the, the money
on building solar panels when you could be wind, building wind turbines. Do you have
a comment to that?
>>S. Keshav: So I'm the wrong person to answer that. I mean I'm not setting policy initiatives.
I'm just telling you the policy initiatives have an effect and it clearly is born out
of the fact that you've looked into it.
So what should be the modeling? My, my answer would be this: I think it's an interesting
research problem from an academic perspective at least to understand the incentive structure
of a particular policy, and what actions can be expected as a consequence of that policy.
And I think we should kind of play it out and see what happens and say, "Predicted this
is what's gonna happen" and see if our predictions are correct; so that, that's what we should
be doing.
>>male in audience: One thing your, your simulation here is seems like it's something that, that
is likely in the future of the grid to be very financially lucrative--
>>S. Keshav: Yes.
>>male in audience: in a sense. If you can sort of identify these points where energy
prices are, are expected to fluctuate a lot in the--
>>S. Keshav: Right.
>>male in audience: grid you build your, your gyroscopes there--
>>S. Keshav: That's right.
>>male in audience: and you sort of arbitrage the, the energy market.
>>S. Keshav: Yes. Yes, they should be financially lucrative. I, I, that's why I say that I don't
know what Google is doing right now because if anything is financially lucrative and has
to do with information, you guys have your fingers all over it.
I would not be the least bit surprised to know there's a simulation going on in one
million computers in the middle of Nebraska or whatever. [laughs]
>>male in audience: Any questions for someone on a VC before they get cut off?
[clicking in background]
>>male in audience: So like the whole system it seems to like geared towards like more
like peer to peer incentives so it becomes like much more sources everywhere?
>>S. Keshav: Yes.
>>male in audience: Right? And with thinkin' about how to give incentives for this sort
of usage--
>>S. Keshav: Yes.
>>male in audience: but once you put like much more power into the nodes--
>>S. Keshav: Um-hum.
>>male in audience: of the system, like there's a question about abuse and control, security,
this sort of thing. So if I have incentive to give something back, I can try to gain
the system--
>>S. Keshav: Yeah.
>>male in audience: so that, you see what I mean, right? So--
>>S. Keshav: Absolutely. And you're, you're absolutely right and that's, let me restate
my thesis. I'm saying that the grid is gonna change. Not because of these kinds of issues
but because of carbon footprint; because of transmission losses; global warming; things
like that.
When it's changing and becoming more distributed these issues will come up. And my claim is
that the things that we've learned from Internet research ought to help us model and analyze
these things. I'm not saying that these problems don't exist. They do exist. They have to be
solved. And I think that Internet engineers have something to contribute over here.
>>male in audience: Okay, so basically just another thing to keep in mind.
>>S. Keshav: Yeah. Remember I showed you this, yeah, so where did it go. Let me see.
Yes. [laughs]
>>male in audience: Thank you.
S. Keshav: Yes, do you have more questions? Someone else?
>>male in audience: Did you still have a question?
>>male in audience: So do you have any recommendations about places where we can learn more information
about the grid how it exists today or the structure of the grid?
>>S. Keshav: Yeah.
>>male in audience: Like books, Website, whatever?
>>S. Keshav: Yeah, it's a great question. There are, so I have a Website called ISS4E
and I really should have put is up on my, if you just search for ISS4E, that's stands
for Information Systems and Science for Energy. And that's the research group that I'm leading
with Professor Catherine Rosenberg at University of Waterloo. And we have a section they call
resources and in that it's, it's a Wiki and in, in the resources section you'll find links
to all sorts of things that have come up including some speeches that Bill Gates and some projects
at Berkeley, for example, this has been focusing on this.
There are some books available. Probably the best one in the Internet grid is called I
think Electrical Grid Concepts by Alexandra von Meier, but if you will send me email I'll
be happy to send you a pointer to that book. Yeah, okay.
But there are, there are books available and there's a free book you can download called
Sustainable Energy - without the hot air, that's from a Professor at Cambridge whose
name escapes me right now, that's also really worth reading as well.
>>male in audience: A link to Professor Keshav's project site is in the abstract if anyone
needs to follow up later.
>>S. Keshav: Oh, good. [chuckles]
>>male in audience: So these very much seem to be first world solutions to this. But like
say carbon footprint in that, North America isn't the only problem. So how does, how do
we, like does this abstract into third world situations at all?
>>S. Keshav: Oh, absolutely. I mean my focus over the last seven years has been a rural
development in developing countries and what got me started on energy was to realize that
the biggest problem in villages is not lack of communication, but the lack of energy,
alright? And a lack of clean energy. Cooking done, is done with cauldron cakes in most
of India, which is an absolute horrible source of particulate matter; causes all sorts of
diseases. Plus those particles settle on the glaciers in the Himalayas and cause the glaciers
to melt causing floods, right?
So what you need to do is basically get them clean grid, cheap grid solar, wind. So these
issues of local distribution, local generation are not just over here, these are worldwide.
And if you can set up a, in fact in Indian is taking a big lead in solar for LED lighting
systems. So you charge it by day, then you have light go on at night using LED's, it's
a, it's a tamper proof, fool proof system that is being used in India. It's being used
in Bolivia, Peru, places like that.
So these issues come up. Now the bigger issues in our planning over here are relevant to
the North American context because that's what I'm talking about right now. I'm delivering
a similar speech in India in three weeks and it will be slightly different, but obviously
the focus is going to be on more local issues. But there is a big relationship between the
spread of electrical energy and rural health actually. So, so I think there are, there
are those kind of issues become relevant.
Now you can say, "What is the role of the Internet in this? Same thing; stochastic sources,
right? How many villagers need to have sun and wind in order for us to get the baseline
electrical volt for that village, even if they're off grid? We need to know that. How
much storage should be put? What is the sizing of the lead-acid cells? And if you over capacity,
if you over provision it, it costs you too much, right? Costs you too much.
So maybe you can have a siren system that says, "Everybody turn off your whatever it
is right now because the grid is going to explode in your face."
So, I mean these kinds of things they sound stupid but, but the early warning system for
the tsunami in South India was actually loud speakers connected to a cell phone which got
a text message saying, "Run for your life." And people-
yeah, it worked. Loud speakers work really well when you live in a dense environment
it works really well.
So these issues are not far from my mind, but they're not in this particular presentation.
>>male in audience: So like I'm thinkin' like, so we have the problem basically with the
current system right?
>>S. Keshav: Yeah.
>>male in audience: And like this seems like a pretty radical change.
>>S. Keshav: Um-hum.
>>male in audience: But I'm thinkin' about how cost effective would be like changing
the whole system comparing to maybe like finding couple things that are like really, really
bad in the current system--
>>S. Keshav: Yeah.
>>male in audience: Let's say like maybe doing one thing, like, like, relatively simple thing
like, I don't know, maybe like adding the storage, like massive storage point somewhere
>>S. Keshav: Yeah.
>>male in audience: So maybe it'll fix the current system so it'll work like another
hundred years--
>>S. Keshav: Yeah.
>>male in audience: it could be like more cost effective than changing the whole system.
>>S. Keshav: So all the research that I've done into this, and I am by no means an expert,
leads me to believe that the system is really at this creaking obsolescent stage, where
the people running it really know it's gonna collapse. And they're really scared. And they
really need help. I mean why else would a monopoly come to university and say, "Please
help us?" They're, they're worried that the whole thing is gonna just, as far as they
can make out they really want to, seem to want to make the change.
I'll give you example of this which absolutely amazing.
In Ontario there are two million wooden power poles; they're two million of them. Each of
them costs one thousand two hundred dollars to replace. So that's 2.2 billion dollars
in just these wooden power poles. Most of them are more than 35 years old and their
life span is 35 years.
[chuckles] So which ones do you change? Now it turns out the people at Waterloo are doing
research on determining which poles to change by taking a helicopter, flying over them,
and looking at the amount of surface rot from image analysis and determining with high probability
these poles are bad.
Other people are using ultrasound measurements. You have a ring of eight ultrasound transducers
you put around the pole and then they compute the ascitic CAT scan examine it with sound
and they figure out the internal structure of the wood and tell you whether it's going
to be falling in the next five years or not.
So these are the kinds of technologies which are coming and so to answer your point, systemic
change seems to be inevitable. It's being recognized by every energy authority in the,
on the planet. So I may be wrong, but hopefully not all of them are wrong at the same time.
Maybe they are better than the economists; [chuckles] who tend to be universally wrong.
I should say the diversity in opinion among, among economists doesn't seem to exist, whereas
with the power engineers that lack of diversity may be less. I have too many negatives in
my sentence. [laughs]
But you get, I hope you get the point.
And second thing is even small things cost a lot, right? Even the poles cost what, like
two billion dollars. Even a small, small nuclear reactor is five billion dollars; there's tremendous
amounts of money in this. Tremendous amounts of money in this.
Change infrastructure; it's a hundred years of infrastructure; hundred years of hundreds
of millions of people paying hundreds of dollars a month every month for the system. It all
went somewhere. [laughs]
>>male in audience: So, so basically you think that changing the system like pretty much
do like complete overhaul of the system in the long term it will be more cost effective--
>>S. Keshav: Yeah.
>>male in audience: then maybe fixing a couple like really big pains of point right now,
sorry points of pain.
>>S. Keshav: If you look at smart grid, that's what people are saying. And I'm inclined to
believe them.
>>male in audience: Okay.
>>S. Keshav: I-I can't say for myself, but fr-from what I've heard that's what I'm hearing.
>>male in audience: Okay.
>>male in audience. Yeah. Have you, have you thought about using existing hydropower plants?
They seem like they are becoming more valuable as you phase in more renewable intermittent
>>S. Keshav: Right. Perhaps--
>>male in audience: Because you can, you can fairly fast scale up the output from scale
the output of a hydroelectric plant.
>>S. Keshav: Yeah, but once you're, I mean there are certain number of turbines--
>>male in audience: Yes, yes.
>>S. Keshav: it's very hard to add more capacity. Yes, you can ramp up over five to ten minutes
you can spin up a new turbine and that's why there what's called pumped hydroelectric storage
is a big deal. You can take water and pump it back--
>>male in audience: Yeah.
>>S. Keshav: up the dam.
>>male in audience: But it seems that you don't even have to pump the water up, you
just don't run those when you have, when the wind is blowing basically. I guess there is
some percentage of your energy mix that is sort of, if you are above this level in terms
of hydroelectricity then you can sort of add more--
>>S. Keshav: Yeah.
>>male in audience: intermittent sources without having to worry--
>>S. Keshav: That, that's right. So the, the, what the power engineers called a baseload
and then the peak load. So the baseload is what you want to meet all the time and typically
it's met today by coal and nuclear and hydro, because these are long running sources and
that's okay. And then you can take the current of the extra deltas from, from wind and solar,
for example.
>>male in audience: So--
>>S. Keshav: Yes, so I agree with you that we want, we want to develop hydro, I-I don't
remember the number off the top of my head, but North American large scale hydro's at
the 80 percent stage or something like that; in that ball park, in terms of usage. All,
basically all rivers have been dammed.
>>male in audience: Yeah.
>>S. Keshav: Okay.
>>male in audience: But, yeah, so one thing here is adding more turbines to existing dams.
>>S. Keshav: Yeah. Yeah. And they're doing this. Niagara is building additional turbines
that's coming on stream I think next two, three years. And the other interest is in
micro hydro, which is a stream running through just put a small generator which is a four,
five, few hundred watts even, right? Can you, and, and that is combined with storage, that
makes a difference. You can take your, and people are putting solar cells in backpacks.
That's really tiny. The same kind of tiny micro hydro can also be built, is being built
and deployed. And so if you live by a stream you could potentially power your house from
there. So, yeah.
The, the problem with micro hydro tends to be it's a mechanical system. The turbines
run out, wear out they get clogged with algae and dust and who knows what. It's pretty gross.
Ever try cleaning out your bathroom sink anytime?
It's similar to that, but worse.
>>voice in audience: [unintellgible]
>>male voice in audience: So is that the, sort of is that where we are?
>>S. Keshav: Yeah, yeah, we, you could put a micro turf in.
Now if you're bald it's much better. Let me assure you of that. [laughs]
>>male in audience: [inaudible]
>>S. Keshav: Yeah. I mean you could generate, I mean people have talked about putting a
tiny wind turbine and you're holding it out of your car and charging your cell phone from
that, right?
There was a proposal about three years ago by a conceptual artist. So please take it
at conceptual artist level, of putting turbines on the middle of the New Jersey Turnpike in
New Jersey. And they kind of estimated that just a wind, the sheer wind from the two cars
moving opposite directions will cause the turbines to spin pretty fast. And that can
generate enough electricity for I don't know what, something interesting.
So, so yeah, you could, you could, all these kinds of harvesting can be done. I think that
it will be done. Once electricity costs goes to, let's say it goes from 10 cents a kilowatt
hour to 50 cents a kilowatt hour in two years from now. You'll be facing half the lights
will be turned off at Bell, I visited Bell Labs in, Bell Labs where I used to work and
I visited them in I can't remember early 2000 something, was it, I don't know. And these
guys, this is before they got bought by Alcatel. Lucent was saving money by turning off every
other li-li-light. Every other light thing was turned off to save electricity. And that
was Bell Labs. If, if electricity goes up by a factor of five you'll see that happening
everywhere. So, and these kinds of harvesting will become popular.
It's a question of money talks. And right now everything is cheap and it's gonna change.
>>male in audience: So I guess one more liability of the current centralized grid is its vulnerability
to deliberate attacks.
>>S. Keshav: Yes.
>>male in audience: And if-if you build your, your grid simulator I'm sure you could find
like one spot that you could go cut a wire and, and--
>>S. Keshav: Yes. Yeah,
>>male in audience: it would fall apart.
>>S. Kesheva: Yeah. About eight years ago I did some research where I identified the
500 data centers where 80 percent of the Internet traffic comes from. I could identify the,
the IP address and with a little bit of work the geographical location as well. And one
of the questions I got asked was, "Aren't they giving people a map of exactly where
to bomb?" [laughs] I never published that work.
>>male in audience: So will this, will this change with the, the smart grid--
>>S. Keshav: Yeah, distribution generally means resilience, that's the standard rule
in computer science.
>>male voice: Any last questions from VC participants?
Alright. I guess thank you very much for coming by Professor Keshav.
>>S. Keshav: Thank you.
[techno music]