Thread by @JesseJenkins

Strongly disagree with this statement that energy models have *declining* policy usefulness in context of rapid technological change. In fact, when used appropriately, energy models should have *increased* relevance in this context (just not the way @EIAgov uses NEMS) 1/

https://twitter.com/MichaelWWara/status/1089201502952902656

The EIA uses their models to create "projections" which are often misinterpreted as "predictions." They are efforts to say what the future "will be" (prediction) or "is likely to be" (projection). 2/

Using a model for these purposes depends on your ability to very accurately characterize the underlying techno-economic dynamics AND the probabilities of changes in key parameters. When it comes to our rapidly changing energy sector, both of these break down. 3/

The underlying techno-economic factors change rapidly as eg the electricity sector includes large shares of variable renewables, which breaks the electricity module of NEMS making its projections next to useless for electricity sector... 4/

In addition, the probabilities of changes in key parameters such as natural gas or oil prices, solar costs, etc are extremely difficult to predict. In fact, given our poor collective track record predicting these trends, one might consider them *epistemically unknowable.* 5/

In other words, we face "deep uncertainty" with regards to technical change in the energy sectors over any time frame longer than say 5 years. See @RobertLempert et al here google.com/url?sa=t&sourc… 6/

So what good are models then if they cant predict or project the future? Doesnt that make them useless for informing policy or markets? ... 7/

...No! But you have to use models in a fundamentally different way: as exploratory tools to map the uncertainty we face and identify robust strategies. We can't run experiments with energy systems, so models are the only way we can gain insight on possible futures. 8/

In the face of rapid and uncertain technological changes, we need models even more as decision support tools. However, those models must be updated to capture key changes in techno-economic forces in each sector. 9/

For ex., electricity models need to capture key engineering & economic constraints & tradeoffs relevant to futures with high shares of variable renewables, distributed generation, tougher climate policies, energy storage, electrification of new demands, demand flexibility etc 10/

In other words, we first need to *change the models* and improve decision support models for a changing economic and technical landscape. See eg bit.ly/GenXModel and bit.ly/JenkinsDissert… 11/

Second, we need to *change how we USE the models.* In this context, models aren't for predicting what *will be,* they are for exploring what *might be.* We face deep parametric uncertainty that can only be explored with models used to perform extensive sensitivity analysis. 12/

Models can help us understand 1) Transition points between techno-economic regimes (eg how cheap does tech a need to be to see widescale adoption and transform rest of energy portfolio in response?). 13/

2) "Stable regions" within which parameter changes have little to no effect on outcomes.

3) Key "signposts" or parameter values (cost levels for fuels, techs, etc) that indicate major changes are coming and can be used to trigger policy decisions

/14

4) Robust actions or strategies that are valuable across a wide range of possible future outcomes.

5) Hedging strategies, or actions that may not pay off but are important to be reduce costs IF a particular set of outcomes materializes. 15/

6) Option creation strategies, or actions that have high value if they make future conditions with favorable outcomes become more likely.

7) Build intuition about role of various technology options, and how they might compete with substitute for, or complement one another

16/

8) Understand the set of cost and performance characteristics for emerging technologies that would make them significant contributors to our energy mix under a wide range of futures; useful for designing technology policy and RD&D objectives, etc. 17/

All of these insights can ONLY be gained from exploratory modeling of our deeply uncertain future with models updated for a changing techno-economic landscape. 18/

EIA's Annual Energy Outlook isn't used in an exploratory manner and their NEMS model isn't sufficient for modeling our changing landscape, particularly in the electricity sector. 19/

Yet models remain extremely relevant for generating policy insight. To do so (as always), we need the right kinds of models to be used in the right kind of way. /End