Technologies like solar photovoltaic systems consist of more than hardware — however, system deployment processes have lagged behind physical equipment in their rate of improvement, a new study finds.
The decline in the price of solar photovoltaic (PV) systems over the past decades is often considered a success story for clean energy technology. But what drove this trend, and how much did improvements in hardware contribute compared to changes in installation processes?
Researchers at HKUST, MIT, and Harvard have identified the most influential sources of cost change and developed a framework that provides a glimpse into future cost drivers. “Improvements in ‘hardware features’ like material usage or device efficiencies not only caused hardware cost declines, but contributed nearly 80% of the total decline in deployment costs,” says Magdalena Klemun, the study’s lead author, Assistant Professor in the Division of Public Policy, and a faculty affiliate at the Energy Institute. “That is a lot, and it is counterintuitive, as we typically associate reductions in deployment or ‘soft’ costs with efficiency gains in processes, not with better hardware, at least not primarily.”
Hardware features not only caused the majority of past cost declines, Klemun adds. “Based on a new metric we developed, hardware features continue to influence a larger share of costs than soft features, although soft costs now exceed hardware costs in many PV markets.” This is another counterintuitive result generated using the new model, which allows separating the contributions of hardware and non-hardware improvements to cost change in technologies. The findings are reported in the journal Nature Energy in a paper by Klemun and colleagues at the Massachusetts Institute of Technology (MIT) and Harvard University. Co-authors include Goksin Kavlak, an associate at the Brattle Group and former MIT post-doc; James McNerney, a senior research fellow at the Harvard Kennedy School; and Jessika Trancik, a professor in MIT’s Institute for Data, Systems, and Society (IDSS), and the study’s senior author.
Overall, while solar PV systems now cost just 1% of what they cost in 1980, only 10 to 15 percent of this dramatic cost drop can be attributed to “soft technology” features. These features include durations of various tasks in system design, installation, and permitting, as well as wages—essentially any price-relevant feature of the services and processes needed to deploy a photovoltaic system. While these features have improved, for instance, through replacing manual design drawings with software, they have done so more slowly than hardware features, and the changes were much less influential for cost change.
These findings directly relate to current challenges in the transition to low- and zero-carbon energy technologies, as the upfront investment costs of many clean energy technologies are now dominated by ‘soft costs’. Extrapolating the study’s findings, these costs may need to be reduced through continued hardware innovations, leveraging the same channels that have been effective in the past, where hardware innovations drove soft cost declines without much contribution from soft improvements. Or, inefficient deployment processes and the associated ‘soft features’ could be tackled directly without changing hardware, trying to increase historically slow improvement rates.
“Being deliberate about soft technology is essential for making it more efficient and effective — to drive down costs, support a high-quality customer experience, and create jobs, among other objectives’’ says Trancik, “Soft technology will be instrumental for supporting a successful clean energy transition.‘’ “However,” Klemun adds, “the kind of systematic thinking typically applied to hardware design doesn’t exist yet for soft technology. So there is a lot of work to do.”
Establishing a science of soft costs
Part of the motivation for the study was to improve traditional approaches to technology cost modeling. In the past, costs have often been modeled as sums of hardware and non-hardware cost components, and changes in costs have been associated with changes in hardware or non-hardware technology inputs. However, additive cost components are just the first layer in the new framework developed by the HKUST-MIT research team. ‘’To really understand why rates of change in technology costs are rapid or slow, we need to go deeper than simply adding up the costs of inputs,’’ Trancik says. ‘’We need to consider the features of technology and how those features are changing and contributing to costs.’’.
In the model underlying the new study, cost components are represented as products of functions of several cost variables, which capture individual hardware and soft technology features—a technology’s representative state at any given point in time and space. The model then splits out the contributions of changes in individual features to changes in cost components, given data on how these features changed over time in a given location.
Using this approach, the researchers were able to estimate how influential better PV hardware (e.g., lighter modules with higher rated capacity) was for cheaper installation. For example, they computed how much the increase in module and inverter efficiency — both changes in hardware features, and the acceleration of mechanical installation tasks with time — a change in a soft feature — contributed to the total decline in installation costs. “Separating those contributions is important because hardware and soft features are encoded in technologies in different ways, which affects how innovations are typically shared across locations and influence costs”, Klemun says. Hardware features are embodied in the design of physical components, which can be mass-produced and shipped around the world, bringing much of the associated information with them. Soft features are encoded in people and institutions, which are typically less mobile. “Both feature types ultimately need to improve to optimize technology performance, but the underlying strategies may differ. That’s why separating the two is a good starting point to consider this difference carefully in engineering design, manufacturing, and policy”.
Interestingly, the researchers showed that the influence of some hardware features on costs was realized almost to the same degree through soft cost reductions as it was through hardware cost reductions. For example, photovoltaic modules were, on average, twice as efficient in 2017 compared to 1980, and that improvement reduced overall system costs by 17 percent. Yet 40 percent of that overall reduction, almost half, could be linked to soft cost reductions driven by higher module efficiency.
One of the reasons for this result is intuitively simple but challenging to formalize in a cost change model: When a technology’s hardware changes, e.g., when components become lighter or change shape due to design and manufacturing innovations, these changes also affect deployment processes and, thus, costs. However, when a process is altered at the permitting office or installation site, the hardware components stay the same, as their features have been “fixed” at the factory gate. Therefore, while most cost components are functions of hardware variables, only a few soft cost components show dependencies on soft variables or “features.” “You can spot this structural difference even before collecting data on how the technologies have evolved. That’s why first understanding and then visualizing a technology’s network of cost dependencies is a promising way to get at potential drivers of change, not just for solar PV but also for other technologies,” Klemun notes.
Looking across countries and to the future
The study also explored the drivers of cost differences across countries, which remain large. The paper shows that in no major PV market covered in their dataset, soft costs improved from a comparatively high level in the past (1980) to a comparatively low level in the present (2017). In other words, countries with robust ‘soft technology’ did not necessarily reach that performance level with time but already had lower soft costs to begin with. Countries with higher soft costs have tended to improve at a similar rate globally, driven by hardware innovations shared through integrated global supply chains. Soft technology innovations weren’t shared across borders to the same degree, or when they were, their influence on costs was smaller.
Going forward, Klemun is interested in exploring whether what they observed for PV — that improvements in soft features haven't been that influential for soft and overall cost reductions — also holds for other technologies. “Maybe there is a silver bullet for soft cost reductions that can be applied to PV. Or maybe not, in which case we would learn something about the importance of building the potential for hardware-driven soft cost reductions into technologies at the design stage.” One of Klemun’s current projects examines the cost evolution of advanced metering infrastructure. “Smart metering systems are similar to solar PV in that some components are standardized, easy to ship, and mass-manufactured, like the meter itself. But then integrating communication and data management systems can lead to high soft cost shares across sites, countries, and years.” These costs are often covered by the public sector through mandated smart meter roll-outs, she says, but the return in terms of cost improvement hasn’t been studied.
Another interesting topic, she says, is to critically examine the desirability of reducing different types of soft costs. “Not all soft costs represent inefficiencies; longer processes can make technologies more attractive to the consumer (due to customization or by enabling participatory processes) or safer. So there are trade-offs between deploying technologies very efficiently to speed up the clean energy transition and leaving enough room for temporary inefficiencies and things like creativity or building clean energy communities.”
This research is funded by the U.S. Department of Energy Solar Energy Technologies Office.
About the Author
Prof. Magdalena Klemun
Prof. Klemun is Assistant Professor at the Division of Public Policy. Prior to joining HKUST, Magdalena was a postdoctoral associate at the Institute for Data, Systems, and Society (IDSS) at the Massachusetts Institute of Technology (MIT). Her research interests are in understanding how the economic and environmental performance of technologies evolves as a function of policy and engineering design choices, with a particular interest in the role of hardware vs. non-hardware ('soft') innovations. Magdalena received her Ph.D. from IDSS at MIT, M.S. in Earth Resources Engineering from Columbia University, where she studied as a Fulbright Scholar, and her B.S. in Electrical Engineering and Information Technology from Vienna University of Technology.
