Advocacy for distributed generation (DG) has a long history. Conservation activists preached the concept in the early ‘70s, and even before. But it never really caught on. Now, suddenly, it’s again become a hot topic. Grid-connected PV systems are blossoming on individual homes, and Capstone Microturbines is having some success in selling small systems for combined heat and power (CHP) at commercial and small industrial sites. A United States Combined Heat and Power Association is raising its voice in Washington, and the Department of Energy has major programs devoted to distributed energy resources (DER).
So what has happened, and what does it mean for electrical power systems in the future?
Not everyone would agree that anything of real significance actually has happened. Overall, the DG market remains quite small, accounting for something on the order of 1% of total US electricity supply. The bulk of that is from a comparatively small number of large industrial CHP applications, many of which have been in place for decades. DG skeptics don’t see much reason for the situation to change significantly.
Nevertheless, a number of things have changed to boost the potential for DG and DER.
As often seems to be the case when markets turn a corner, it’s not really one big thing that does the trick, but an accumulation of smaller things. Many individual developments have come together to tip the economic scales on DG from mostly unfavorable to increasingly favorable.
Perhaps the most significant development is simply the ongoing cost-performance improvement in power electronics and controls. That has made it feasible to develop smart inverters that are efficient, economical, and safe for connection to the power grid. The new IEEE 1547 “Standard for Interconnecting Distributed Resources with Electric Power Systems” has facilitated the process by giving equipment designers a single well-regarded standard to work to. Underwriter Laboratories standard UL 1741 references IEEE 1547 for its technical requirements. The UL standard is now the basis for certification of DG interconnection equipment. Although the utility industry is highly fragmented, UL certification carries weight. Together, IEEE 1547 and UL 1741 bring a degree of confidence and regularity for the DG equipment industry that was sorely lacking in the past.
Another change has been the shifting economics of various power resources. The price of gas and coal, the chief fuels for central power generation, have risen sharply in the last few years, while the economics of wind and solar power generation have been steadily improving. Also, the recent development of reliable, low cost micro-turbine systems has brought the efficiency of CHP systems within reach of smaller businesses for whom it wouldn’t previously have been a viable option. The turbines, which can run on very lean gas mixtures, have also made it feasible for cities to tap gas from landfills and sewage treatment plants as DG resources.
At least as significant as the changes that have already occurred are those that are expected in the near future. Chief among those is the economic feasibility of fuel cells for stationary power generation. These will be high temperature units that can potentially be fueled by hydrogen, internally reformed natural gas, or synthesis gas from coal or biomass. High temperature operation will enable them to be coupled with microturbines for very efficient power generation. Several competing technologies are already established well enough for test deployment. They are currently too expensive to threaten micro-turbines in CHP applications, but with the amount of development attention they are getting, that will almost certainly change. They promise to double the thermal efficiency of power generation compared to present micro-turbines, and rewrite the economic equations for CHP.
And then there’s wind power…
Wind power is perhaps the strongest driver for DG technologies. In good locations—and disregarding the cost of backup power and new transmission lines—wind turbines can now deliver energy at a lower cost per kWh than either gas or coal fired plants, and with zero CO2 emissions. Transmission issues are a big concern, as is intermittent supply, but wind power is nonetheless an increasingly attractive option on an economic basis alone. If CO2 emissions are ever taxed, its advantages will be impossible to ignore.
Wind power is an odd duck in the usual scheme of central vs. distributed power. Wind turbines are necessarily spread out geographically. Individually, they produce power as distributed resources, yet they don’t look much like any other DG resource. The output from dozens to hundreds of turbines is aggregated in wind farms whose net capacity may compare to that of central power plants. Wind farms tend to lack the signature DG advantage of low transmission costs by virtue of being physically close to where most of their output is consumed. On the contrary, their irregular output and distance from markets can be a major challenge for power transmission systems.
Ironically, the most cost-effective strategy for utilities to integrate high levels of wind power may be to take advantage of other distributed energy resources. However, that will need to include discretionary (dispatchable) loads as well as distributed generators. It will also require a more advanced standard for interconnecting these resources than what is provided by IEEE 1547. To see that, let’s first take a closer look at what’s involved in 1547.
The key requirement that shaped IEEE 1547 is reliable anti-islanding. “Islanding” refers to the creation of an “island” of energized lines drawing from a local generator running independently of the grid at large. It can be deliberate, as when an industrial facility with its own generating capacity isolates itself from a stressed grid, in order to avoid voltage sags and risk of blackout. But it can also occur unintentionally, if a DG resource fails to isolate itself when the grid to which it is connected goes down. An island of energized lines persists in the middle of a region of what are supposed to be dead lines. That can be a life-threatening hazard for line crews working in the area.
The approach specified in 1547 for anti-islanding involves a combination of voltage and frequency feedback. Basically, the inverter will only output power to the line if it detects AC voltage on the line indicating an active connection to a central generator. If the voltage goes either too high or too low, the inverter immediately disconnects.
To make sure that the voltage it is sensing is truly that from a central generator, and not the output of another DG resource or the echo of its own output from a reactive load, all 1547-compliant devices share a certain design feature. They “try” to output power at a frequency slightly higher than the standard line frequency. But they also use phase-locked loop methods to continuously sync their output with that of the power line. As a result, they never actually go out of phase with the central generator, as long as the connection holds. If the connection is lost, however, and a coincidental match between active DG resources and the set of loads prevents the voltage from immediately dropping out of range, the set of DG resources driving the line will find themselves “freewheeling” at their set frequency somewhat above line frequency. An internal clock quickly detects the resulting phase drift, and shuts down the output.
This approach has the advantage that it is relatively simple and depends only on the line connection. It does not require any side channel communication, and it can be implemented in a certifiable fail-safe manner. But it also has intrinsic limitations that will eventually necessitate something more advanced.
Limitations of 1547
The most obvious limitation of IEEE 1547 is that, by design, it requires a strong central generator to provide the “heartbeat” for the system. DG resources are assumed to be minor contributors to what remains predominantly a central power system. If the level of DG “penetration” gets too high, the phase drift features intended to make DG device operation fail-safe will overwhelm the central generator and cause the system to shut down.
For now, that limitation of 1547 is not a problem. There is no way that DG will provide a large enough fraction of our net electricity supply to manifest the problem within the next few years. However, the situation could start to change fairly quickly once high temperature fuel cells for stationary power become economically competitive.
Hybrid systems of high temperature fuel cells integrated with gas turbines can achieve very high thermal efficiencies. Moreover, they can do so in small modules of a megawatt or less. 53% efficiency was demonstrated in 2002 in a small prototype, and analysts working for DOE expect that 60% to 70% will be common in the near future. For advanced designs further down the line, their target is 75 – 80%. These targets far exceed the thermal efficiencies of today’s most advanced GTCC central power plants.
Manufacturing economies for hybrid fuel cell / gas turbine systems appear to favor mass production of relatively small modules. These would be in the range of hundreds of kW to a few MW. There are no obvious scale advantages from grouping large numbers of modules into a central plant. For greatest efficiency, the modules can and should be distributed among facilities that can make use of their waste heat. But robustly controlling a grid with the bulk of power coming from thousands of such distributed units is not something that IEEE 1547 was designed to handle.
A less obvious limitation of IEEE 1547 is that it addresses only positive DG resources—those capable of delivering power to the grid. To efficiently cope with the characteristics of wind and solar power resources, it will be essential to have a standard that covers negative DG resources as well. Negative DG resources are otherwise known as discretionary (or dispatchable) loads. Shutting down or throttling back a discretionary load is just as effective for meeting peak demand as turning on a peaking generator. Where it can be made available, it’s also likely to be a much cheaper and more efficient way to meet peak demand.
While there is provision for dispatchable loads in the national electricity market, it is applicable only for utility companies and industrial energy users large enough to play in the bulk energy trading market. For ordinary commercial and residential users, the only option at all similar has been dual metering. Dual metering allows customers to pay lower rates for power used during off-peak hours. Its main residential use has been for water heating in homes that are not supplied with gas. But it’s tied to a fixed daily schedule. It’s useless for integrating wind power, whose availability follows no set schedule.
Recently, control systems have been introduced in the New York region that allow big air conditioners in commercial buildings to be remotely “throttled back” by the utility when power demand is critical. In return for installing the controls, commercial customers are offered lower power rates. The real-time remote command capabilities of this system make it closer to what’s needed for the dynamic “use it when it’s there” strategy that wind and solar resources favor. However the system is quite limited in scope. It’s aimed at curtailing a relatively small number of heavy loads at times when supply is critically short. It isn’t aimed at shifting discretionary loads to times when excess supply happens to be available.
A discretionary load, by definition, has some latitude in when it operates. The nature of the load determines how much latitude. An example is the electrical hot water heating mentioned earlier. The total energy needed for water heating in one day is determined by usage, and is not discretionary at the level of the water heater. But because hot water is fairly easy to store, there is considerable latitude as to just when the energy is supplied.
Water pumping from wells to storage tanks is another example. In that case, the timing latitude is even larger than for water heating. While an insulated hot water tank can hold a day’s worth of hot water for a household, a cold water tank can easily hold many days or even weeks of supply. That flexibility makes water pumping probably the single most naturally suited application for wind and solar power.
Many electrical loads that are not currently discretionary could be made so, given incentive. For example, the microturbine systems that Capstone Turbine Corporation sells for small-scale CHP are normally operated to satisfy real-time heat demand, with electricity produced as a byproduct. However, they could easily deliver heat to tanks of eutectic salts, which would store it for later use. That would allow the turbines to operate mostly at peak demand times, when their power could command a premium.
Unfortunately, the fixed rate system that applies for most residential and commercial customers gives no incentive for such operation. Utilities mostly sell power at less than their own marginal purchase cost during peak hours, recouping by selling above the marginal purchase cost at other times. That simplifies metering and makes it easy for customers to understand their utility bills, but it insulates the system from an efficient source of matching supply and demand. That source will be increasingly important as the fraction of power from wind and solar resources increases.
Beyond IEEE 1547
What might a more advanced standard for interconnecting distributed resources look like? It clearly must provide for anti-islanding as robust and fail-safe as 1547, while enabling very high penetration by distributed resources. Ideally, it should make it possible to interconnect a set of distributed resources in a semi-autonomous grid with no central generator at all. The standard should also allow for real-time rate variability, in order to match available supply and demand.
What’s needed is functionally similar to what already exists for central power stations connected to one large regional grid. Power plants linked in a regional grid are, in their own way, “distributed resources”. But the control system for regional grids involves human operators in busy control rooms. Reports flow in, status displays are updated, decisions are made and disseminated to the affected parties. Clearly, that sort of operator control is totally infeasible for a system integrating tens of thousands of small power sources and discretionary loads. But is fully automated control possible, and can it match or exceed the safety and stability of the current system?
From a purely technical perspective, the most efficient system might resemble the following:
- Signal generators at local distribution stations inject a low-amplitude signal at a frequency of perhaps a few hundred kHz on the distribution lines. The injected signal is modulated to provide a precise time reference for any DG resources connected to the line.
- DG resources monitor the injected signal and use it to synchronize their output. Loss of this signal causes an immediate disconnect.
- The signal is also modulated at a low bit rate to convey a real-time demand level. A low demand level encourages discretionary loads to connect and discourages DG sources from connecting. A high demand level does the opposite. Metering rates might or might not be directly tied to the demand level.
- DG resources also monitor the current flowing through the distribution line at their point of connection. This serves two functions:
1. It allows them to sense the direction of power flow and to perform automatic power factor correction; and
2. It allows distributed generators to detect and respond instantly to line faults.
The most critical feature of this scheme is the real time demand level and how it is managed. If not done properly, it could result in disastrous instability. In response to a signal of low demand, large numbers of discretionary loads could simultaneously connect, momentarily overloading the system, only to all cut out a moment later in response to a signal of high demand. At least, that would be the natural fear of the system operator.
The stability issue is actually not hard to deal with. The solution involves hysteresis and random delays in the connect / disconnect control logic of individual resources. That causes them to respond in a statistically predictable manner that preserves system stability. Thus, in response to a lowering of the signaled demand level, a few discretionary loads with short delay times would connect, and/or a few discretionary generators would disconnect. A short time later, if the demand level remained low, more units would respond. After a few minutes at a given demand level, all the units that could be expected to respond to that level would have done so. If at that point there remained an excess of supply over demand, the central controller, as the system’s “auctioneer”, would need to further lower the signaled demand level, triggering more units to respond.
Distributed Resources and the “Smart Grid”
Fully automated and distributed control over connection of distributed resources as outlined above should be technically feasible. Whether it is institutionally feasible is another matter. It’s foreign to the way that utilities and system operators currently manage the power grid. It’s also foreign to the architecture for the “smart grid” of the future that is proposed by the Electricity Innovation Institute in its “IntelliGrid Architecture”. E2I, as the Electricity Innovation Institute is known, is affiliated with the Electric Power Research Institute, and with CEIDS—Consortium for Electric Infrastructure to support a Digital Society”. The IntelliGrid Architecture does not envision distributed control of individual energy resources in the manner suggested above. It punts the issue of control for DERs by proposing that “aggregators” manage them.
An aggregator presents much the same interface to the system operator as a power plant. The aggregator contracts to supply blocks of power or load in the same manner as a power plant or a large industrial power user. It’s then the aggregator’s responsibility to somehow manage the distributed resources under its purview so as to meet the contracted commitment. How it achieves that is not covered by any standard, so it’s up to individual aggregators and DER equipment manufacturers to work out a system design.
The aggregator model is one with which incumbent utilities and operators are comfortable; it does not disrupt their current methods of operation to any significant extent. Within limits, it is certainly workable, as shown by the existence of Celerity Energy. Celerity is the country’s largest, most successful energy aggregator, managing DE resources for the University of California system and for many other clients. It focuses mainly on enabling clients to get better value out of their under-utilized backup generators. It installs internet-connected controllers on the DE equipment that allow it to be remotely started, stopped, and monitored. The company maintains a database of equipment under its management. Custom software uses the database to negotiate contracts with the grid operator and devise dispatch schedules for individual pieces of equipment.
The aggregator system is able to handle discretionary loads. In other respects, it is subject to much the same limitations as IEEE 1547. An internet connection is unable to provide a time reference that is sufficiently precise for regulating phase angle, nor is it fast enough or reliable enough to provide a trip relay. Hence the DERs under the aggregator’s control must rely on grid sensing and a 1547-compliant grid connection for power quality and fault response. They are therefore subject to the same limitations on penetration level that apply under 1547.
What’s the Alternative?
If institutional barriers prevent the adoption of standards facilitating deployment of small DE resources, it will likely hinder the adoption of wind and solar energy resources. It will tend to make them more costly and less efficient. But it won’t stop them altogether. Instead, developers of wind and solar resources will be forced to co-develop and integrate large discretionary loads and energy storage systems. These will allow the integrated system to look more like conventional power plants, as far as grid operations are concerned.
There are various ways to accomplish that economically. One leading candidate, for example, is compressed air energy storage (CAES). It is both a large discretionary load, able to consume surplus power from wind or solar resources when a surplus is available, and a very efficient source of generation when supply is needed. In the future, small-scale CAES may be married with super-efficient fuel-cell / combustion turbine power generation. That will enable maximum energy “mileage” to be extracted from limited resources of natural gas or gasified biomass.
Nonetheless, it would be better for the country’s energy future if appropriate standards for self-management of small-scale DERs could be put in place. I’ll write again if I learn more about efforts in that direction.
NREL background paper on IEEE 1547 development.
Home page for EERE’s Distributed Energy program.
Home page for Distributed Energy journal. Many articles on DG market.
Good article on microturbines, focusing especially on Capstone.
“Intelligrid Architecture” announcement, with links.
Home page for “US Combined Heat and Power Association”.
Home page for Celerity Energy, Inc.