If you compare distributed generation (DG) units to service transformers, you're bound to find some similarities. Both are located at or near end users, both come in approximately the same power ratings, and both can supply electric power at utilization voltages to homes, businesses, and industrial plants. Even their looks are similar. Both are housed in large rectangular boxes and come in rated capacities ranging from 5kVA to 1500kVA and higher. Comparisons, however, can only go so far. As sizes become smaller, DG units are not able to supply the same level of peak demand and power quality as service transformers of equal rating. This difference becomes most pronounced when DG units ranging from 20kVA to 75kVA are used to serve individual or small groups of homes and businesses. At these small sizes, consumer load coincidence is low, and DG units can find it difficult to serve peak demand and/or track changes in load.

Consumer Load Coincidence

Engineers have traditionally sized major utility equipment (e.g., feeders, substation transformers, and transmission lines) based on coincident load curves. This is possible due to the large numbers of customers involved. Service transformers and small DG units, however, only serve a small number of customers and must deal with noncoincident loads.

Fig. 1, on page 18, shows coincident and noncoincident load curves for a typical home in the Midwestern United States. The blue curve represents the coincident load behavior for the average user in the home's particular rate class. The green curve shows the actual noncoincident load curve of the home.

No household's actual load curve looks like an expected load curve. Instead, most homes have daily load curves that resemble the green curve in Fig. 1. This demand pattern reflects large load changes as electrical appliances (e.g., water heaters, air conditioners, refrigerators, lights, etc.) cycle on and off. A typical household might have appliance sizes and duty cycles as described in the table, on page 18.

Averaged over many houses, household peak demand in Fig. 1 is 6.5kW. The peak for an individual household, however, can be much higher. During the course of a typical day, several major appliances might cycle on at the same time and result in a short “needle peak.” This noncoincident load behavior is what actually happens at the customer's meter.

Although every consumer has a load curve with needle peaks and valleys, no two homes have appliances that cycle on and off at exactly the same time. As a result, individual customers experience peaks at different times, and an average peak tends to be much lower.

Noncoincident Load Curves

Service transformers typically supply between one and 10 customers. As a result, they see mostly noncoincident load behavior. Depending upon the numbers and types of customers, the ratio of noncoincident peak demand to coincident peak demand (referred to as diversity factor) typically ranges from two to three.

The impedances of service drops, service transformers, and low-voltage circuits are engineered to handle voltage drops associated with needle peaks. But even though service transformers must serve noncoincident peaks, the transformers' kVA ratings are usually chosen based on coincident peaks. This is because transformer temperature, not instantaneous power, typically limits its transformer loading. Because the thermal time-constant of service transformers is typically 1 hour or more, needle peaks do not create appreciable rises in transformer oil temperature.

A 25kVA transformer is designed to serve a continuous load of 25kVA, but it can easily serve more for short periods of time, without cause for concern. It is not uncommon for utility personnel to allow a service transformer to experience short peaks above 200% of its kVA rating before considering replacement. As long as coincident peaks stay below the nameplate rating, internal transformer temperatures will usually remain within design limits, and engineers can expect years of satisfactory service.

The situation would be quite different with a DG unit. For example, if an engineer installs a 25kVA DG unit to replace a 25kVA service transformer, the DG unit would have to serve the same needle peaks, which could reach 50kVA. DG units, however, are electrically (not thermally) constrained and cannot easily handle loading above their nameplate ratings. A DG unit must be rated for 50kVA to handle a 50kVA peak load. Thus, DG units must be larger than service transformers to provide equivalent service, especially for small applications with low diversity factors.

Fortunately, there are other ways for DG units to handle needle peaks. The most common is the use of energy storage devices for temporary power boosts. These devices can recharge during periods of low loading but still must have inverters sized to handle noncoincident peaks.

For a DG unit to deliver satisfactory service, it must:

  • Provide steady-state power output equal to the coincident peak of its loads.
  • Store sufficient energy during off-peak periods to satisfy on-peak energy requirements.
  • Provide peak power output equal to the noncoincident peak of its loads.


The last requirement is the most difficult for baseload DG units to meet, and it's also the greatest power quality challenge typically faced in DG applications. Addressing needle peaks while maintaining adequate voltage and power quality is often expensive and commonly erodes the attractiveness of DG for baseload applications.

Load Tracking

Noncoincident load behavior creates another problem for DG units: handling the relatively large blocks of load that instantaneously and regularly cycle on and off. In traditional T&D systems, individual loads are miniscule when compared to central generation. Large loads are easily served, and only small reductions in service voltage occur. Stand-alone DG units, however, experience power mismatches when large loads cycle on. These mismatches can result in short but noticeable voltage sags, which continue until the DG unit adjusts its power output. If loads cycle frequently, voltage fluctuations become a source of irritation for customers. In worst-case scenarios, these fluctuations cause sensitive loads to misoperate.

Fig. 2, on page 18, compares a typical transient voltage response of a 25kVA service transformer to an inexpensive and an expensive 50kVA gas microturbine. The red line shows the voltage response when an air conditioner is turned on, creating a sudden increase in demand from 23.5kW to 30kW. This increase is representative of three houses being served by the same device (service transformer or DG) when the air-conditioning units turns on.

The service transformer (solid green line) responds to the increase in demand by slightly lowering its voltage — a reaction to the increased current interacting with system impedances. Customers will barely notice this voltage reduction, except possibly for those with poor house wiring.

The low-cost DG unit (dashed line) actually provides better steady-state voltage than the service transformer. However, a transient voltage sag of 5% occurs as the rotor accelerates to meet the 30% increase in load.

Fig. 3, on page 18, shows flicker curves derived from empirical studies. Based on this, customers will notice virtually every 5% voltage sag and will become irritated when they occur more frequently than 0.03 times per minute (about once every 30 minutes) — a very real possibility as multiple air conditioners at three houses cycle on and off.

More expensive DG units have better load-tracking ability. The dotted green line in Fig. 2 shows the response of a 50kW unit with a battery storage unit and power inverter. Using stored energy from the batteries, this DG unit has a near-instantaneous ability to alter its output to track changes in load. This particular unit has a 1% sag when load increases 30%, which would have to occur more than 10 times per minute before irritating typical customers.

One final way to address load tracking is by connecting the baseload DG unit in parallel with the electric utility supply. The DG unit serves daily coincident loads, and the electric utility provides load tracking and voltage regulation services to address noncoincident peaks and sudden load changes. Over the course of a day, the utility provides little energy but considerable voltage and peak-shaving support.

If electricity rates are based on energy consumption, then baseload DG connected in parallel with the utility is a bargain because only a small amount of utility energy is consumed. Ultimately, however, the utility must be able to recover the cost of this connection, and it will have to impose capacity charges roughly equal to 30% of typical energy-plus-capacity charges.

Conclusion

Small distributed generators have been successful in backup applications. In the era of microturbines, hourly rates, and retail competition, there is a tendency to predict similar success for small DG units in baseload applications. But all things are not equal. A 25kVA service transformer is able to supply temporary loads well above its rating; a 25kVA DG unit cannot. Service transformers can easily track large changes in load without causing voltage sags; DG units cannot. Of course, you can size and engineer distributed generation to overcome these problems, but the added costs make baseload DG hard to justify.

From our perspective, DG will continue to find economic applications in niche markets where electricity prices are high enough and gas prices are sufficiently low. No one can predict the future, but it's safe to say that baseload DG applications won't make the electric grid obsolete anytime soon.

Mr. H. Lee Willis is an IEEE Fellow and the vice president of technology and strategy for ABB Consulting. You can reach him at lee.willis@us.abb.com.

Dr. Richard E. Brown is the director of consulting IT for ABB Consulting and a senior member of the IEEE. He can be reached at richard.e.brown@us.abb.com.