World Energy Outlook (WEO) is the annual flagship publication from IEA, regarded as the most comprehensive study of the energy system, and probably the most influential energy-publication there is. In the 2016 edition of WEO, which was presented in Oslo last week, the focus is on renewables. What is the outlook for renewables? What is the economics of renewables, and how do you integrate renewables? Particularly the economics and integration of variable renewable energy (VREs) like wind and solar.

In my masters thesis (in Norwegian) I looked at how variable renewable energy functions in wholesale markets and how the World Energy Model (WEM), used to make the analysis for WEO, models this. I specifically looked at photovoltaics (PV), so that will be the scope of this article, but some of the insights can also be applicable to other VREs like wind and/or other distributed energy production.

But first, let’s look at how the renewable projections in the New Policies Scenario differ from WEO 2015 to WEO 2016. In WEO 2015 solar PV starts with a total installed capacity of 127 GW, growing with an average of 34.4 GW/year (not accounting for retirements) until it reaches a total installed capacity of 1066 GW in 2040. In the 2016 edition, solar PV starts with a total installed capacity of 176 GW (2014), growing with an average of 47,3 GW/year (not accounting for retirements) until it reaches a total installed capacity of 1 405 GW in 2040. An increase of 339 GW or 32%.

If we look at historical capacity additions, the 2016 projection looks like a better projection than the one in the 2015 edition. We know that the annual capacity addition in 2013 was approximately 40 GWi, almost 50 GW in 2014ii and that the capacity additions in 2016 is forecasted to be somewhere around 70 GW, with a slowdown in 2017.

This is actually more in line with how IEA projects capacity additions in the 450 Scenario: In the WEO 2016 450 Scenario the PV market more than doubles. With an annual average growth of 74,3 GW/year (not accounting for retirements), gradually increasing until it surpasses 90 GW per year by 2030 and in 2040, reaches 110 GW per year. In 2040 the projected cumulative capacity is 2108 GWiii.

So what happened?

1. (Too high) Cost assumptions and learning that was left out

In my master thesis I found that the cost assessments used for WEO 2014 was higher than the actual prices in the following years as well as compared to other assessments at the time.

This leads to a development where solar PV is regarded as uncompetitive by the model and all capacity additions are policy driven. Figure one from IEA PVPSiv shows that this was not an entirely unreasonable assumption, as the majority of the solar PV market has been driven by subsidies. What these high cost assumptions did nevertheless, was that the added capacity in the model was lower than what we saw in reality, which in return lead to low learning and continuously high costs (despite using a high learning rate) and concluding in a conservative projection for capacity additions.

As there are no public cost assessments after WEO 2014 (more on this later), it is difficult to give a qualified comment on cost assumptions used in WEO 2015 and 2016. But it is reasonable to assume that the cost assessments have been updated according to the price decline the last years, and that this means that solar PV experiences larger learning effects due to both more capacity built per subsidy and also because solar PV becomes cost competitive in several markets (especially in the 450 scenario).

2. Issues with the World Energy Model and how it models renewables in the power market

Apart from high cost assumptions, the main critique in my master thesis was that the WEM used an approach with one merit-order and an annual aggregated load curve for each region. This means that the load profile even 25 years into the future is set by exogenous data based on the current demand profile. This is a general problem with techno-economic models. How do you describe change in cultural preferences? Is it probable that when the Indian population gets richer, they will adopt the same car-dependency of the Western middle class, or will urbanization and other cultural trends lead to radically different approaches for personal transport and ways of living?

This way of describing the power market, also ignores important characteristics for PV, characteristics that gives this technology market segments (both in time of day and in types of consumers) where it outcompetes other power producers. This again makes the model underestimate the effect increased PV capacity will have on the economics of baseload power plants.

The introduction of the WEM Hourly Model is a huge improvement to the WEM and WEO 2016, and there is a thorough discussion on integration and variability of VREs. In a world where the majority of capacity additions comes from renewables, understanding the inherent traits of different VREs and their effect on the power system is key. Thus the analysis done by the WEM hourly model is very interesting, but after reading the report and also the Model documentation it is still unclear how the results from the WEM hourly model analysis is used in the general WEM framework. For example:

• How do the results impact the capacity factor (CF) of new, and existing, power plants? In other words: Does the CF of new/old power plants (both VRE and dispatchable) change based on the results, and is this then taken into account when deciding new generation capacity/closure the following year?
• How do the results impact the relative financial attractiveness of VRE compared with dispatchable power plants?
• How do the results impact dispatch/ramping costs for baseload plants?

And does the WEM hourly model make it possible to look at the competitiveness of distributed generation (DG) in different sectors of the economy? Different sectors of the economy have different demand profiles and these may diverge from the aggregated demand profile of the region.

These are examples of where the WEM might have a “baseload”-bias. That is, an overvaluation of production from baseload power plants compared to production from VREs. That said, there are also model choices that gives a “VRE-bias”. A concrete example is that the hourly dispatch model does not represent the transmission and distribution system, and the costs this might add to VRE expansion.

3. The “Black box”-approach

For readers not so interested in energy models, IEAs “Black box”-approach should be the most problematic. Even though the WEO 2016 gets reviewed by almost two hundred reviewers from different organizations and experts, neither their comments or IEA’s assessment of these are public. The same goes for the models used and the data IEA relies on. Even today – the latest public available cost assessment is from the WEO 2014, meaning that the data probably are from 2012 or earlier. In short, the general public must take the results and underlying data and analyses from IEA at face value.

This is in contrast to other large contributors to the energy debate, like the US EIA and IPCC. EIA publishes all data and a thorough model documentation for their NEMS-model on their web page, and IPCC had an open review process. This resulted in 140 000 comments, all of whom was assessed and commented by specialists.

Conclusion

First of all, if we look at the historical capacity additions, it is the 450 scenario that best fits the historical trends. How large will the slowdown be in 2017 and what progress will we see in 2018? Is it realistic that the capacity additions will fall to a 2014 level and stay there (with a slight increase due to replacements of retired capacity)?

It looks like IEA have revised their cost assumptions, and that we start to see PV competing on a commercial basis as we see in the market today, especially in the 450 scenario.

The introduction of the WEM Hourly Model is a huge improvement to the WEM and WEO 2016, as is the thorough discussion on integration and variability of VREs for stakeholders. Next year, the focus should be on how large amounts of VRE affects the economics of existing and new baseload power plants. This would be an important tool for policy makers and stakeholders.

More problematic is the “Black box”-approach. A more open review-process and publication of data, model documentation and an overview of the research used, would both give IEA and the World Energy Outlook more credibility and be a huge addition to the scientific community. Thus IEA should be inspired by the openness of EIA and IPCC.

Literature