Impact of Variable Generation Philip M Gonski

8/3/2019 Impact of Variable Generation Philip M Gonski

1/25


2/25

The current power system in the United States consists of meeting customer demand in

real-time and balancing supply and demand. With the proposed building of 145 GW1 of

installed variable generation capacity in the United States, the existing grid will need to

be examined for the long terms effects and limitations that this might place upon our

system. Interconnection requirements and standards must also be updated to reflect the

addition of a new source of power generation. With the establishment of IEEE

subcommittees as well as the IEEE 1547 standard, it appears that this topic is deservedly

receiving much debate and research. As several countries already undertaken vast

research and experience on this topic, the Untied States will have a large amount of

information and data available to aide in the adapting of the grid to the upcoming demand

for more renewable generation.

Findings and Discussion

Modeling Characteristics

One of the crucial issues facing wind power is the modeling and dynamic of wind turbine

generators on the current power grid. Coal and Nuclear turbine generator dynamics are

well-understood and have been use in models for decades. Thus, for conventional

generation, feasibility studies and impacts on the grid are well-known and well-

documented. Per the NERC, conventional generators are mandated to provide

comprehensive steady-state data and reporting procedures to model the dynamic

performance of the system. Such information is required to build a representation of the

system for planning, as required in NERC MOD-014 & MOD-015 2. Information gained

from model simulations is critical as it is required to perform load flow calculations, short

circuit analysis, as well as stability studies to determine the impact on the overall system.

Model characteristic data is usually provided by the manufacturer, however, in the case

of wind power, such information is not yet developed. Even more so, NERC standardsfor modeling and model validation have not yet been applied for wind generation. Until

models have been developed, several companies have relied upon using conventional

generator information. At the best-case scenario, user-written models for generators have

1 http://www.nerc.com/files/IVGTF_Report_041609.pdf


2


3/25

been developed which are both difficult to maintain as well as use. Thus, one of the

critical steps towards developing a renewable portfolio is the adaptation of a well-

recognized, industry-standard model, as is the case for traditional generation. One of the

hindrances facing such a practice is the current use of four different turbine generator

types with different characteristics.

The four turbines models currently in use today are classified into four types: squirrel

cage induction generators (type 1), wound-rotor induction generations with variable rotor

resistance (type 2), and double-fed induction generators (type 3), and asynchronous

generators with power converters (type 4). All of these generators, especially types 3 & 4

have dramatically different short circuit characteristics than synchronous generators as

can be seen in the figure 1 below. In this graph, the top waveform represents a wind

plant fault contribution, while the bottom represents a conventional synchronous

generator of a similar size. Such a graph clearly displays the errors that might arise in

using conventional generators to model the slightly more complicated wind generators.

Figure 1 Short Circuit Contributions of Double-Fed Induction vs Synchronous Generator3

Induction generators of types 1 and type 2 provide initial fault conditions similar to that

of a synchronous generator; however the current rapidly decays as flux collapses. Types

3 & 4 posses a high degree of controllability of both frequency and power output. Type 3

generators are especially difficult to model as the characteristics can be drastically

3 A Whirl of Activity http://www.ieee.org/organizations/pes/public/2009/nov/index.html

3


4/25

changed by closing a switch to purposely short the rotor during a fault for protection. To

further compound such issues, these types of generators are not well-represented in the

industry and are often little understood by the utility engineers who are involved in

performing grid reliability testing. Currently, to resolve these complex issues, the IEEE

has established a working committee tasked with fostering cooperation between turbine

generator manufacturers and vendors of short-circuit modeling software to ensure that

renewable generation is accurately modeled and analyzed.

As discussed previously, wind generation is vastly different from the characteristics of

conventional synchronous generation. Thus wind interconnections may require the use of

more detailed analysis methods. Wind systems often have very weak fault current levels,

series compensation, as well as complex protective relaying to accommodate the rapid

changes in output that might occur. Wind generators are often located in remote

transmission regions and are thus subject to small short circuit toleration, and a very large

voltage drop possibility due to the lower voltage that is often used for transmission.

Developing models for such transients is required to compare and select generators,

ensure compliance with grid regulations, as well as ensure full-voltage control and power

quality4 via the usage of shunt capacitors. Figure 2 below displays one example of a

complex wind generating system that must be modeled as part of analysis

Figure 2 Complex Wind Turbine Models5

4 A Whirl of Activityhttp://www.ieee.org/organizations/pes/public/2009/nov/index.html

5 A Whirl of Activity http://www.ieee.org/organizations/pes/public/2009/nov/index.html

4
http://www.ieee.org/organizations/pes/public/2009/nov/index.htmlhttp://www.ieee.org/organizations/pes/public/2009/nov/index.htmlhttp://www.ieee.org/organizations/pes/public/2009/nov/index.html


5/25

Grid Interconnection Standard Overview

In order to curtail any potential affects on the stability of the power system, certain,

interconnection standards have been established. The most important requirement is for

low-voltage-ride through (LVRT) in the event of system faults. During three-phasefaults, the generator must stay online for the normal fault-clearing time of up to nine-

cycles. Turbine generators built before 2008 must also remain online despite a voltage

dip as high as .15 p.u at the high side of the generator step-up transformer. Systems

installed after this period have recently been extended to 0.0 p.u. At the point of

interconnect, reactive power control of +/- .956. Certain groups, such the Canadian

Province of Ontario, have established very strict LVRT and interconnection

requirements7which are quickly setting the groundwork for future requirements to come.

Although power terminals of wind generation have very different behavior from

conventional generation, they are largely compatible with existing power systems in

operation. The Grid Codes specified by FERC 66A as well as IEEE 1547 which holds

true for conventional generation is consistently being met by commercial wind

generation. Voltage, output, and ramp control requirements have also been met when

requested.

In a majority of the United States grid, the output of dispatchable generation (generation

which may be controlled in output) resources follows change of demand. A small

percentage of generation in any area is designating as providing Automatic Generation

Control (AGC) in order to cope with the rapid and uncertain demand changes that may be

6 http://www.nrel.gov/docs/fy07osti/41329.pdf

7 http://www.ieso.ca/imoweb/marketdata/windpower_CA-ME.asp

5


6/25

experienced in a short period of time. AGC control will be crucial to the power grid in

managing the short-term uncertainty of variable generation, as well as the short-term

impacts that may arise from forecasting error.

AGC control consists of frequency and algorithms working together is maintain systemfrequency. The system is often located inside the system control center and monitors any

potential imbalances that might arise from generation and demand within a control area.

In this manner, the output of generators will be modified whenever a frequency change is

required to ensure equilibrium in the system. As variable generation is utilized, AGC

controls will need to trigger conventional generators outputs whenever there is any

dramatic drop-off of wind generation. Thus, if there is a meterlogical event that requires

the termination of wind generation on the grid, the system can react and ensure frequency

limits are within boundaries. With the usage of speed governing in modern wind

turbines, as well as their instant response to dispatch instructions, it is currently envisaged

that wind turbines may be allowed to participate in AGC systems in the near future8.

Nonetheless, as wind power penetrations continue to increase, AGC algorithms and

parameters will need to be optimized in order to ensure maximum system performance.

IEEE 1547 Distributed Generation Standard

To cope with the consequences of a move towards connecting distributed generation intothe existing grid, the IEEE 1547 Standard was developed in 2003. This document

provides a uniform standard for the operation, testing, safety, and maintenance of

interconnecting distributed resources. To prevent degradations of the grid, the source

must possess adequate voltage regulation, integrate with the grounding system of the

transmission, synchronize with the main grid, and avoid any inadvertent energization of

the surrounding power grid9. Thus, whenever the grid downstream of the turbines is

offline, the turbines must cease from outputting power to the grid to allow for operators

to service the necessary transmission section. During any faults that are experienced,

turbines must be able to remain isolated from the main system, as well as ensure

interconnection reliability. Any relevant breakers in the system must be designed to

8http://www.nerc.com/files/IVGTF_Report_041609.pdf

9 http://www.nrel.gov/eis/pdfs/interconnection_standards.pdf

6


7/25

coordinate with breakers at the transmission side to ensure full levels of fault protection.

Once a voltage, frequency, of synchronism event occurs in the system, wind farms must

be able to fully isolate themselves and then reassure their reconnection to the area grid.

Power quality concerns should also are also as plants must maintain a balance of real and

reactive power . The following figure 3 below details a typical interface as required by

IEEE 1547.

Figure 3 Typical IEEE 1547 Interconnection10

As can be viewed in the figure, there must be coordination at all times between the grid

area and local protective relaying system. Through such methods as paralleling

switchgear, the system should be allowed to synchronize from the grid, as well as open

breakers when necessary to ensure islanding during abnormal operation. Coordination is

also critical as distributed generation may lead to an increase in the available fault current

that can be provided to the system, as well as the breaker closing and fuse closing time.

Thus, care should be taken by the utility to ensure that the system has been modified and

can handle the new margin that has been added.

10 http://www.nrel.gov/eis/pdfs/interconnection_standards.pdf

7


8/25

Voltage Control

Wind Turbines also face the potential issue of voltage drop as a consequence of the

variable power output. Modern plants today have developed voltage control and reactive

power compensation that is increasingly comparable to conventional thermal plants.

Wind plants also may provide dynamic and standard reactive power support as well as

voltage control to increase stability. As can be seen in figure 4 below, modern wind

plants are able to maintain a somewhat constant voltage level despite power output

changes. Such control is done via a voltage regulator system which continuously adjusts

the reactive power output to maintain constant voltage output. Thus, as far as the

immediate power grid is concerned, wind turbine voltage does not have as much as a

negative affect as it might be expected if such complicated control methods were

designed and in place. In certain locations, this can be accomplished via power electronic

transmission technologies such as SVC ( Static Var Compensators) or STATCOMS. In

a typical SVC system, inductors and capacitors can be switched on to generate reactive

power, or generate negative reactive power. In one example of this system at CalCement

bus, without reactive power compensation the voltage variation drops to .905 p.u, while if

reactive compensation is added this number climbs to .95 p.u.11 Adding multiple

generators to a point of interconnect will also help to reduce voltage. Such a process is

called aggregation and will be referred to in greater detail further on. Figure 4 displays

an example of the variation reduction when multiple turbine generators are connected.

As discussed previously, the main voltage concerns arise due to the remoteness and low

voltage in use on available transmission lines rather than the wind generators themselves.

Voltage concerns were primarily a concern in the early days of wind turbines when direct

fed induction generators were used. However, with the advent of the double-fed

induction type these concerns have been greatly diminished as it is often a built-in

capability.

Figure 4 Voltage Variations as Multiple Generators are Added12

11 E. Muljadi and C.P Buttefield. Power Quality Aspects in a Wind Power Plant IEEE Conference

Paper. January 2006.

12 E. Muljadi and C.P Buttefield. Power Quality Aspects in a Wind Power Plant IEEE Conference

Paper. January 2006.

8


9/25

Figure 5 Wind Voltage at Point of Interconnect13


9


10/25

Double Fed Induction Generator

The double fed induction generator, unlike most conventional generators, deals with a

constantly varying torque by interconnecting both the stator and rotor to the power line.

Rotor windings are connected to the grid via slip rings and back-to-back voltage

converters which control both the rotor and grid current. In this manner, the rotor

frequency can differ from the grid frequency. Thus, it functions asynchronously. By

means of controlling the rotor current via the voltage converter, the active and reactive

power fed to the grid from the stator operates independently of the generators turning

speed. A basic design of this device can be viewed in figure 6 below

Figure 6 Double Fed Induction Generator14

Rotors for the induction generator are typically wound from two to three times the turns

of the stator. As a consequence, the rotor experiences higher voltages than the rest of the

system. Any voltage dips in the grid will have a magnified affect on the rotor of the


10


11/25

generator. To protect such occurrences, turbines employ a device referred to as a

crowbar which is able to short circuit the windings of the rotor if overvoltage or

overcurrent is detected on the system. Another benefit of the controllability of the rotor

voltages and current enables the induction generator to remain synchronized with the grid

even when the wind turbine speed is variable. Double-fed generators power converters

can also perform built-in reactive power control in order to regulate voltage and the fast-

response of the converters enables improved voltage recovery and ride-through

characteristics. As one can easily see, before the development of such technology,

regular induction generators were seen to be very complicated to integrate into the

existing system as they did not provide regulated output characteristics. With the advent

of this new technology, moderate variability of wind power will have minimal impacts

upon the system.

Wind Impact Affect on Operations

Perhaps one of the largest concerns plaguing wind power is the impact on the grid

operation. At any given moment, wind operators must now exactly how much power is

dispatchable to meet applicable load on the system. In the current market structure, day-

ahead forecasts are utilized which allow operators to consider the anticipated levels of

wind generation for the next operating day when making unit commitments. Real-time

forecasts are also implemented which re-distribute the available balance of power on the

grid at a set time period. As a direct consequence, the impact of wind variability on

operations is impacted largely upon the time sample upon which the entire system is

redispatched to match generation to demand. As is the case with the NY Independent

Operator, the entire system is redispatched every five minutes15. In doing this, the affect

of wind variability is reduced from one dispatch interval from the next. For some power

operators, the redispatch period is one hour upon which the variability of wind in this

time period may have adverse affects. Thus, for such power grids, a serious wind event

that occurs within this hour period can lead to system emergencies.

15 W. Grant, D. Edelson, J. Dumas, J. Zack, M. Alhstrom, J. Kehler, P. Storck, J. Lerner, K. Parks, and C.

Finley. Change in the Air.IEEE Power Energy Mag, vol. 6, no. 6, pp 47-58. Nov/Dec 2009.

11


12/25

Based upon the day to day forecasts placed upon wind turbines, system operators need to

organize sufficient spinning reserves to that any adverse affects on system balance will be

avoided. Net loads of wind turbines must be served after accounting for the fact that

wind has more variability than the load itself. Often, it is not seen as economical for

operators to counterbalance every wind variation with a load change in the system.

Perhaps the most drastic wind variation affect is ramping, a situation where there is a

drastic upturn or downturn of wind generation. A recent study performed by Xcel

Energy has shown that there are more high-ramp requirements with wind that without

wind. This implicates that higher penetrations of wind will likely increase the ramp

requirements for many hours of the year.

Figure 7 Ramping and Timing Challenges16

As a direct result of producing steeper ramping rates, generators may often be required to

operate at reduced output. The recent Western Wind and Solar Integration (WWSIS)

study has recently been completed on this topic and considers the overall operational

impact of higher penetrations of renewable resources. Results of this study can be seen in

Figure 8 below. In this study, 35% energy by renewables in the balancing area and 23%

renewablese in the rest of the WECC were modeled for the year 2017 using historical

weather data. Figure 8A highlights the base case when no new wind or solar power is

available. Nuclear and coal remain to base loaded to meet demand, while gas turbines



12


13/25

and hydropower remain for additional peaking abilities. Figure 8B highlights the

situation in an April month with 35% renewable. In this result, coal plants and nuclear

plants are required to be scaled down and cycle somewhat in order to balance available

power on the system with demand. At such a high level of penetration where wind power

is plentiful, forecasting accuracy is critical to meet this variability. During such times,

wind power output must be curtailed as there is more power output than current demand.

Figure 8C displays a 35% penetration level in July where the wind is relatively constant.

In this situation, coal and nuclear are base loaded, while any of the variations are done

with wind and available peaking units. Such behavior has a very minimal impact on

operations in the system.

Figure 8 Affect of Wind on Existing Generation17

17 D. Corbus, D. Lew, G. Jordan, W. Winters, F. Van Hull, J. Monobianco, and B. Zavadil. Up with

Wind.IEEE Power Energy Mag, vol. 6, no. 6, pp 36-48. Nov/Dec 2009.

13


14/25

At even higher rates of penetration as discussed in Figure 8B, existing units variability is

difficult to manage without the required ramping capabilities. On a real-time basis,

operators must be able to economically dispatch available power to match load. As a

result, accurate predictions of wind power will greatly aide operators to discern whether

or not spinning load is achieved. The uncertainty that wind power introduces into the

day-ahead forecasting system have been shown to increase system operating costs by up

to $5.00/MWh18 at wind penetrations of 20% or 30%. These numbers, however, are

strictly dependant on the nature of the associated dispatchable generation and its cost. . .

Recent history events have also dictated the importance of accurate forecasting, as well as

providing lessons upon some of the struggles that the Unit States has overcome to

incorporate variable generation.

Major United States Wind Events

The large impact of variable generation in the United States had a substantial even on

February 26, 2008. On this date, a very large downward ramp in wind production led

directly to a major system emergency. Figure 9 highlights the dramatic downturn in MW

output as a result of this downward ramp. Forecasts that were currently in use at the time

relied primarily upon each individual day look-ahead that was provided by each wind

generation source. Several of the methods used to provide these forecasts depended

solely upon the wind generators themselves and as a result, several of these report results

possessed great variations. Some forecasts were made by a commercial centralized

forecasting system which was in the process of testing by ERCOT at the time of the

incident. One of these tests is the 50% probability of exceedance, while another 80%

probability of exceedance was performed19. This accurate data was available to ERCOT

at the time of the incident; however, it was still in testing and was not considered

operational.

For a majority of the day prior to 5pm, forecasts were in line to the previous forecasts,

however, after this period there was a dramatic shift in predictions. Resource plans made

from the day-ahead forecasts indicated a gradual 1,000 MW decrease over the next three

18http://www.nrel.gov/docs/fy07osti/41329.pdf

19 These tests are based off a normal distribution curve and displays the likelihood of a event occuring

14


15/25

hours, while the commercial forecast had indicated a 200-500 MW rapid decrease from

5-7pm. Actual production, as can be seen in Figure 9, closely matched the commercial

forecast that was performed. As a result of this dramatic drop off in load, the grid did not

have available resources to match the drop-off in load. Although such an even did not

have a dramatic meteorological even associated with it, there was an atmospheric

decrease in wind speed due to a weakening atmospheric gradient. This even was

predicted in the commercial forecasting software which was not available to the system

operators at the time of the incident. Such an event highlights the importance of how

moderate ramps can cause large grid management issues, as well as underscoring how

even fairly common weather events can have a significant grid impact.

Figure 9 Wind Output Forecasts20.

Another significant wind event occurred on April 4, 2009. In this event, there was a high

wind cutout which leads wind output to drop from 650 to 450 MW within an hour time

span. Wind output continued to decline until it reached around 310 MW. In total, thewind reductions lead to a 10% decline in overall grid power availability. However,

unlike the ERCOT event, the commercial forecasting tools were in use and predicted



15


16/25

such an event. As a consequence, grid operators were able to schedule other spinning

reserves during this time period to ensure enough power was available to match the

roughly 3000 MW of power that was required. Clearly, with the growth of forecasting

tools now available, operators are now in a better position to predict weather patterns and

minimize operational impacts on the system.

Forecasting

Clearly, the largest impact on the power grid comes from the inherent variability of wind.

As can be seen in the previous examples, tools are now available which can minimize the

impact of wind by using a variety of commercially-available wind forecasting software.

Wind power production forecasts rely mostly on a combination of physics-based and

statistical models. Physics based models are referred to as numerical weather prediction

models (NWP). Physic based models are highly reliable as they are based upon sets of

equations which do not require training samples and are not limited by usage of

historical data. In disjointing themselves from historical data, they ensure that the model

may be able to predict events which had never occurred in quite the same way. Due to

this ability, such prediction tools are very costly and are limited by the incomplete

knowledge of the state of the atmosphere as well as the simulation parameters.

Statistical models are based upon the relationship between input (predictor) and output(forecaster) variables. Unlike physics-based forecasting, this analysis relies upon large

sets of historical data in both input and output. Thus, these models have the benefit of

learning from experience and planning based upon what was seen in previous years

without reliance upon the underlying physical relationships. Statistic models can be used

in a variety of ways to aide in the forecasting process. In man cases, statistical model

data is incorporated into NWP models to account for any terrain or native landscape

affects that can not be represented in the NWP model by them. As they learn from

experience, statistical models tend to predict typical events in any region and thus are not

able to prevent rare events which were not part of the historical data. To ensure greatest

accuracy, most generators reliable upon a variety of individual forecasts which can help

to ease any uncertainty due to faulty input data and any differences in model

16


17/25

configurations21. For the most part, short-term forecasting relies heavily upon statistical

models which can use recent data from nearby locations. Longer forecasts rely upon

NWP forecasts incorporated into statistical models. Perhaps the best path used by

generators is to use many different providers to generate forecasts. By doing this, they

will be able to gauge for themselves which model appears to work the best in the subject

area. In several recent studies, commercially available wind-forecasting can be shown to

provide 80% of the benefits that would result from perfect forecasting technology.

Depending on the mixture of available generation, a GE Energy integration study has

reported an approximately $95 million cost savings with more accurate forecasting

methods.

Figure 10 displays a snapshot of available system demand and wind power in two

situations. In the first graph, wind output and available demand matched quite well. As

can be seen in the second graph, there is simply a lack of wind power available when

demand for power increases. Such patterns can create operational challenges, however, if

such events can be predicted, their impact is greatly lessened on the system if there is

enough available generation that has been scheduled to meet demand

Figure 10 Alberta Load and Wind Output22



22http://www.aeso.ca/downloads/Wind_Integration_Consultation_Oct19_website_version.ppt

17
http://www.aeso.ca/downloads/Wind_Integration_Consultation_Octhttp://www.aeso.ca/downloads/Wind_Integration_Consultation_Octhttp://www.aeso.ca/downloads/Wind_Integration_Consultation_Oct


18/25

Factors Influencing Forecasting

Despite all the rewards of forecasting, there remains many variable issues which can have

a drastic impact on results. One of the most important factors is the quality of data. At

all moments, wind plants must report the turbine availability and output back to the

independent system operator. Lack of accurate data can often be attributed by the

arrangement of the anemometer stations which are used in forecasts. As can be seen in

Figure 11 from two neighboring sites, multiple measuring stations distributed throughout

the plant produces vastly more accurate data than the site with just one measurement

tower. The first chart displays information received from six separate wind stations,

while the second chart displays information from a nearby wind farm with only one

measuring station. One can easily view that having accurate measurements will allow for

much better forecasting and information to ease operational difficulties.

Ramping events can be very difficult to account for and are caused by a wide variety of

meteorological events. Some of these events which can cause ramp-ups include a cold

thunderstorm passage, rapid intensification of low pressure systems, sea breezes, as well

as thermal stability or vertical mixing. Ramping-down events can be caused by pressure

changes after thunderstorms, decrease in wind speed as a warm front passes, and

18


19/25

boundary-layer stabilization at sunset/nightfall23. Down-ramps are much more difficult to

predict as they are often not related to any meteorological events, except for the influence

of thunderstorms. Current NWP models can sometimes predict rapid changes in wind

patterns, however, the inherent variability of ramping events makes it very difficult to

accurate obtain the right forecast for every event as the causes are often very complex.

Figure 11 Impact of Multiple Weather Stations24

The variability patterns will have a magnified impact on wind generation if turbines are

concentrated in the same geographic region. If turbines are all located in neighboring

areas, they are all affected by the same weather patterns. In this manner, 100,000 MW ofwind output in the same region often acts as a single turbine 25. Aggregating wind

turbines can turn sudden interruptions in output into a more manageable multihour

downward ramp. This can be displayed in Figure 12.

Figure 12 Impact of Aggregation in Texas Event26


Finley. Change in the Air.IEEE Power Energy Mag, vol. 6, no. 6, pp 47-58. Nov/Dec 200



25 M. Milligan, K. Porter, E. DeMeo, P. Denholm, H. Holttinen, B. Kirby, N. Miller, A. Mills, M.

OMalley, M. Scheurger, and L. Soder. Wind Power Myths Debunked.IEEE Power Energy Mag, vol.

6, no. 6, pp 89-99. Nov/Dec 2009.

19


20/25

In this figure, the output of one individual generating station drops rapidly, while the

aggregated output of multiple turbines located throughout the state declines more slowly

as the weather event passes over the region. Aggregating has also been shown in a

recent study to decrease forecasting error by up to 50% if turbines have been located

throughout a 750km region27. In the current situation, many turbines are located in

remote areas and are serviced by only a few transmission lines. Often, these lines are at a

lower voltage which carries more current & has more losses. Due to the great distance

involved in distributing weaker voltages, voltage drop due to I^2R losses are critical.

Such complexities display that most of the wind turbines impact on the grid has been

lessened by technology and statistical analysis. As the United States moves more

towards renewable energy, we will need to look to those governments which already have

large quantities of wind power in place and learn from their experience.

Hawaiian Experience

The main island of Hawaii possesses one of the largest percentages of wind generation in

the world. In fact, upwards of 40% of their power portfolio can come from renewable



6, no. 6, pp 89-99. Nov/Dec 2009.



6, no. 6, pp 89-99. Nov/Dec 2009.

20


21/25

generation28. Since there is not many normal spinning sources present on the island, the

resources they do have must remain online at all times to provide adequate inertia to

preserve grid stability. As a result, as per Figure 13, excess power can often be produced

in off-peak hours, requiring that wind generation be curtailed. Wind variability has also

had impacts on the system frequency. Without wind, the system frequency deviated +/- .

06 Hz, while on a mildly variable period with wind generation the system frequency can

vary +/- .1 Hz29

Figure 13 Hawaiian Power Distribution30

To minimize system variability, ramp-rate limits of 2 MW per minute have been set for

both upward and downward ramps. By controlling its ramp rate in these directions, the

turbines are able to keep the power output relatively uniform. Problems were later

encountered in the programming of the AGC system. For a while, efforts were made to

make the system more responsible to variations in the overall system frequency.

However, in affect the variations were made worse as the AGC made efforts to chase

every random variation in frequency caused by wind turbines. As a result, generators

would increase their output to match the drop-off in wind generation, while in the mean

28 M. Matsurra. Island Breezes.IEEE Power Energy Mag, vol. 6, no. 6, pp 59. Nov/Dec 2009.



21


22/25

time wind output picked up once again to its former levels. Such issues were resolved

once the AGC system had been detuned to make it less reactive to variable changes in the

system. However, with this modification, the overall system frequency can deviate +/- .2

Hz before issuing commands to dispatchable units.

Spinning reserves must still be present at all time in the system and are driven at any

moment by the wind plant total output. Sustained ramping events and any frequency

variations must all be accounted for in the scheduling of both available reserves and

quick-starting generation. Figure 14 displays a ramping event experienced upon which

diesel generation reserves were brought online to ride through a severe ramping event.

As mentioned previously, these ramping down events can be very hard to forecast, thus

quick-starting capacity can be crucial in order to preserve system integrity. As is the case

with many other countries, the island of Hawaii is learning through experience to provide

an example of system success.

Figure 14 Hawaiian Wind Ramping Event31

Wind Power in Spain

Wind Power has dramatically increased in Spain due to strong local manufacturing as

well as policy support with feed-in tariffs and low-interest loans32. In total, Spain

receives around 11.5% of its generation from wind. For such a weakly interconnected


32 http://www.nrel.gov/docs/fy07osti/41329.pdf

22


23/25

system which possesses feeble ties to the European system, wind variability and

forecasting plays a crucial role. Any imbalances in the system must be less than 1,300

MW for ten minutes to prevent major system failure. Wind power has had the largest

affect on the availability of spinning reserves in the system. Once day-ahead forecasting

has been scheduled, the available reserve capacity is scheduled which should be available

to balance any possible ramping events on the system. If there is a predicted lack of

availability of spinning reserves, thermal users can be switched off in order to ensure that

there is enough available generation to meet the most users load demands.

Although wind has been widely used in Spain, there are still problems that will persist for

long periods of time. Unlike the United States, Spain is essentially an island grid with

very little interconnections to draw power from outside sources if required. In several

situations, they have needed to shed wind load, while in one situation, the difference

between real and scheduled wind production was greater than 7000 MW33. To cope with

these issues, the Spanish Grid Operator, Red Electrica has established a Control Center

for Renewable Energy to supervise and control generators in real time. Hydro-pump

storage is also being developed as a method to cope with the overproduction of wind

during several hours of the day. Interconnections have also been planned and are being

developed to reinforce their ties with other countries as another outsource of excess wind,

or to help during periods of low wind outputs. With the feed-in tariffs and other

government-run programs, Spain is currently believed to be on track to be more than 40%

renewable by 2020.

Conclusion

Despite a few growing pains, wind power in the United States will continue to grow and

develop into a more reliable and clean source of power. Although initially facing crucial

inherent difficulties such as voltage variation, frequency, and lack of proper forecasting

tools, new technology has arisen which alleviates a majority of these concerns. Wind is

inherently variable and thus can never be a base-loaded source of power, however, with

the highly accurate commercial forecasting software available today, it can definitely

help to meet the soon-to-be growing United States power demand. The evidence of the

33 http://www.ree.es/ingles/publicaciones/pdf/030409_MIT_WindpowerdevelopmentinSpain.pdf

23


24/25

growth of recent forecasting tools can clearly be seen in the example of the Texas system

emergency in February, as compared to the commercially forecasted April ramping

event. Although large variations in wind will continue to have impacts on grid

operations, allowing system operators adequate time to schedule available resources will

alleviate the major concerns of variability. Luckily for the United States, lessons can be

learned from Hawaii and Spain in regards to integration of wind resources into the power

grid. Despite all the operational issues which must be resolved, the main issue that will

hinder the usage of wind resources is the lack of transmission system capability. As

discussed previously, wind turbines are often located in remote areas and thus

interconnect at lower voltages and weaker transmission systems. Recent discussions of

smart grid improvements may help to account for the variability of wind by fostering

more interconnections and distribution of available generation. Perhaps one of the most

important modifications that will be made will increase the voltage of the transmission

lines entering the wind farm. By increasing the voltage, the voltage drop across the

cables will be minimized. It will certainly be a learning experience in the United States

as we move towards a less-predictable source of energy; however, with a few

modifications, these additions appear quite feasible, especially given the current political

climate.

Bibliography

http://www.nerc.com/files/IVGTF_Report_041609.pdf

http://www.ree.es/ingles/publicaciones/pdf/030409_MIT_WindpowerdevelopmentinSpai

n.pdf

http://www.nrel.gov/docs/fy07osti/41329.pdf

24


25/25

M. Matsurra. Island Breezes.IEEE Power Energy Mag, vol. 6, no. 6, pp 61. Nov/Dec

2009.

M. Milligan, K. Porter, E. DeMeo, P. Denholm, H. Holttinen, B. Kirby, N. Miller, A.

Mills, M. OMalley, M. Scheurger, and L. Soder. Wind Power Myths Debunked.IEEE Power Energy Mag, vol. 6, no. 6, pp 89-99. Nov/Dec 2009.

A Whirl of Activity

http://www.ieee.org/organizations/pes/public/2009/nov/index.html

http://www.ieso.ca/imoweb/marketdata/windpower_CA-ME.asp

E. Muljadi and C.P Buttefield. Power Quality Aspects in a Wind Power Plant IEEE

Conference Paper. January 2006.

http://www.nrel.gov/eis/pdfs/interconnection_standards.pdf

W. Grant, D. Edelson, J. Dumas, J. Zack, M. Alhstrom, J. Kehler, P. Storck, J. Lerner, K.

Parks, and C. Finley. Change in the Air.IEEE Power Energy Mag, vol. 6, no. 6, pp 47-

58. Nov/Dec 2009.

D. Corbus, D. Lew, G. Jordan, W. Winters, F. Van Hull, J. Monobianco, and B. Zavadil.

Up with Wind.IEEE Power Energy Mag, vol. 6, no. 6, pp 36-48. Nov/Dec 2009.
http://www.ieee.org/organizations/pes/public/2009/nov/index.htmlhttp://www.ieso.ca/imoweb/marketdata/windpower_CA-ME.asphttp://www.nrel.gov/eis/pdfs/interconnection_standards.pdfhttp://www.ieee.org/organizations/pes/public/2009/nov/index.htmlhttp://www.ieso.ca/imoweb/marketdata/windpower_CA-ME.asphttp://www.nrel.gov/eis/pdfs/interconnection_standards.pdf

Impact of Variable Generation Philip M Gonski

Documents