Can we predict the future growth of renewables based on past trends?
By Avi Jakhmola
In its recently released 2021 World Energy Outlook, the International Energy Agency (IEA) projects the growth of wind and solar power up to 2030 in four different scenarios. Bloomberg New Energy Finance (BNEF) has also recently published its projections for solar power additions in 2021-2030. In this post, I compare these projections to estimates derived by extrapolating historical growth data into the future using the Gompertz growth model.
The Gompertz model of an asymmetric S-curve was used by our team in a recent publication to measure the growth of renewables in the 60 largest countries. We have also published code which allows fitting a Gompertz curve to any data series.
Here, I introduce a tweak and fit Gompertz data not to electricity generation, as was done in the article, but to the capacity of solar power normalised to the total electricity supply. Capacity data are usually more recent and fluctuate less than generation. Normalising to overall electricity supply takes into account the changing size of the global electricity system, which is important for a long-term analysis.
The figure below shows the historical growth of solar power (black circles for IRENA data and green squares for BNEF estimates, which are slightly different). The IRENA data are extrapolated by the Gompertz model (orange line) for 2021-2030. The four IEA scenarios – Stated Policies (STEPS), Announced Pledges (APS), Sustainable Development (SDS), and the Net Zero Emissions by 2050 (NZE)– are shown as dotted straight lines connecting the current value to the 2030 values (i.e. the projected capacity in 2030 divided by electricity supply in 2030 in the same scenario). The BNEF projections are normalised to electricity supply from the IEA scenarios for the high estimates (normalised to STEPS, dark green triangles) and low estimates (normalised to NZE, light green triangles).
Historical (2000-2020) and projected (2021-2030) global installed solar PV capacities normalised to the total global electricity supply.
It is easy to see that the Gompertz extrapolation of the empirical trends aligns very well with the BNEF projections as well as with two of the IEA scenarios – STEPS and APS (being a bit closer to the Announced Pledges). The SDS and NZE scenarios envision larger normalised capacities (1.17 and 1.49 times by 2030 respectively) than the Gompertz projection, BNEF and the two other IEA scenarios.
Another way to look at the same data is to measure annual additions of solar capacity (annual growth rates) as shown in the figure below. The growth of solar power is still accelerating (there are more additions every year) and the Gompertz model reflects this gradually rising speed, eventually stabilising at about 7 MW/TWh per year soon after 2030 (not shown on the figure). Once again, these growth rates extrapolated by the Gompertz model for 2020-2030 match the BNEF (6.0-6.9 MW/TWh) growth projections as well as the IEA STEPS and APS scenarios (4.9 and 6.2 MW/TWh respectively). The SDS scenario projects a slightly higher rate of 7.7 MW/TWh per year and, but the NZE projects a significantly (40-50%) faster growth of 10.6 MW/TWh.
Historical (2000-2020) and projected (2021-2030) global, normalised solar PV growth rates.
To summarise, my extrapolation of historical trends using an asymmetrical S-curve model is a good match for sophisticated global scenarios, especially if normalised to the size of the electricity system. For a deeper dive into our method for fitting the growth models and estimating G check out our publication and our interactive curve fit explorer.