Why wind and solar power are such a challenge for energy grids

Last week, I described a modeling study showing that it is possible to run the entire US economy on renewable energy: wind, water, and solar power. Technologically, the tools are available. Economically, the total system costs would be lower than a business-as-usual scenario. But politically, the plan is wildly ambitious, to the point of fantasy.

Among other things, it would require that policy and investment decisions be approached holistically, coordinated across multiple sectors, and made on the basis of multidecadal cost-benefit horizons, with enormous upfront investments paying off in health and climate benefits that unfold over decades.

That is not the way humans typically approach big challenges. Engineers aren’t granted the power to redesign large systems from scratch. Energy is not just a physical system, it’s a social and political system too, and social and political change is unpredictable and messy. It lurches and stalls. Progress, when it comes, is often kludged, backward-looking, and incumbent-protecting. It’s a fallen world we live in, but we muddle through.

Barring an unexpected sociopolitical mobilization on the scale of World War II, it’s likely we’re going to muddle through the transition to clean energy. That puts the challenge in a different light. It forces us to grapple with the system as it currently exists, with all its entrenched incumbents, sunk costs, and behavioral inertia. And it forces us to confront trade-offs between our ideals and political realities, trade-offs that cannot be waved away with ritual invocations of “political will.” Thinking this way is more vexed and less fun than blue-sky modeling, but it’s a necessary complement.

So here’s our question: given current energy infrastructure and institutions, how much variable renewable energy (VRE) — wind and solar — can we integrate into the energy system?

This is a complex and contentious topic, in part because energy systems, and our understanding of them, are rapidly evolving. This post will offer a broad overview the kinds of challenges wind and solar pose to grids, and the kinds of solutions grid managers use to address those challenges. In my next post, we’ll look more closely at how far those solutions can get us.

There’s been a lot of study on this recently, rounded up in this paper from the 21st Century Power Partnership, a coalition of research institutions including the National Renewable Energy Laboratory (NREL). See also this similar NREL paper, which draws on this massive 2012 study and others.

Here’s NREL’s list of the five features of VRE that pose challenges to grid managers:

1) Variability: This is the biggest and most vexing.

Power plants that run on fuel (along with some hydro and geothermal plants) can be ramped up and down on command. They are, in the jargon, “dispatchable.” But VRE plants produce power only when the wind is blowing or the sun is shining. Grid operators don’t control VRE, they accommodate it, which requires some agility.

This figure shows why, using wind power as an illustration:

This shows one week of electricity supply and demand (details and location not particularly important). The green at the bottom is power coming in from wind. The yellow at the top is total demand. The orange in the middle is the gap between the two, the amount that has to be supplied by conventional power plants.

Another way of looking at it: from the perspective of the grid operator, who has control over a set amount of dispatchable power, VRE energy supply is functionally equivalent to reduction in demand — large, rapidly rising and falling fluctuations in demand for dispatchable power.

On the chart above, “shorter peaks” refers to times when conventional plants are supplying the day’s “peak load,” which is when power is most valuable. VRE reduces or “shaves” the peak, thus screwing with the economics of conventional plants. “Steeper ramps” refers to times when conventional plants have to increase or decrease their output quickly in response to fluctuations in VRE — often more quickly than they are designed or regulated for. And “lower turn-down” means that in times of high VRE supply, conventional plants will have to run at the lowest output they are capable of, i.e., “minimum load.”

All these effects of variability pose challenges to the rules and economics that govern existing power infrastructure.

2) Uncertainty: The output of VRE plants cannot be predicted with perfect accuracy in day-ahead and day-of forecasts, so grid operators have to keep excess reserve running just in case.

3) Location-specificity: Sun and wind are stronger (and thus more economical) in some places than in others — and not always in places that have the necessary transmission infrastructure to get the power to where it’s needed.

4) Nonsynchronous generation: Conventional generators provide voltage support and frequency control to the grid. VRE generators can too, potentially, but it’s an additional capital investment.

5) Low capacity factor: VRE plants only run when sun or wind cooperates. According to the Energy Information Administration, in 2014 the average capacity factor — production relative to potential — for utility-scale solar PV was around 28 percent; for wind, 34 percent. (By way of comparison, the average capacity factor of US nuclear power was 92 percent; those plants are almost always producing power.) Because of the low capacity factor of VRE, conventional plants are needed to take up the slack, but because of the high output of VRE in peak hours, conventional plants sometimes don’t get to run as often as needed to recover costs.

So those are the challenges. Now let’s take a look what NREL calls “emerging best practices” in addressing them.

Improved planning and coordination: This is the first step, making sure that VRE is matched up with appropriately flexible dispatchable plants and transmission access so that energy can be shared more fluidly within and between grid regions.

Flexible rules and markets: Most grids are physically capable of more flexibility than they exhibit. Changes to the rules and markets that govern how plants are scheduled and dispatched, how reliability is assured, and how customers are billed, says NREL, “can allow access to significant existing flexibility, often at lower economic costs than options requiring new sources of physical flexibility.”

Flexible demand and storage: To some extent, demand can be managed like supply. “Demand response” programs aggregate customers willing to let their load be ramped up and down or shifted in time. The result is equivalent, from the grid operator’s perspective, to dispatchable supply. There’s a whole range of demand-management tools available and more coming online all the time.

Similarly, energy storage, by absorbing excess VRE at times when it’s cheap and sharing it when it’s more valuable, can help even out VRE’s variable supply. It can even make VRE dispatchable, within limits. (For example, some concentrated solar plants have molten-salt storage, which makes their power available 24 hours a day.)

Flexible conventional generation: Though older coal and nuclear plants are fairly inflexible, with extended shut-down, cool-off, and ramp-up times, lots of newer and retrofitted conventional plants are more nimble — and can be made more so by a combination of technology and improved practices. Grid planners can favor more flexible non-VRE options like natural gas and small-scale combined heat and power (CHP) plants.

Cycling conventional plants up and down more often does come with a cost, but the cost is typically smaller than the fuel savings from increased VRE.

Flexible VRE: New technology enables wind turbines to “provide the full spectrum of balancing services (synthetic inertial control, primary frequency control, and automatic generation control),” and both wind turbines and solar panels can now offer voltage control. (See this paper for more.)

Interconnected transmission networks: This one’s pretty simple. Wind and solar resources become less variable if aggregated across a broader region. The bigger the geographical area linked up by power lines, the more likely it is that the sun is shining or the wind is blowing somewhere within that area.

For example: this paper models a grid that connects 11 wind-power sites up and down the US East Coast and finds that “the output from the entire set of generators rarely reaches either low or full power, and power changes slowly.” Geography, in other words, reduces variability.

NREL helpfully arranges these solutions according to their cost (these are loose estimations, as actual costs will depend on the particular system in question and the current state of technology):

As you can see, changes in rules and markets are generally the cheapest way to go — and those solutions are available to every grid operator, today. They just require (d’oh!) political will.

Also notable: on average, energy storage is more expensive than other flexibility options, and batteries are the most expensive of all. This is what prompts many analysts to predict that storage will play a smaller role in the short term than conventional wisdom has it. All these other flexibility solutions will be slotted ahead of it, while it continues to get cheaper. (Then again, the surprisingly affordable Tesla home battery has other analysts even more excited about storage. In short, nobody knows anything.)

For a look into some case studies on how current grid operators are boosting flexibility, see this paper from the Brattle Group. There’s some really cool stuff in there, including new technologies that enable much faster and more accurate real-time forecasts for wind and solar energy.

Okay, so we have a set of challenges and a set of solutions. Where does that leave us?

NREL clarifies two key points. First, the challenges to integrating high levels of VRE into the grid are technically solvable:

[R]egional grid integration studies conducted to date have indicated that there is nearly always a technological fix that can be adopted at some cost (e.g., a change in operation or piece of hardware that can be added to the grid). So, simply deploying extremely large amounts of transmission and storage (or some other set of technologies), and modifying the RE generation to maintain system operational parameters could enable 100% penetration of wind and solar.

So the VRE carrying capacity of a grid is technically 100 percent, if cost is no issue. But cost tends to be an issue. So NREL posits a difference kind of carrying capacity:

[T]he limit to RE penetration is primarily economic, driven by factors that include transmission availability and operational flexibility, which is the ability of the power grid to balance supply and demand. This limit can be expressed as economic carrying capacity, or the level of variable RE generation at which that generation is no longer economically competitive or desirable to the system or society.

This notion of “economic carrying capacity” clarifies our original question. Technically speaking, we can integrate as much VRE as we want, as long as we’re willing to keep spending more money on grid-integration solutions. The question is, at what point is it cheaper, from a total cost-benefit perspective, to resort to low-carbon alternatives to VRE? And wherever that point is, will it still be there when we actually reach it?

We’ll tackle that issue in my next post.