Integrating Distributed Energy Resources into Wholesale Markets and Grid Operations
Distributed Energy Resources (DER) encompasses distributed generation including combined heat and power, wind and photovoltaic systems, demand response, energy storage, vehicle-to-grid systems and microgrid. DER will have impacts on system operations and energy markets, which will require grid and energy market operations to embrace it. But what will be DER’s impacts, and how can we model DER economics and adoption from the user point of view?
There are two aspects to integrating distributed energy resources (DER)— what is known collectively as combined heat and power, wind and photovoltaic systems, demand response, energy storage, vehicle-to-grid systems and microgrids —with wholesale markets and grid operations. One is the impact of DER on the grid’s physical stability, the other the effects of price responsive DER on wholesale market behavior. These are two very different issues, but both have implications for wholesale operations visibility and retail-level interactions.
Almost all DER is connected to the grid via power electronics, specifically by inverters. Such resources in effect decouple the grid’s physical dynamics from the dynamics of the DER technology. That is, inverter-based resources do not generally exhibit frequency response or response to the rate of change of frequency; in other words, they do not have inertial or governor response.
This does not seem alarming if DER is thought of only as a type of load. Resistive load also does not have inertial or governor response, and loads based on power electronics such as variable speed drives also do not exhibit inertial or governor response. But if we think of DER as substituting for conventional generation, then the question arises as to whether the system still has sufficient inertial and governor response. There may also be locational issues (as opposed to interconnection or control area issues) to consider.
Operators of isolated island grids have long worried about these questions, as it is well known that wind generation and distributed PV can pose risks to maintaining sufficient primary frequency response. But large control area operators faced with high renewables and DER penetration are starting to look at this as well. Studies assuming the 15-20 percent penetration typical of renewable portfolio standards indicate there will be no problem in many cases. But at higher penetration levels concerns build about whether NERC performance criteria can be met or whether maximum frequency deviations on large unit outages can be managed.
One tool being increasingly employed for examining system dynamic performance over the time scales of interest for this problem is DNV GL’s Kermit, which simulates grid, generation and DER dynamics on a scale of sub-seconds to hours. Some existing studies focus on the theoretical inertia on line from conventional generators as a function of projected unit commitment factoring in renewable production. Other studies look at the aggregate primary response available from conventional units on line based on standard droop speed settings for unit governors.
This latter evaluation works today when there are many conventional units on line and spinning reserve is provided from a handful of units. But in a future scenario where at a given moment renewable production may displace 50 percent or more of the conventional generation in terms of on line capacity, there may be less than apparent total headroom available from conventional generation for primary or governor response and for automated generation control or secondary response.
Put differently, the primary governor response may eat into the capacity that was thought available for spinning reserves.
If it can be shown that under some scenarios there is insufficient inertial or governor response available, then market operators have to face increased dispatch costs so as to keep conventional units on line for this purpose. And what if conventional plant operators plan or threaten to retire plants due to inadequate revenues? Then market solutions have to be developed for the provision of inertial and governor response as ancillary services products rather than as conventional unit interconnection standards. Conceivably, those involved in capacity markets or capacity planning would have to consider these factors as well.
Once market product solutions to ensuring adequate inertial and governor response are on the table, then DER and renewables developers and technology communities will demand access to these markets and revenue streams. This leads to the use of synthetic inertia and synthetic governor response provided by inverter-based DER or wind farms. Depending upon the DER technology, this can be provided by, for instance, limiting the power delivered to the grid to a level below the instantaneous capability of the physical resource. For instance, a set of PV panels could be controlled to deliver less than full potential.
Wind farms without storage may have the ability to provide limited response by using the inverter to accelerate or decelerate the turbines for a brief period, but stress on the turbine blades is a concern.
Today NERC standards would not allow grid operators to make use of synthetic inertial or governor response. And indeed, except for a few special cases, the dynamic performance of synthetic and governor response is not well understood. But as DER penetration increases and the capacity to exploit these potential capabilities becomes more and more attractive, standardization will result.
At the other end of the dynamic spectrum, one of the major advantages claimed for future smart grid technologies is that end use loads will be able to respond to energy prices autonomously. What is needed here are regulatory and tariff structures allowing retail customers direct access to day-ahead, hour-ahead, or real-time energy prices from the wholesale market. Their development is not a straightforward as one might think.
Under the hood of the price-responsive load model is a small matter of price instability in what is called a sequential market in the context of economists’ cobweb theory. The energy market is sequential in this context in that the supply side (the market operator) clears the market using supply offers and estimated demand (load forecast or actual demand) and then publishes the price. Then the price responsive load reacts to that price. Ignoring the time dynamics of the relative speed of response of generators and end use load for the moment, all is well and prices will converge over time. But if the demand side is more elastic than the supply side, the cobweb expands over time and prices diverge.
This phenomenon is not imaginary. Consider a situation in which large industrial loads that have time flexibility in energy usage are subject to real time pricing. Assume that a generator outage causes the market operator to signal a combustion turbine to come on, and the real time price spikes as a result due to the instantaneous supply-demand imbalance. That price spike could cause the industrial load to interrupt its consumption—and the load could respond more quickly than the generator could come on line. So when the generator does come on, in say 10 minutes, the load has dropped and there is now an imbalance. This leads the operator to decrease the price and signal generation, prompting the load to switch back on.
When we consider the stability of the market process including the different time dynamics of generation and load, the answer is more complex than just the relative elasticities. The relative time delays matter as well. Without the math, suffice it say that if the load is faster than the generation, watch out.
What does this mean to the market operator? If the market operator manages to estimate the load elasticity with reasonable accuracy and clears the price anticipating the price response, then all is well. If load elasticity is ignored, then there can be problems depending upon a host of factors. So the market operator will need short-term and day-ahead load forecasting that incorporates the effects of load elasticity.
One path to being able to estimate elasticity is to gather data from all those price-responsive loads, a formidable task even on the optimistic assumption that all consumers know their price elasticities in advance. Another approach is for an aggregator to capture all this information and pass it on to the market operator. This, however, runs counter to the very idea of autonomous price responsive load and leaves the aggregator with the forecast risk.
Yet another solution is for the market operator to invest in developing big data analytics and data bases to forecast demand elasticity. That is a topic for another issue of the newsletter.