Content

Chris, your recent work, including with the Alan Turing Institute, focuses on managing uncertainty in decision-makers’ models. When we talk about “uncertainty” in energy systems modelling, what does that mean? Is it technical, behavioural, climate-related, or something else?

If there's one unifying theme in the work I've done over the years, it's modelling to support decision-making in situations where you don't have a lot of relevant data in the traditional numerical sense. Sometimes that’s simply because you're dealing with the future, which, by definition, you cannot directly measure. Other times, it’s about rare events in the changing climate. By definition, we don't have a lot of historical data for rare events, because otherwise they wouldn't be rare. Or it might be new technologies. You might also be dealing with situations where people could have recorded some data, but didn’t. And then there are climate models and other model-generated data, where you need a layer of expert interpretation when applying them. So, for one reason or another, there are often situations where expert judgment is needed in the analysis. You then deal with how to systematically record expert scientific or decision-maker judgments about what matters. You’re applying a relatively conventional hard science to work through the consequences of these various sources of judgments and other evidence.

Can you think of any real-world examples?

At the Alan Turing Institute, we ran the Managing Uncertainty in Government Modelling project, which looked at some methodological aspects of these issues.  The project took place during the main COVID years, when our ability to interact with people in government was much more limited. You couldn’t drink coffee with people. Another example is the work I’ve done in electricity around the security of supply and the electricity capacity market. This work is closely related to REMA (reform of UK electricity markets, ed.) Here, you’re looking at how you procure the right amount of generating or supply capacity to ensure an appropriate level of risk of shortfalls. Similar to REMA, it’s about taking ideas from economic theory and applying them in practice, given the various complications the real world throws at you. There are also related points around cost-benefit analysis. In both cases, controversies around how well conventional methods of cost-benefit analysis are implemented in policy. In the case of the capacity market, it's always difficult to “monetise” rare but severe events. Another challenge is managing uncertainty in the background against which you're planning, which has been necessary in the REMA process. There’s often a lot of emphasis on the cost-benefit numbers, and less on the uncertainties underlying them. One major uncertainty that is very difficult to understand, especially if you're not part of the finance or generation development sector, is the effect of a reform like REMA on financing costs.

What exactly do you mean by that?

Take, for instance, the proposed zonal market option and the associated uncertainty around investments in renewables in regions where electricity might be cheaper. The costs of these investments are dominated by capital costs, which are fixed. If generators are directly exposed to the market, they face both price and volume risk. Under a locational market, you only get paid for what you produce, and the volume risk is particularly difficult to mitigate. This increases the uncertainty about future income, which inevitably drives up financing costs. The scale of this risk is sometimes difficult to understand for people outside of the renewable generation sector. At the same time, there are very good reasons to consider locational pricing. Production is simply worth different amounts in different places, and it costs different amounts to supply people depending on where you are in the network. This creates an interesting effect: doing something that naturally reflects the differences across the network affects financing costs. This cost increase seems comparable to the benefits that you get.

'The most important point is a shared understanding of the decision context between analysts and decision makers.'

Chris, some of your work has been on how to take extreme weather events into account. Let’s imagine a decision maker is involved– what are the right and wrong ways to present data to them?

I’d actually turn that question around a bit. One thing that often goes wrong is forgetting that communication between analysts and decision makers needs to be two-way. Often, in practice, analysts produce an analysis, and then the focus is on how to communicate it. But it’s essential to understand what aspects decision makers are particularly interested in, what their objectives are, and what sort of risk tolerances they have in uncertain situations. Only then can the analyst properly understand what they should be doing and how they should be doing it. One of the most important things for analysts is to grasp how the decision makers think about the decision. At the same time, it is equally challenging for the decision makers to understand that the relationship between modelling results and real-world outcomes is quite uncertain. To make good decisions, you want to confront these uncertainties head-on. So, I guess, I might answer a slightly different question and say that the most important point is a shared understanding of the decision context between analysts and decision makers. In the context of REMA, for example, tools like the PyPSA GB model, developed by my colleague Andrew, can help decision-makers access the analysed data in a transparent way, unlike when consultancies publish their reports and provide little detail about how these numbers were produced.

You and your colleague Lars Schewe have worked on real-time control of electricity systems, including dynamic voltage regulation and active power adjustments. Looking to the future of the UK grid, where fast digital control and coordination could be key, are today's operators and technologies ready for that level of responsiveness, or is there still a gap between what's possible in theory and what can be reliably delivered at scale?

Recently, Lars has worked on the technical side more than I have, although I’ve recently sat on the National Energy System Operators Technology Advisory Council. In general, there's a challenge around just how quickly we're looking to move. We’re certainly trying to force the pace on innovation in the sector more rapidly than technologies typically progress from basic research to deployment. I remember speaking with someone in the industry who understands this slightly better than me. He mentioned that technology often moves in 20- cycles. Whereas here we have these very ambitious targets for 2030 – which is not 20 years away. So, when you look at some of the scenarios for the 2030s, people really are relying on technologies that are not yet in widespread commercial use.

'The broad vision is that potentially every customer meter point could actively interact with the network. That could mean going from a few hundred to low thousands of entities directly interacting with the transmission network operator, to tens of millions that somehow need to be coordinated.'

What technologies are these? Can you give an example?

Well, there are questions about what will eventually happen with different nuclear technologies. Will Europe be able to produce new conventional nuclear reactors at a reasonable cost? What about the development of small modular reactors? There are also emerging technologies like carbon capture and storage. All of these are being discussed in the context of the 2030s. But to bring it back to your previous question, one of the big challenges is what Lars Schewe, Miguel Anjos, and I wrote about for the Energy Systems Catapult a couple of years ago: how can we develop an overall architecture for the operation of the system – one that can manage both the uncertainty associated with renewables and the complexity of having a vast number of local resources and network constraints? I think we need a somewhat different paradigm here. There are two types of networks: high-voltage transmission networks and local distribution networks. The former aren’t particularly complex, in the sense that the number of large entities connected to the transmission network isn’t going to grow by orders of magnitude. Local distribution networks are very different. At present, we don't manage these. But the broad vision is that potentially every customer meter point could actively interact with the network. That could mean going from a few hundred to low thousands of entities directly interacting with the transmission network operator, to tens of millions that somehow need to be coordinated. Unlike the transmission networks, local distribution networks will have to be run in a decentralised way. If you're going to have active interaction with essentially everyone, the situation is just too complex to think about bringing all the data together and controlling it centrally.

What’s the relationship between this kind of technical, localised transmission operation and market design? Does the way the market is set up influence how the networks physically work in practice?

That’s a good question. People from the mathematical sciences often approach problems by asking what the question is in principle, and then looking for a practical market or engineering scheme that can provide a solution. Sometimes, whether you have a market or a monopoly doesn’t make much difference to the control setup required. My colleagues and I don’t claim to be professional economists or market design experts. We tend to focus on technical, algorithmic aspects, whether that’s general approaches to running a system, which can be independent of a market, or the optimisation involved in running an auction and deciding which bids and offers are accepted. One point that certainly interests Lars and me is how the optimisation for clearing a locational energy market would be carried out. You need an appropriate optimisation problem, and one thing that often gets neglected until later is that you can have a brilliant market design, but to make it work in practice, you have to be able to crunch the numbers. When designing the market, it’s important to ensure you haven’t introduced features that make the computation unnecessarily difficult.

How did you, as a mathematician, get interested in energy and decision-making under uncertainty as your applied field?

Well, one of the reasons I got involved was simply that I was looking for a job. Back in 2006, it was also clear that research and innovation in energy systems were going to be areas with opportunities. I'd previously been in theoretical physics, where you don't have quite the same external drive for innovation coming from society. I think there were particular opportunities for someone like me because, historically, there had been less contact between the mathematical sciences and the energy network industry than there could have been. I also became interested in questions around reliability and situations where you don't have a lot of data. Partly because it seemed important, and partly because not many other people seemed to be focusing on it. Those questions around uncertainty management have probably been one of the threads running through the almost 20 years I've been in the sector now. Another consistent thread for some colleagues and me has been building a community of relevant mathematical science researchers and linking them to challenges coming up from industry – even when we’re not the ones developing the methods to solve them. More recently, this has expanded further: a few of us have started actively trying to bring together energy networks with mathematical science and climate communities. It's very rare to have a project of any nature where all three of these disciplines work in close collaboration. One feature we have here between the Schools of Mathematics and the School of Geosciences at Edinburgh is that you can have all three disciplines present with only two people in the room.

'We had the idea of creating this forum where the organisers don't particularly have skin in the game, so we can bring a range of people together to hash out the issues. I think the range of people from across the sector that we had at the First Scottish Forum on Future Electricity Markets shows that there was a demand for this.'

So with that in mind, how do you yourself hope to contribute to the UK’s evolving energy system?  

I'm really thinking about my role as being, by background, a sort of mathematical sciences person. I tend not to be naturally a “high policy” person. So I see myself being part of certain areas where I can work directly. I'm doing quite a number of projects at the moment around climate resilience. We're looking to do more work in managing systems with large volumes of energy storage. And then there's the ongoing work on the security of supply. The reform of electricity markets is probably a good example of where I'm not particularly a specialist, but where there are areas where I can contribute some wider thinking about cost-benefit analysis, bringing perspectives from different sectors. And also, where I can play that linking role and help bring together a range of specialists from the mathematical sciences who can contribute to solving some of the questions where we need answers quickly.

Is that kind of what you did with the First Scottish Forum on Future Electricity Markets?

Yeah, well, my understanding is that in this area of market design, we're not particularly specialist at Edinburgh. There are certain aspects where we do have the expertise. For instance, in the School of Maths, we have very strong capabilities in the relevant computation. But to some extent, we've turned that into an advantage: we had the idea of creating this forum where the organisers don't particularly have skin in the game, so we can bring a range of people together to hash out the issues. I think the range of people from across the sector that we had at the First Scottish Forum on Future Electricity Markets shows that there was a demand for this. We've been discussing what comes next, which is a little bit more complicated, because planning a second event will depend on the government's big decision. We are thinking about selecting several people who are able to discuss the consequences of that decision and the next steps, whatever the government decides.

This interview took place in May 2025. 

Jan Žižka