Gökşin Kavlak, postdoctoral associate, MIT Institute for Data, Systems, and Society
The research often involves understanding, what are the fundamental drivers of technology evolution? Why do technologies improve over time?
Jessika Trancik: My name is Jessika Trancik. I’m an associate professor at the Institute for Data, Systems, and Society.
Gökşin Kavlak: My name is Gökşin Kavlak. I’m a postdoc at the MIT Institute for Data, Systems, and Society.
JT: We both work on energy systems and on understanding the drivers of technology evolution over time, which is important because one of the challenges with addressing climate change is to drive the evolution of energy technologies toward affordable, low-carbon emission energy infrastructure. Two of the primary questions that we work on are the following. One is to ask, how can we combine and optimize different technologies in order to rapidly transition away from carbon emitting energy infrastructure today? There we look at how would you optimally combine energy storage technologies with combinations of solar and wind and other low-carbon, electricity-producing options? And what kinds of energy storage technologies do you need? How do you integrate transportation energy demand, for example from electric vehicles, into the electric power grid, and in that way decarbonize transportation energy services along with electricity. That’s one set of questions that we work on.
Another set of questions that we work on has to do with understanding how technologies change over time. In this first area, answering these first questions, we might come up with targets for energy storage technologies or for the cost of solar energy. In the second area we ask, how can you accelerate the cost decline in solar energy, for example, or in battery technologies? There the research often involves understanding: What are the fundamental drivers of technology evolution? Why do technologies improve over time? This is something we’ve worked on. Recently, Gökşin, we published a paper on photovoltaic modules and why they fell in costs, why the cost dropped by 99% over the last 40 years. What do you think were the most interesting conclusions to come from that paper?
GK: PV—photovoltaics—PV module costs came down dramatically over the last four decades, as you mentioned. What we did in this paper was to develop a method that brings a structure to answering this question about why this cost reduction happened in PV. We answer the question by identifying mainly two levels of mechanisms of cost reduction, low-level mechanisms and high-level mechanisms of cost reduction.
Low-level mechanisms are mechanisms that are about the technology itself. I can give some examples, such as module efficiency for PV modules, the amount of silicon, the amount of other materials used in a silicon module, yield, manufacturing efficiency, material utilization, and so on. Changes in these types of variables are what we call low-level mechanisms. For PV modules, we found that improvements in module efficiency was the main low-level mechanism and contributed to about 25% of the cost reduction since the 1980s. Fifty percent of the cost reduction since the 1980s came from improvements in materials-related low-level mechanisms. These are silicon price, reduction in other material usage, reduction in material prices, and so on. These are low-level mechanisms.
The second level is high-level mechanisms. These are the drivers behind the low-level mechanisms. These are drivers such as R&D—research and development, learning by doing, economies of scale—that are more closely aligned with the policies that are used to stimulate improvement in technologies.
Most of the cost reduction came from R&D-related improvements. For example, the low-level mechanisms that I mentioned, changes in efficiency, changes in material use, and manufacturing efficiencies. These were mainly driven by R&D, especially in the beginning of the technology starting from the 1980s and so on. Later on, after 2000 or so, economies of scale increased in importance with increasing manufacturing plant sizes and economies of scale became a high-level mechanism of equal importance as R&D.
After identifying—quantifying—the effects of low-level and high-level mechanisms, we also looked at the question about policies. Which policies drove these mechanisms? The main question we actually addressed in this paper was the question about whether it was public R&D or market expansion policies—such as feed-in-tariffs, renewable portfolio standards, and so on—was the main driver of cost reduction. In this paper, we found that both public R&D and market expansion policies were important but market expansion policies were slightly more important than public R&D and accounted for about 60% of the cost reduction in PV modules since 1980.
JT: It was interesting because one of the big questions that’s been out there is: Should and can government policy be effective in driving a technology’s improvement, first of all? We see that, in this case, government policies really were critical and drove this 99% cost decline in this low-carbon technology so that the cost today is 1% of what it was 40 years ago. That’s a major change and improvement in this technology.
Another question that people ask is about whether government should focus on funding research and development or whether they should also be involved in incentivizing the market growth of technology like solar energy, which is intended to address climate change. These climate impacts do not appear in the price of electricity today so it requires other kinds of policies to bring that into the price. People ask, should government be interfering, in a sense, in markets by bringing the climate impacts into the price or through policies specifically targeting the market growth of solar energy? Or should they focus just on public R&D, on funding research and development?
As you mentioned, Gökşin, in our work, we found that, in this case, in the case of solar energy, these market expansion policies were really important in bringing about this cost decline. When we go deeper into the technology and look at what changed these materials-related costs that you mentioned—the conversion efficiency and so forth—we see that some of them happened as a result of government-funded R&D but a lot of them happened as a result of private sector R&D and economies of scale and learning by doing. These things likely would not have happened without these market expansion policies.
A couple of takeaways—and, Gökşin, it will be great to hear your thoughts on these as well—but one is that government policy can drive a rapid change in improvement in a technology that is an important part of the solution to climate change. We need more of these kinds of improvements in other low-carbon technologies. I think we can replicate this example that we see from solar and other technologies. We also learned that market expansion policies were really important and should be pursued alongside government funding, I think, for R&D. As researchers, it’s not so much our role to say what government should do or shouldn’t, but looking at these results, it’s pretty clear that these market expansion policies were important. Those included things like feed-in tariffs in Germany and early subsidies in Japan and renewable portfolio standards in the U.S., really a mix of different policies that were important. I don’t think we would have seen what we saw in solar energy if we had just focused on government funding for research and development. I think that’s an important takeaway.
Finally, one of the features of this technology that is probably also very important is that we saw a number of these low-level mechanisms. We had on the order of five or more low-level mechanisms that contributed 10% of the cost decline or more over that period. What I take away from that is that there were a number of different knobs to turn in improving this technology. That may be why we saw steady improvement over 40 years in these solar modules. That’s again something we could look for in other technologies in creating and developing technologies that have a number of different knobs to turn.
One of the things we’ve been looking at now is actually going even further inside the black box of the technology to look at a lower level, even lower than these low-level mechanisms that you talked about, Gökşin, like efficiency improvements and yield improvements and so forth. What happened? What were the specific innovations that mattered in this case? I’d be curious to hear from you, in addition to these conclusions, maybe what surprised you the most from these results? Then maybe looking ahead to some of this new work that we haven’t yet published, but that we’re writing up, what are some of the insights from that? I think there are a number of fascinating insights. What are some that you think are most interesting to you or most surprising?
GK: As you mentioned, Jessika, there were actually multiple knobs that were effective in reducing the costs of PV modules. I think what was expected, to me, were improvements in efficiency and how it affected the costs. That was somewhat expected. What was surprising to me was this multiplicity of knobs and the fact that multiple low-level mechanisms were affecting costs. As you mentioned, now we are looking into these low-level mechanisms. We are asking, what happened and what changed? What made these low-level mechanisms effective? What were the innovations behind each of these low-level mechanisms?
Basically, we start with our cost model of a PV system. This means that we’re now including the balance of systems to our equation, to our system boundary. We previously looked at modules. But now we’re also thinking about the processes, about installing the module, getting permits, and interconnections, and so on. That’s called balance of systems. Innovations in PV modules have been studied extensively over the last decades in different fields. Right now, we’re also looking at innovations that reduce the cost of balance of systems, because balance of system costs are much higher and they declined less rapidly than module costs. That’s one of the reasons that now we’re including balance of systems in the study. Basically, we are looking at each low-level mechanism, such as module efficiency, and asking, what were the innovations behind this low-level mechanism? After a comprehensive literature review and expert elicitation, we identified a number of innovations for each low-level mechanism. We identified innovations at the cell manufacturing level, module level, and so on.
What we are doing in this paper right now is we are developing a typology for innovations. Basically, we want to understand what types of innovations affected module costs and balance of system costs. After identifying a large number of innovations, we can now see patterns of innovation types. We can basically see differences between these two types of technologies, so to speak, with this new research.
This is at the low-level. We are also looking at the high-level. We are actually also looking into the policies in our current work. Our goal is to build a complete picture of technological improvements starting with policies and driving high-level mechanisms, low-level mechanisms, and identifying innovations in between. Maybe would you like to talk about this complete idea of this project and how we are going to integrate all of these?
JT: Often times, when you see this debate in the public discussion around climate change and people asking, do we have all the technology we need to address climate change? Some would claim we have the technology we need, what we need is policy. Or others saying, let’s focus on a few key technologies and not wait for the policy. Or we don’t need policy, the technology should be able to develop on its own. You see all these different perspectives out there.
I think one of the key insights from this work is that it’s not one or the other. At all of these levels, there’s this kind of layered approach. The way to explain why a technology improves, and also the way to improve it in the future, is to look at efforts at all of these levels. If you have more policy, you bring about certain human activities in companies, as we saw in the case of solar energy companies around the world and people within those companies working hard. They wouldn’t have done that work had we not had government policy incentives that brought some of the price of carbon emissions into the equation by stimulating markets for this low-carbon technology. We wouldn’t have seen those efforts without that government policy. We wouldn’t have seen people working at universities around the world or government labs on these technologies without government funding for their research, because there just wasn’t the incentive for it in the market.
At all these levels, if you have policy, it brings about a lot of efforts that bring about these high-level mechanisms that you mentioned and that we described. Then those human efforts bring about changes to the technology itself and these changes that we observe in our cost equation that are the low-level mechanisms that we study.
We can think about other technologies in a similar way. When we’re looking forward and trying to address climate change, we can think about addressing the problem at all of these levels. I think that this layered, conceptual and theoretical framework for explaining technology evolution over time is quite useful. I’d be curious to hear your thoughts on this, Gökşin, and also to hear what technologies you’re most interested in maybe studying in the future using this approach.
GK: About your first question, yes, definitely, I agree with you that if we ask the question, why did solar costs come down? There are multiple answers, there are people, there are institutions, policies. We can say it’s China or Germany, we can identify certain countries. There are different ways to answer this question and there are different mechanisms at all these different levels.
What we did in developing this method was we basically introduced a structured way to think about this question and propose the hierarchical way of answering it, low-level and high-level. Some of these are overlapping, but by separating these levels from each other—low-level, high-level, and policies—we can think about them systematically.
This approach can be definitely applied to other technologies. In our lab, we are already applying this method to other technologies: batteries, nuclear power plants, and so on. Personally, as a next step, I would like to apply it to wind, first of all, because I feel like we’ve, and I personally, have spent a lot of time looking at solar. Wind energy is definitely another important renewable energy technology. I definitely would like to apply this method to wind, and I’m actually really curious about the energy storage application.
JT: I guess what I would hope people might come away with from this research—and some of the conclusions that I think are exciting and also actionable—is that, first of all, we’ve seen that low-carbon energy technologies—and solar energy provides a really important example of this—that these low-carbon energy technologies can improve rapidly. Cost can come way down. The solar module costs fell by about 99% over the last 40 decades. We see that this happened because of a number of different efforts. The incentives were there because of government policies. A number of people around the world in the private sector and public sector did important work on this technology to bring costs down and that brought about changes at the level of the device. This kind of model could be applied to other technologies. We can improve other low-carbon technologies and accelerate that improvement. We can use these kinds of models to forecast improvement, but also to inform strategies for how to use what are always limited resources—limited financial resources, limited time to address climate change—to drive down the cost of low carbon technologies.
Of course, we still need curiosity-driven research, but I think combined with that curiosity-driven research can be an understanding of the fundamentals of why technologies improve and why some technologies improve faster than others, how we can design technologies for rapid improvement. Also, the analytics and the forecasting tools and the models we’ve developed can be applied to many new technologies to assess their potential and to inform policy. Is there anything that you’d like to add to that?
GK: I’d like to add that our paper on PV modules came out in December 2018 in the journal Energy Policy. If anyone is interested in learning more about the method and our results, please check out the paper. We have also several other papers in the pipeline following up on this, so please follow us at our website, which is trancik.mit.edu. Thanks so much, Jessika, it was really fun to talk to you about this.
JT: Thanks for talking about this research, I enjoyed it.
The research discussed in this episode was supported by the U.S. Department of Energy Solar Energy Technologies Office under award numbers DE-EE0006131 and DE-EE0007662.