# Using maths to solve power problems

Peter Grindrod is working on a pilot scheme in Bracknel to examine how maths could help improve the networks effectiveness...
13 May 2014

## Interview with

Peter Grindrod, Oxford University

## Big_Bend_Power_Station.jpg

So we can improve the grid, but with added fluctuations from more renewable energy being used, we have to make sure that our power networks work efficiently to predict and manage energy use.

Mathmatician Peter Grindrod from Oxford University is working on a pilot scheme in Bracknell in Berkshire to examine how we can improve the networks effectiveness.

Kat - So, can you basically very briefly explain what is the current problem with how our network is working.

Peter - Well, we're working on a project with Scottish and Southern Energy Power Distribution who own the network or distribute energy through towns like Bracknell in the south of England. The problem that we've got is that people are going to be putting more and more technology, more and more demand, and more and more generation onto the network. So, the old wires that were put in there for one thing and they're going to have to be used in a much more flexible way. We're particularly thinking about low voltage networks. So, that's the last mile if you like from the substation to your house. And on such networks, there aren't lots and lots of customers that might be 100 houses on a network. So, when we look at their profile, it's not smooth at all. Every time somebody produces a demand like turning their dishwasher on or their kettle, actually, that produces spikes.

Now this is an opportunity really because over the next few years, we're all going to get smart meters put into our homes. So for the first time, we'll have some idea of the different patterns of use by different households. When we look at such patterns in typical residential areas, we find that some people are very volatile and some people are very predictable and do the same thing every day. Well, if you've got some means of storage and we've heard about some storage with hot and cold gravel, or you've got just a hot tank of water in your home. If you've got some means of storing energy then you'll be able to take advantage of that situation because it's likely that the price of energy will vary during the day. And now, you're able to monitor or rather just a smart meter is able to monitor your usage, actually, what it can do is take energy out when it's cheap and then give you that energy back. But it must do so, just as we heard from Argonne, from Guenter, it must do so in a way that we have a passive customer. We can't have everybody worrying about when they're going to use energy, when they're going to turn the cooker on.

Kat - I was going to say you know, I'm coming home from work in half an hour. Once I've done the radio show, I want to turn the cooker on. It needs to know that I'm coming home. So, how can the kind of stuff that you're working on, how can maths help to predict this kind of demand?

Peter - So, that's quite challenging. It's a really good example of modern maths actually. What we do is we want to write algorithms that monitor what you do from day to day and from week to week and gradually, that chip which sits in your home nice and securely will learn what you do. And so, it will automatically take care of those predictions of high spikes and those predictions of times when you're using less energy. Really, the problem for maths is that we don't want a sort of medium forecast. We want to predict where the spikes are going to become, how big they're going to be and in particular, we'd like to predict them slightly early so that then we can control any storage or any other devices we've got on the network.

Kat - But I guess the challenge is, I for example don't do this radio show every week. So, how do you guess that I'm not going to use the cooker at half past seven?

Peter - Well, so some of your demands will be fairly fixed. A lot of people have to catch a school bus in the morning and some of your demands will be extremely discretionary and won't be fixed. But actually, quite a lot of usage is predictable. The difficulty is that you're getting in on living your life. So, we have this notion in Bracknell of looking at lots of domestic users, trying to figure out how many of them are so volatile that no amount of forecasting and no amount of energy storage or time shifting demand shifting would be useful. And how many customers and how many substations are going to be able to deal with this in a very profitable way. It's a kind of win-win really.

Kat - So, I was going to ask about the Bracknell projects. I mean, why Bracknell?  I used to live there and how does it work? What are you actually doing?

Peter - Okay, well we chose Bracknell because it's got a bit of everything. It's not a space tent like Milton Keynes or something like that. It's got lots of house in the states and it's got some older properties which are sort of mixed or what we say heterogeneous. They've got a mixture of small retail and housing. Another of course, it's also got some big business parks, many companies that part of our project have their headquarters there. So, we chose it because it had lots of typical areas. And so, that we could meter up some residents. Of course, eventually everybody will get a smart meter in the UK, but what we've done is intensively meter up some residences so that we could have a look at how many substations, how many local networks are in trouble and might need smart management. And of course, there'll be others that aren't in trouble at all. They've got lots of capacity. At the moment, we don't monitor those so we don't really know. Of course, at the same time, people are getting electric vehicles, they're getting solar power, they may be getting heat pumps. But there's going to be new technology invented over the next 5 to 10 years which we haven't thought of yet. And all of that is going to have to go on these low voltage networks and kind of domestic consumers. They're just going to expect it all to work seamlessly. So, what we want to do is to put a sort of smart layer in that helps them do that.