Frequency Regulation: Is Your Plant Compliant?

Introducing wind and solar into the grid highlights the importance of optimizing power plant frequency regulation capabilities.


By Ibrahim Abdur-Rahman, Sydney Niemeyer & Ricardo Vera, PE

The introduction of large amounts of intermittent renewable power (namely wind and solar) into electrical distribution grids has highlighted the importance of optimizing the frequency regulation capabilities of power plants.

The general decline in primary frequency response in all interconnections has prompted the regulatory entities to address the issue. Electric grids such as the one in Texas are especially sensitive to this issue due to their relatively small overall interconnected capacity as compared to other regional grids. The Texas Regional Entity (Texas RE) is actively working on a regional standard for frequency regulation. The Federal Energy Regulatory Commission (FERC), through Order 693, has also implied a minimum frequency regulation of the electrical grids.

This article reviews the requirements to attain optimum frequency regulation for the most common type of power plants, conventional boiler-turbine power plants and present the procedure for calibrating the boiler and turbine frequency regulation parameters. Also included is a discussion of the activities in the regulation and evaluation of the units’ frequency control.

Frequency Regulation

Electric grid frequency regulation is attained by the response of the turbine governors to deviations from nominal synchronous speed, the operation of the boilers-turbine controls in response to the frequency change and the actions of the dispatching system.

Frequency regulation success for any given boiler-turbine plant depends on many factors, primarily:

  • Steady state and dynamic stability of the unit
  • Load following capability
  • Linearization of the turbine governor valves steam flow characteristics
  • Proper calibration and coordination of the boiler and turbine frequency regulation parameters
  • Proper high and low limiting of the boiler and turbine frequency regulation based on unit conditions
  • Proper dispatching actions to restore the frequency to its normal operating value.


Another factor that influences a unit’s capability for frequency regulation is the available boiler energy storage. The larger the storage, the less the initial pressure drop caused by the quick opening of the governor valves and the better the initial unit frequency regulation.

The standard speed regulation setting for the turbine governors of the boiler-turbine generating units is 5 percent. This is a ± 5 percent change from rated speed (0.05*3,600 = 180 RPM), causes the turbine governor to change the governor valves position demand ± 100 percent. It is also generalized industry practice to add a small deadband (DB) to the calibration of the governor speed error bias to minimize the movement for very small speed deviations. The selection of the DB affects the fidelity of the regulation, as shown in Figure. 1.

The figure’s regulation curves are calculated by developing the equation ΔGVD= f (ΔRPM) for each DB, where ΔGVD is the change in the turbine governor valve demand as a function of the change in RPM.

Knowing the ΔGVD for any given ΔRPM enables the regulation calculation via the equation:

REG (%) = (100 * ∆RPM / ΔGVD)*(100/3,600)

The Electric Reliability Council of Texas (ERCOT) Protocols Section 5 has specific requirements for governor deadband settings. The maximum allowable deadband is ± 0.036 Hz, which has been the industry standard for mechanical “fly-ball” governors on steam turbines for many years. With the development of energy markets in the early 2000s, generators with electronic or digital governors began implementing this same deadband in their primary frequency response implementation. Unfortunately, the protocols were not clear on how to implement the droop curve at the deadband. Since the protocols required 5 percent droop performance, many generators introduced a “step function” or modified “step” once the deadband was reached in order to achieve near 5 percent droop performance outside the deadband.

A ±0.036 HZ deadband corresponds to a ± 2.16 RPM deadband. Based on the corresponding droop (regulation percent) for this deadband (Fig.1), a generator’s performance to frequency deviations during disturbances inside and near the deadband would be much greater than the 5 percent without some “step” function. These governor settings result in an abnormal frequency profile for the interconnection.

Figure 2 is the frequency profile for March and September 2008. It is clear that the “flat top” of the profile is centered around the +/- 0.036 Hz deadband. This provided challenges for both ERCOT and generators, as frequency spent as much time at the governor deadband points as it did at any point in between. It made it difficult for ERCOT to deploy the correct regulation ancillary service to bring frequency to 60 Hz and to meet the NERC CPS1 Requirement 1, since ERCOT had an epsilon-1 limit of 0.030 Hz. It also contributed to generator instability at the deadbands with the implementation of the various “step” functions in the governors.

Members of the ERCOT Performance Disturbance and Compliance Working Group (PDCWG) became increasingly concerned about the frequency instability and the realization of the risk of the step function in the governors. If generators that implemented step functions were separated electrically from the grid, they would experience extreme instability because of the step function. This would be caused by the governor providing excessive frequency response to the island to small generation load imbalances, resulting in large frequency swings and unit instability.

As a result of this analysis, a member of the PDCWG discussed the issues with one large generating facility to seek a solution. A team at the facility was willing to try different deadband settings along with a specific droop curve implementation. This implementation required a straight linear curve from the deadband to full range of the governor, eliminating any step function (see Table 1).

After testing different deadbands, a 1 rpm deadband (+/- 0.01666 Hz) was chosen. Four generator governors were set in this manner in November 2008. Fig. 3 illustrates the results of this governor implementation. The profile of ERCOT one-minute average frequency changed to a more “normal” distribution.

The general industry mindset for governor deadbands is to minimize generator movement due to frequency regulation. This may work for an interconnection where generators have various deadband settings. But when a majority of the generators set the deadband exactly the same and with a step function, the benefit completely disappears because frequency will move to the deadband frequently as demonstrated in the profile in Fig. 2. How much additional generator movement can one expect from this lower deadband and is it going to increase operating costs? This can be demonstrated by calculating the one-minute megawatt average movement of a hypothetical generator exposed to actual measured frequency using the two different governor settings.

An Example

Consider one generator with +/- 0.036 Hz deadband and droop step function and a second generator with +/-0.01666 Hz deadband and no step function. Using 2008 one-minute average frequency data, the generator with the lower deadband would have had 893,164.2 MW-minutes of primary frequency response. The generator with the larger deadband would have had 662,574.0 MW-minutes of primary frequency response. This is a 34.80 percent increase for the lower deadband generator.

This looks like a competitive disadvantage for the lower deadband generator. However, if the same comparison is made for ERCOT frequency data in 2009 where the new deadbands had an actual impact on frequency, interesting results are obtained. The lower deadband generator would now have had 692,039.8 MW-minutes of primary frequency response as compared to the larger deadband generator at 446,244.0 MW. That is a 55.08 percent increase in movement for the lower deadband generator. One interesting detail is that the megawatt-minute movement of the lower deadband generator is 4.45 percent higher than the movement of the larger deadband generator of the previous year: 692,039.8 MW-minutes compared to 662,574.0 MW-minutes.

Having the lower deadband in service for the entire year reduced the frequency movement of the interconnection and reduced the primary frequency response movement as well. The lower deadband generator megawatt-minute movement decreased 201,124.4 MW-minutes or 22.518 percent between 2008 and 2009.

This indicates the reduced impact on the generator with the non-step governor implementation with the small primary frequency response impact when the governor becomes active as compared to the “step” implementation.

Why not just eliminate the step function and leave the deadband at +/- 0.036 Hz? Referring back to Fig. 1, droop performance at 59.900 Hz would be around 7.72 percent with a +/- 0.036 Hz deadband as compared to 5.97 percent droop with the +/- 0.0166 Hz deadband. The difference is greater at 59.950 Hz with a 17.64 percent droop performance for the +/-0.036 Hz deadband versus 7.46 percent droop performance with the 0.0166 Hz deadband. If this larger deadband were selected, it is doubtful that the interconnection or balancing authority could meet the FERC interpretation of minimum frequency response with this larger deadband, not to mention the loss of the normal frequency profile. Without the primary frequency response of the lower deadband, the frequency profile would return to the “flat top” profile spanning the +/- 0.036 Hz deadbands, a less reliable state for the interconnection.

Based on this justification for the use of a 1 rpm deadband without a step function for the turbine governors, this value of DB will be used in the article. The governor speed error calibration for this DB is listed in Table 1.

Boiler Frequency Regulation Calibration

With the turbine speed regulation already decided to be 5 percent with a DB of 1 rpm, the next step is to calculate the boiler calibration that will match the turbine response to frequency upsets.

When the unit is synchronized to the grid, the governor’s changes in response to the speed deviations are translated into unit load changes that the boiler has to support. The boiler load demand is therefore biased based on the expected turbine-created load changes. The boiler load bias is normally in terms of frequency (Hz) instead of turbine speed (rpm).

The boiler load demand frequency bias DB is set the same as that of the turbine, which in terms of frequency will be: DB = 1 RPM * (60Hz / 3,600 RPM) = 0.016667 Hz.

The first step in the calculation is to find the relation between the turbine changes in megawatts with respect to the changes in frequency, namely the turbine (ΔMW/ΔHz).

As an example, consider a well-linearized turbine governor with a megawatt vs GVD characteristic defined by the coordinates (25, 200) and (90, 700). Proper linearization is attained by the proper calibration of the governor valve curves and the valves’ sequence.

For the above defined governor, the turbine load to governor valve demand ratio is given by:

∆MW / ∆GVD = 500 / 65 = 7.69 MW/GVD

This value will be referred to as the “5 percent frequency load regulation.” It is not based on the unit load rating.

Per previous discussions, the slope of the governor valve demand and the turbine speed deviation is:

∆GVD / ∆SPD ≈ 100 percent / (180 — 1) RPM ≈ 0.5587 percent/RPM

The relation between the change in speed and the frequency is:

∆SPD / ∆HZ = 3,600 RPM / 60 HZ = 60 RPM/HZ

Combining all the above incremental equations, the relation between the turbine changes in MW to the changes in frequency outside the governor deadband is:

∆MW / ∆ HZ = ∆MW/∆GVD * ∆GVD/∆SPD * ∆SPD/∆HZ =

7.69*0.5587*60 ≈ 258 MW/HZ

The turbine ∆MW/∆ HZ relation must be matched by a corresponding boiler load demand calibration in order to attain acceptable unit frequency load regulation.

The boiler frequency bias must bias the load demand at a fast rate because of the speed at which the turbine moves the governor valves. Depending on the type of boiler and the unit load rating, maximum limits need to be placed on the boiler bias to maintain the boiler within reasonable operating parameters. (For illustration purposes, a frequency bias limit of ± 200 MW will be selected.)

Each unit needs to be evaluated carefully to prevent the loss of the unit by a frequency load bias beyond its capabilities. It is also important to design as part of the unit frequency regulation scheme boiler and turbine frequency response limits based on the operator established unit “high” and “low” limits. These limits are normally set to maintain the unit equipment within existing operational capabilities.

Continuing with the boiler calibration, the desired boiler load demand calibration slope is:

MW / ∆ HZ = 258 MW / 1 HZ = 258 MW/HZ

As stated before, the boiler load demand frequency bias DB will be 0.016667 Hz to match the turbine’s 1 rpm DB. The end point of the boiler calibration curve for a 200 MW change can be found when the slope (258 MW/Hz) and the DB are known:

(200MW) / (Hz-0.016667) = 258 MW/Hz or END POINT Hz ≈ 0.792

The end point of the boiler load demand frequency bias is then (0.792Hz, 200 MW).

The definition of the boiler frequency bias is now complete and listed in Table 2.

With the proper boiler and turbine calibration, as described above, an optimum unit frequency load regulation can be attained. This, however, only takes place if the other factors (mentioned at the beginning of this article) that influence this regulation are also optimized.

Effect of Non-Linearity

One common culprit of variable frequency load regulation is non-linearity of the turbine governor flow characteristics.

The frequency load regulation of a boiler-turbine unit depends on the slope of the turbine steam flow characteristics. If the flow characteristics are non-linear, the percent frequency load regulation will vary according to the value of the slope at the operating point of the governor. At some operating point, the unit will not meet the desired 5 percent frequency load regulation.

Consider the variations in the steam flow characteristics of the linearized governor discussed before, as presented in Figure 4. The value of the slopes of the various segments is:

Ideal 5 percent regulation slope = 7.692 ∆MW/∆GVD

First slope = 10 ∆MW/∆GVD

Second slope = 4 ∆MW/∆GVD

Third slope = 13.333 ∆MW/∆GVD

Using the 5 percent frequency load regulation slope as reference, the regulation attained for any of the slopes is obtained by the ratio of the reference slope and the actual slope:

SLOPE REG percent = 5 percent*(REF SLOPE ∆MW/∆GVD / ACTUAL SLOPE ∆MW/∆GVD)

Therefore, the regulation for the slopes of Fig. 4 is:

1ST SLOPE REG percent = 5*(7.692/10) ≈ 3.85 percent (over-regulation)

2ND SLOPE REG percent = 5*(7.692/4) ≈ 9.62 percent (under-regulation)

3RD SLOPE REG percent = 5*(7.692/13.333) ≈ 2.88 percent (over-regulation)

This shows the importance of the linearization of the turbine steam flow characteristics in the optimization of the unit frequency regulation.

Unit Frequency Control Performance

The variability of the units’ turbine flow characteristics, dynamic behavior, boiler storage capacities and so on requires the establishment of a comprehensive standard that can fairly evaluate the units’ compliance with frequency control based on their individual characteristics.

The generator performance metric and measurement techniques must account for the deadband of the governor, the droop setting, stored energy available during the measurement period and the dynamics of the steam generator and its ability to restore steam conditions during the frequency recovery period. This includes loss of initial steam conditions due to the sudden change in steam demand caused by the frequency deviation, operation of the steam generator at a reduced inlet steam pressure and compensation for steam expansion at different governor/control valve openings.

Operating a steam turbine continuously at a 100 percent governor/control valve position at reduced loads (full sliding pressure operation at valves wide open) is a reliability trade-off for economics that is not considered justified. This method of operation will lead to cascading outages during severe grid events since all primary frequency response is lost. This requirement does not include steam turbines in a combined cycle configuration with combustion turbines since the steam turbine will follow the primary frequency response of the combustion turbines that provide their steam supply. Operating a generator at 100 percent output is also acceptable with limited or no primary frequency response to low frequency deviations, as it is the responsibility of the balancing authority to maintain adequate frequency responsive spinning reserves at all times.

All these factors are being studied and evaluated by the regulating authorities in their efforts to establish industry standards for frequency regulation.

More demands will be placed on power generating plants to contribute correctly to the electric grids frequency regulation as large quantities of intermittent capacity are connected to the grids.

In this article, the specifics for a conventional boiler-turbine plant were addressed, but each unit type has inherent characteristics that need to be understood to optimize its contribution to frequency regulation.

The correct coordination and calibration of the frequency regulation parameters of the boiler-turbine controls is a must to attain the desired results, but this will only be true if the other factors that influence the unit response are also optimized. The review of all these factors should include the evaluation of the adequacy of the unit control systems.

Standards and regulations are being developed to require and evaluate compliance of all units to the grids’ frequency regulation. These standards strive to achieve a fair evaluation based on the design and capabilities of the units.

Authors: Ibrahim Abdur-Rahman is an instrumentation and control supervisor with 32 years of power plant experience. He is the leader of the Electrical Apparatus Group (Electrical Specialist) at the NRG WAP plant. Sydney Niemeyer is a control system specialist with 38 years of experience in the power industry. He is a member of numerous ERCOT and NERC task forces, working groups and committees. Ricardo Vera is a power industry operations and controls consultant with more than 30 years of experience in the power field.

More Power Engineering Issue Articles
Power Engineerng Issue Archives
View Power Generation Articles on

No posts to display