IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 2, NO.

3, JULY 2011 321

Reserve Requirement Impacts of Large-Scale
Integration of Wind, Solar, and Ocean
Wave Power Generation
Douglas A. Halamay, Student Member, IEEE, Ted K. A. Brekken, Member, IEEE, Asher Simmons, Member, IEEE,
and Shaun McArthur, Member, IEEE

Abstract—Many sources of renewable energy, including solar, Wind power is now a very mature and established renewable
wind, and ocean wave, offer significant advantages such as no fuel resource throughout the world. However, other renewable
costs and no emissions from generation. However, in most cases power sources such as solar (photovoltaic (PV) or concen-
these renewable power sources are variable and nondispatchable.
The utility grid is already able to accommodate the variability of
trating/thermal) and ocean wave energy also have significant
the load and some additional variability introduced by sources potential. Solar has recently been the beneficiary of a number
such as wind. However, at high penetration levels, the variability of large-scale initiatives (e.g., California Solar Initiative), and
of renewable power sources can severely impact the utility reserve there is currently a good amount of industrial activity regarding
requirements. This paper presents an analysis of the interaction wave energy, particularly for development in the Pacific North-
between the variability characteristics of the utility load, wind west of the U.S. [3], [4].
power generation, solar power generation, and ocean wave power
generation. The results show that a diversified variable renewable A. Wind, Solar, and Wave Generation Characteristics
energy mix can reduce the utility reserve requirement and help
reduce the effects of variability. Each of these renewable power sources can be described by
three major characteristics.
Index Terms—Load forecasting, load modeling, marine tech-
nology, power systems, power system stability, reserve require- 1) Variable: The output power of a large-scale wind, solar,
ments, solar power generation, wind power generation. or wave power plant varies over time. The majority of the
time, the variability from one minute to the next is very
small, and even the hourly variation is usually small. How-
NOMENCLATURE ever, on occasion the output of a large plant, as high as sev-
eral hundred megawatts, may go from full output to low
BPA Bonneville Power Administration. production or vice versa over several hours [5].
BAA Balancing Authority Area. 2) Nondispatchable: As implemented now, the system oper-
ator has very limited control of the output of large-scale re-
newable generation. In general, the operator must deal with
I. INTRODUCTION whatever the renewable generation outputs are in much
the same manner as dealing with the load. Therefore, it is

M ANY balancing areas within the U.S. are faced with a
rapidly expanding wind power resource. For example,
it is predicted that as much as an additional 5000 MW of new
common in the analysis of the impact of renewable power
generation to subtract its contribution from the load. At low
to moderate levels, renewable power generation appears as
wind power could come online within the next five years in the a negative load.
Pacific Northwest [1]. 3) Energy source: Due to the nondispatchable nature of wind,
At low penetration levels, the variable output of wind power solar, or wave, they generally have a relatively low ca-
plants is easily absorbed within the variability of the load pacity credit [6]–[9]. That is, they do not make a signif-
[2]. However, as the penetration level increases, the added icant contribution to the power requirements of the grid
variability of the wind resource can cause greater ramp-rates, for planning purposes. However, each Joule of energy con-
greater interhour variability, and greater scheduling error. verted by a renewable source is one Joule saved for “tradi-
This ultimately increases the amount of generation the system tional” generation, such as coal. Therefore, renewable en-
operators must hold in reserve (i.e., the reserve requirement) to ergy sources can make a significant impact on the energy
accommodate the unplanned excursions in wind generation. requirements of the grid.

Manuscript received September 14, 2010; revised January 04, 2011; accepted B. Reserve Requirements
February 07, 2011. Date of publication February 17, 2011; date of current ver- The variable, nondispatchable nature of wind, wave, and
sion June 22, 2011. This material is based upon work supported by the National
Science Foundation under Grant 0846533. solar has a significant impact on the utility reserve requirements
The authors are with the Department of Electrical Engineering and Com- [10]. Analyzing the effect of these renewable energy sources on
puter Science, Oregon State University, Corvallis, OR 97331 USA (e-mail: the reserve requirements provides a meaningful and concrete
brekken@eecs.oregonstate.edu).
Color versions of one or more of the figures in this paper are available online
method of characterizing the variability of a given renewable
at http://ieeexplore.ieee.org. energy source, including its short- and long-term correlation
Digital Object Identifier 10.1109/TSTE.2011.2114902 with the load.

1949-3029/$26.00 © 2011 IEEE

and wave power data is 10 min. and wave data sets is clear.322 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY. 5% solar. including reserve requirements [11]. If the supposition holds. as this more closely reflects reality for the Pacific and is much more geographically restricted. defined as the Reference [16] investigates a parallel branch of research. and wave power plots are at a 5% penetration level. and 5% wave penetration.g. which uses the moving water due 6) 5% wind. The quantifying the increase in reserve requirement as a function wind data is scaled directly as necessary to achieve the desired of penetration. JULY 2011 C. The solar and ocean wave data were generated. and solar. Wind and Using the data from the measurement buoys. The advancing the analysis of the reserve requirement calculations. and 2. power. Fig. and solar and ocean power data generated from resource measurements. using actual load and wind power data. scenarios in which there is a greater A. in place. a day in January). covers the cal. ratio of peak power to peak load demand over the year. [12]. The various data sets at 5% is then averaged over 10-min intervals to generate the power penetration are plotted in Fig. Research Objective A number of research groups have analyzed the impact of large-scale wind integration on utility operation. The reserve requirements for six scenarios are compared. B. the all scenarios. The converter wind.5 MW. This total is then scaled as necessary to achieve the de- eration appears to be much more seasonally uncorrelated. wave or solar. it is clipped at that level. and the analysis. The area of study is within the Pacific dominant period. Reference [14] demonstrates a technical method for reserve requirement calcu- lation for high levels of renewable power penetration in general. 2. such that there are 52 704 data points produces a capacity factor of 50% for an average winter day in all year-long data sets. wind). 2) 15% wind power penetration.5% solar. The power from each of the three parks imum load for 2008. Wind diversity of renewable power will have lower reserve require- ments than for the scenario with only 15% wind penetration. 1) no renewable energy (only load). data.5-s sample time for a 5 by 80 from BPA [17]. time-series water load data for the BPA BAA are both freely available directly surface elevation data at a 0. The supposition is that the temporal and spatial variability characteristics of wind. . For for a few populous areas. whereas this paper focuses on the impacts of penetration rate for a given scenario. 3. (e. All data. with daily averages substituted time-series (52 704 points). solar. wave. mostly within or near the Bonneville Power Admin. and ocean wave energy specifically. solar.. the spectral significant wave height. solar. In short. Tidal energy conversion. cations in the grid is occupied by a 250-kW generic wave-fol- endar year 2008 starting January 1. The wind power data used is for the approximately 1600 MW This paper builds off of [15] by adding enhancements to of wind within the BPA BAA for 2008. VOL. This includes roughly the wave power data generation methodology and further 15 wind farms throughout the Lower Columbia region [18]. However. 10 754. 4) 10% wind and 5% wave penetration. 5) 10% wind. to tidal variations to generate power.g. 2. different measurement buoys in the Pacific Northwest for 2008. If the initial data power output is assumed to be proportional to the vertical water was available at a higher sample rate. such as Puget Sound. wind. while the wind gen. The sample time for the load. sired penetration rate for a given scenario. Unless stated otherwise. 1. Plots of year-long data (averaged daily) for load. wind penetration is greater than or equal to either overall tidal resource is not as large as wind.S. The research presented in this paper analyzes the reserve re- quirements for the Pacific Northwest of the U. 1. Some research has also been extended to include large-scale solar [13]. The tidal resource in the Pacific Northwest is strong load within the year to the peak generation within the year. grid with 100-m spacing is reconstructed. it was down-sampled to surface velocity squared. and solar will allow a greater combined penetration rate than using only one predominate type of renewable power source (e. wave. NO. The power from the three locations for the raw 10-min data for clarity. was not considered in this Penetration is defined in this research as the ratio of the peak research.. Each of the 400 lo- as described below. If the instantaneous power production 10-min intervals by block averaging the higher resolution data exceeds 250 kW. diversification. and simulated sources. Wave II. DATA SOURCES The methodology for generating the wave data is described The data used in this study came from a variety of both real in detail in [19]. solar. All data was normalized to the max. and wave 3) 10% wind and 5% solar penetration. The proportionality is within each 10-min block. The wind. lowing point-absorber wave energy converter. where a large and growing wind capacity is already be extended to include tidal. solar. Future research will Northwest. and dominant direction is collected from three Northwest. istration (BPA) Balancing Authority Area (BAA).5% wave penetration. unadjusted power represents a 14% penetration. including wind. all data is set such that the combined output of the wave energy converters at a 10-min sample time. The seasonal variability of is added together to produce the total wave power generation the load.

it even if the forecast is accurate. it is essential the accuracy and frequency of the forecasted generation/load. in to implement. “Variability” refers to the natural variation in resource output. METHODOLOGY average power generation/load for that hour. including wind. while having a large impact on the hourly or daily re. An analysis on the reserve requirements impact as a function of CSP storage time is a good topic for future forecasting method for the highly variable renewable resources. using it as a method to forecast load is not optimal. Solar Irradiance (power per area) data was gathered for 10 dif- ferent locations in the Pacific Northwest from the University of Oregon Solar Radiation Monitoring Laboratory: Aberdeen. troduced the concepts of “variability” and “uncertainty” [12]. III. Dillon. more advanced forecasting methods that include The second time scale of interest. For the concentrating solar data. solar. The first. However. Given the focus on reserve requirements. is defined as the difference between the 10-min average B. While previous studies [22]–[24] have focused on examining Both the following and imbalance reserve requirements are the correlation between various renewable resources and loads. the imbalance component of the reserve requirement is forecasted to grow rapidly given current A. to generate the total output power for that site. may offer longer energy storage times. The load . and imbalance (difference between hour average and forecasted/scheduled) re- serve requirements. Portland. To generate the PV data. and ocean wave. is defined as the difference the renewable energy resources in this paper. so the regulation reserve com- the variability of these resources on reserve requirements for ponent is not represented as it requires 1-min sample time data. With re. serve requirements. and ocean wave. the variability of quirement is consistently smaller than the following and imbal- both the generation and the load must be examined. or daily basis. of the available generation can present a particular challenge. a BAA. illustrated in Fig. For relatively short forecasting hori- 1A time constant of six hours was chosen as a conservative estimate to in. This total is scaled as necessary to achieve the desired Fig. Forecasting power generation/load and the minute-to-minute power genera. in order and the 10-min average power generation/load. between the hourly forecasted generation/load and the hourly solar. oil-based. hour requirements. 2. the tion/load. The data was processed to a 10-min sample time over the 2008 calendar year. ference between a perfect forecast and the actual forecast. Eugene. research. Salem. It should also be noted that some research groups have in- which. Bend. definitions are essentially the same as “following” and “imbal- Three different time scales are currently used by the BPA to ance” as defined above. particular molten-salt-based systems. In order to calculate imbalance reserve requirements. These ferent types/time scales of reserves is necessary. a six-hour thermal time constant is assumed. 2. In practice. Each of the 10 sites is then added together to produce the total solar data. calculate reserve requirements for the BAA [11]. Each site is weighted equally. It is assumed that each of the 10 locations is 50% PV and 50% concentrating solar (CSP). hourly. While the one-hour persistence method is viable as a baseline This may decrease the solar reserve requirements. and unspecified CSP systems. This time scale accounts for small changes in power scheduled or forecasted power must be determined for both the demand or supply that can be readily met through Automatic renewable resource and the load. the irradi- ance is used directly without processing. over the 2010–2012 time frame. often has less of an impact on the intra.: RESERVE REQUIREMENT IMPACTS OF LARGE-SCALE INTEGRATION 323 C. Time Scales for Reserve Requirements forecasting tools and methods [11]. forecasting minute resource variability [11]. persistence methods are reasonably accurate. The available data for use in this paper this paper presents the results of an analysis on the impact of was at a 10-min sample time. following. each weighted at 50%. Both of these data sets are added together. and simple clude molten-salt. reg- ulation. The ir- radiance data is run through a simple one-pole low-pass filter with a time constant of six hours1 [21]. The imbalance In order to fully characterize both the variability and fore- component of the reserve requirements is directly impacted by castability of various renewable energy resources. Definition of following (difference between 10-min and hour averages) penetration rate for a given scenario. zons. Ashland. ance reserve requirements due to the generally small minute-to- newable resources like wind. Some systems. results. This time scale to keep the forecasting component from introducing bias in the accounts for larger changes in the power demand or supply. respectively. to define a consistent set of tools and methods that can be uti- With the large increase in wind power generation in BPA’s BAA lized for analysis. the one-hour persistence method was used for all of The final time scale. Generation Control (AGC) via spinning reserves. In order for a BAA to balance generation with load This should not affect the analysis as the regulation reserve re- on a minute-by-minute. is defined as the real-time meteorological measurements can provide improved difference between the hourly average power generation/load forecasting over persistence methods [25]. Burns. “Uncertainty” refers to the dif- readily becomes apparent that a clear understanding of the dif. and Twin Falls [20]. Hermiston.HALAMAY et al. imbalance.

2007 ( points) was hourly average power was computed (imbalance). and seasonal basis [11]. ments involves sorting the following and imbalance calculated sents an average day of December 2007. 4). the bottom plot. This correction power generation for the BPA BAA. Both the fol- averaged element-wise with the load power time-series for the lowing and imbalance components were stored in an array for 24-hour period of December 2. and forecasted load (similar to Fig. period. Utilizing the forecasting methods in the previous section. a simple C. VOL. NO. 3. historical load growth. The power at 2:00 is mean [11]. Unlike wind.25% of outliers. 4 demonstrates the various data sets for wind forecast at 1:50 to the forecast at 2:10. the calculated following and imbalance reserve In these situations. while eliminating those extremely rare events where data points 2:00 to 2:50 for the average December day of 2007. 2008 and 1:00 to 1:50 December 1. For example. and solar. weekly. required reserves are beyond three standard deviations from the as described in the preceding paragraph. the difference between the hourly average power and essary given the seasonal variability in the shape of the daily each 10-min power was computed (following) and the differ- load curve). The end result is a load power time-series that Industry-standard practice for computing reserve require- covers a 24-hour period (at a 10-min sample time). wind.5% of the data points. Fig. determined as the midpoint of the linear ramp from the previous The plot in Fig. as seen in Fig. reserves and then keeping the middle 99. 5 demonstrates the following and imbalance 1:50 December 1. While BPA utilizes a sophisticated load forecasting eration) is simply the sum of the forecasts for each component: model that considers parameters such as temperature variation. this trimming of the data still enables the controllers to over the hour for the corresponding hour in the previous year meet the necessary North American Electric Reliability Corpo- average day for that month. According to The basic load forecast for the coming hour is the average power BPA. For example. The A correction term is added to this basic forecast to account solid line in the plot is the collection of actual combined wind for the load-growth from one year to the next. the algorithm uses a logic test to examine those situ. the load power time-series for the ence between the scheduled/forecasted hourly power and each 24-hour period of December 1. December 2007. eliminating the top and bottom 0. for the purposes of this paper. a simple 1-h persistence forecast is used for components are illustrated. The top plot of Fig. the correction term added for that hour). The 2007 and imbalance reserve requirements. Following and Imbalance Calculations model was used. it is easy to see that the forecasted . the basic forecasted ration (NERC) and Western Electricity Coordinating Council power for the hour 2:10 to 2:50 (five data points at a 10-min (WECC) reliability guidelines for balancing power supply and sample time) on December 1. Load variation over a four-day period in January 2008 [17]. 2007. 3. it is In order to forecast load for 2008 (the period under study). 2. JULY 2011 Fig. of 2008. curve follows a consistent pattern on a day-to-day basis. reserve calculations for the load demand for the same one-day Finally. for the period the scheduled load for that transition hour. Example of comparison between wind power output versus hourly av- erage versus scheduled output. power generation for an example one-day period in 2008. 2008 is the average of the six demand. load. while in day and the next to prevent discontinuities in the forecasted load. The dashed line is the term is the difference in the hour-average power in the previous hourly average power (an average of the six 10-min data points hour for 2007 and 2008. This consistency allows for a much more accurate load The total forecast for the net load (load minus renewable gen- forecast. hourly ations where there is a transition between one month’s average average load. [26]. For example. Moving through each hour data was processed to calculate each month’s average day (nec. then possible to process the 2008 data to calculate the following historical load data for 2007 was used as a baseline. 3. 2007 and so on for all days in ease of access and analysis. the load is forecasted 10 min prior to the hour. 4. The dashed–dotted line is the scheduled/fore- to the forecast for 2:10 to 2:50 December 1. 5 shows the actual load.324 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY. and repre. starting at time marker 2040. above) is the difference between the average power for 1:00 to The plot in Fig. As an example. wave. 2008 (as described casted power for the hour. and load variation on a daily.

Example of comparison between load demand versus hourly average Fig. The versus scheduled demand with calculated instantaneous following and imbal. In comparison. dashed–dotted black lines represent the overall maximum and minimum ance reserve requirements. 6 versus the larger discrepancy be- tween the two in Fig. Even though the forecast for the wind ation. 6 show the following and imbalance reserve than it is in Fig. It is essential to note. which correspond to the need for the system operator to decrease or increase generation. the actual maximum imbalance requirement is 0. For ex- the much more predictable nature of the load demand curve.25% outliers. 5 is generally much more 6. For this specific example.25% outliers.5 to 0. while the “trimmed” requirement is Fig. With the expected rapid growth of installed . 5. Fig. is useful in order to better examine the impact of the renewable resource on the power system. Example of load minus wind imbalance over four weeks.25% of out- liers. pact. the power generation from renewable resources looks similar to a negative load to the system operator. and it can thus be treated in this manner. 4. 6.HALAMAY et al. This is simply due to on the following and imbalance reserve requirements. 6 is how close the scheduled/forecasted power is to the actual load minus wind line in Fig. and the red solid lines represent the maximum and minimum of the bottom curve over the entire year. discussed further in Section V. excluding the top and bottom minimum reserve requirement excluding the top and bottom 0. 0.0428 pu.05 to 0. this same order of 0. In comparison to the plot in Fig. These plots clearly show how the one-hour persistence generation is much less accurate than the load forecast. It is the “trimmed” reserve requirement that is used erage versus scheduled demand with calculated instantaneous following and for planning purposes and that is discussed in Section V. 4. The final reserve requirement is the maximum Of particular interest in a comparison between Figs. the order of 0. when method can cause the imbalance reserve requirements to be the two are combined the wind forecast has relatively little im- larger than the following reserve requirements. imbalance reserve requirements. or load demand minus gen. maximum and minimum refer to the most positive and most negative values for the reserve.25% of outliers. Upon appli- cation of the outlier trimming procedure. 5. creating a negative growth in the imbalance impact that the relatively small wind power generation (on the plot below. 5 and it is clear that the load forecast in Fig. Nevertheless.8 pu). for the same one-day period in 2008. To better clarify the calculation of the reserve requirements. or load minus resource as will be 1600 MW [18]. with a maximum installed wind capacity of artifice of load minus wind.1 pu) has on the much larger load demand (on time-period demonstrates an over-forecast in the wind gener. calculation for load minus wind. discussed herein. The strate 2008 data. The final reserve requirement is the maximum and reserve requirement. comparing the bottom plots in Figs. This difference is due to the limited load was too low. however. ample. respectively. that these plots demon- erated wind power. 7. 0. Example of comparison between load minus wind versus hourly av. 5.: RESERVE REQUIREMENT IMPACTS OF LARGE-SCALE INTEGRATION 325 Fig. Note that in this context. 4. excluding the top and bottom 0. the red solid lines rep- resent the new maximum and minimum imbalance reserve re- quirement values that exclude the top and bottom 0. the maximum following requirement is larger in Fig. it is apparent that the addition of the wind has had an effect accurate than the wind forecast in Fig. 7 demonstrates an example of the imbalance reserve over a four-week period with the actual maximum and minimum values denoted by the black dashed–dotted lines. 4.0386 pu. and and minimum of the bottom curve over the entire year. 6 The plots in Fig. At low pene- tration levels.

2. pected. it is 412. for the imbalance time scale (Table II).3 MW (0. which presents the load minus 15% wind wind it is 447. by which the forecasted net load exceeded the actual net load. this implies an increase in the number of events renewable generation than expected. the system The next data column presents the load minus 10% wind and operator must increase dispatchable generation elsewhere in the 5% solar scenario. Table II). This can be directly compared to the to be 379. corresponds to cast on the imbalance requirement will grow. The following and imbalance reserve requirement distribu. The variance also increases. the maximum row.3 MW (0. hourly average (for imbalance.25% of outliers are removed. with the second tions that were generated from the load.5 MW) TABLE II RESULTS FOR IMBALANCE RESERVE REQUIREMENTS DISTRIBUTION (PU. differences over the year between the hour average and 10-min the reserve requirement is improved compared to the wind average (for following.03832 pu).04159 pu). The max and min repre. and 5% wave. Because larger errors are weighted more heavily in row. sent the necessary reserve requirement to cover the variability More significant is that the RMSE increases more than the for a given time scale (following or imbalance). and for the load minus 5% scenario. This is the decremental reserve. JULY 2011 TABLE I RESULTS FOR FOLLOWING RESERVE REQUIREMENTS DISTRIBUTION (PU. 3. This is called incremental reserve. 5% solar. with the diversification of renewable resources. for the load minus 15% wind distributions. added by wind to the net load increases the reserve require- tics presented in Tables I and II. being negative. it is obvious that by . respectively. NO.03527 pu). The minimum MAE. represents cases in which there was less the RMSE.5 MW) wind capacity to over 6000 MW. while for the load minus 15% second data column. requiring larger reserves. Similarly. VOL. or some combination of both. Pbase 10754. This scenario most closely represents the current re. decremental The minimum (min) and maximum (max) are the largest reserve). Pbase 10754. or a combination of both. This is discussed cases in which the renewable generation is greater than ex- further in Section V.1 MW (0. after the top and The mean absolute error (MAE) and root mean square error bottom 0. wind. The operator must decrease dispatchable generation to IV.. The first data ments. On both These numbers represent the maximum amount over the year the following and imbalance time scales (particularly the imbal.326 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY. ance time scale). the impact of the wind fore. wind. generation to accommodate the over-forecast (i. Comparing the first column (only load). the column in each table represents the base load reserve require. or greater load than ex. For example.e. solar. pected. RESULTS balance the system. and wave column (load minus 15% wind penetration). being positive. the deleterious effect of wind on utility reserve and thus the necessary amount of decrease of dispatchable requirements is obvious. or the load is less than expected. Comparing the results in this column to those system to cover the deficit. Therefore. decremental reserve required for the load alone is calculated ment for each time scale. newable resource portfolio in the Pacific Northwest. Thus. Table I) and the hourly forecast and alone case. (RMSE) both increase significantly for 15% wind penetration. the variability data for the various scenarios are characterized by the statis.

In these two charts. 8 and 9. however. which means that the net load Opportunities for future work abound in this research area. more apparent. the increasing strain on BAA reserves can utility perspective. . the overall reserve require- RMSE. would require significant rent portfolio of wind alone. with the important note that presenting the load minus 10% wind. 9. amount in the equal mix scenario.: RESERVE REQUIREMENT IMPACTS OF LARGE-SCALE INTEGRATION 327 Fig. Comparison of incremental and decremental imbalance reserve require- ments (per unit) for each of the six scenarios analyzed. In the im- quirements for the following and imbalance distributions for balance case in Fig. be lessened. (load minus renewable generation) is less than forecasted more Improvement in forecasting accuracy alone can greatly help the often than not. wave. the imbalance reserve require- The final data column proves to be the most illuminating.HALAMAY et al. confirmation of the supposition that the various characteristics For the equal mix scenario. imbalance reserve requirements. 8. Diversification. ments (per unit) for each of the six scenarios analyzed. as the renewable resource contribu- as expected given the equal mix of solar and wave penetration. which presents the load minus the label “dec” refers to the decremental reserve requirement. espe. cases the skewness is negative. the addition of variable. or under-forecast the renewable generation. 8. severe as more and more wind generation is installed. A summary of the reserve requirements for the six analyzed the reserve requirements on both the following and imbalance scenarios is presented in Figs. 9. the situation is only expected to grow more of the diverse renewables did not cause a disproportionate in. The incremental and decremental reserve re. In all on a scale as large as the current level of wind penetration. implying that the addition In the near future. the difference between the var- Similar results are demonstrated in the fifth data column. This di- verse renewable resource portfolio provides the largest benefit V. adding a second large-scale renewable resource to the portfolio. able generation sources such as tidal power to the mix.5% all of the scenarios with a combination of renewable resources wave scenario. In general. and solar generation will allow a greater com- is very close to those of the load alone. By diver- crease in the number of events requiring larger reserves. 10% wind and 5% wave scenario. the differences become much this scenario fall between those for the previous two scenarios. In this scenario. as a study to determine the optimal combination of the var- Also. this is an important difference from the cur. the increased diversity renewable power source. CONCLUSION to the reserve requirements for both the following and imbal- ance time scales compared to any of the other scenarios. ment tends to decrease. of renewables is most helpful in reducing both the MAE and the solar. the time scales are reduced. resources to implement wave and solar generation technologies Another interesting observation is in the skewness. the current portfolio of renewable resources in the Pacific alone. For the following case in Fig. The include the investigation of the impact of adding other renew- practical significance is that utility operators will need decre. an equal mix of 5% wind. renewable generation tends to ious renewable resources to improve reserve requirements even decrease the skewness. however. unlike the load minus 15% wind scenario where ments are reduced compared to those for wind alone. Unfortu- the RMSE increased more than the MAE compared to the load nately. however. tion becomes more diversified. and 2. Comparison of incremental and decremental following reserve require. and wave power generation. further. the RMSE and MAE increased approximately the same Northwest is almost exclusively composed of wind generation. 2. Possibilities for future study cast the load. Fig. This could be caused by a tendency to over-fore. label “inc” refers to the incremental reserve requirement and sults in the next data column. the variance of both distributions of wind. 5% solar. bined penetration rate than using only one predominate type of balance reserve requirement distribution. From a sifying the portfolio. By utilizing an equal mix of wind. This result is also borne out by the re. In fact. require fewer reserves than that with wind alone. as well mental reserve available more often than incremental reserve. Particularly for the im.5% solar. The results of the analyses in this study lend evidence to the cially the current portfolio of wind alone. ious scenarios is relatively small. and 5% wave penetration is subtracted from the load.

“Characteristics of the UK wind resource: long-term pat- [1] Northwest’s wind power could quadruple [Online].” IEEE Trans. 2.bpa. Ted K. 2010. D. pp. A. (summa cum laude) with an emphasis in power. Soder. and Y. Available: http:// power electronics and electric drives.S. T. D. 28–33. 1–6. Winters. no.” IEEE Power Energy Mag. Power and Energy Society graduate student at Oregon State University fo- General Meeting. Oct. 22–31. Milligan. [11] 2010 BPA rate case wholesale power rate final proposal genera. L’Energia E L’Am- thors would also like to thank Dr. 23. pp.uoregon. K. “Large-scale integration of optimal combinations of PV. Ente Per Le Nuove Tecnologie. Musial. NO. 2009. F.pdf&contentType=applica- Storck. Agilent Technologies. VOL. Jun. 16–21. His research interests include control. Mills. vol.” in Proc. J.” Energy Policy. Stage Gate 2 no. [16] D. 112–127. Apr. REFERENCES [23] G. Tech. “Wind-solar resource complementarity and its combined correlation with electricity load tion code. Lew. D. power electronics. Brekken. T. having worked for VLSI Technologies. May 2009. The au. Ackermann. “Change in the air.” IEEE Power Energy Mag. E. J. 2009. Brekken (M’06) is an Assistant Professor 7./Dec.S.. vol. He has variability of renewable energy sources by reserve requirement calcu.D. no. 65–75. Simmons. W. Remote Sensing. 6. Tech. Groome. 2010. 2010.com/5tds3s no. Musial. pp. “Comparison of capacity credit calculation methods for Douglas A.edu/index. 2007. Bermejo of the BPA for their technical assistance. Jun. 2006. Kehler. Shaun McArthur (S’08–M’10) is a graduate student currently pursuing a Ph.” IEEE Power Energy Mag. vol. and renewable energy device simulation General Meeting.. da Fonseca Manso. After working for three years as sources. 31. Previsic. V. Jul. 2010. no. WA. K. 1. “Re- serve requirement impacts of large-scale integration of wind. Bedard. solar. F. Holttinen. Halamay and T.bpa. 2009 [Online]. de la Torre.” in Proc. and Marvell [17] Total load & wind generation in the BPA control area (Balancing Semiconductor. returned to school after a 20 year career as an ASIC lation. T. pp.” IEEE Trans. [2] M. K. 1999. “Capacity credit for renewable energy re. Eriksen. Power Engineering Society 1999 Winter Meeting. and [21] Solar Thermal Energy Program Production: Guidelines and Future Pro- J.E. 685–691.328 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY. tinyurl. M. 2009 [On- line]. Trondheim.E. [4] R.pdf also studied wind turbine control at the Norwegian [12] D. Lange. Power Syst. 3.. no. “Wind and wave power into the electricity supply. Available: http://www. M. Pudaruth and F. smart grids. South Korea in 1999.transmission. 2. vol.bpa. C. and Ph. in applied [19] S. M. B. IEEE Energy Conversion Congress and Expo (ECCE). Integrated Defense Systems in Renton.aspx? [5] W. 27. Lozovoy. Hewlett-Packard.gov/Business/Opera. G. no. Edelson. He received his B. L. 2001. J. A. Sales.html The authors would like to thank B. 24. 24. May 2009. and J. Li. He received his B. with renewable sources. 1072–1079.nrel. and C. M. Mohrlen. [Online]. G. J. 2. Lund. Agelidis. P. Final Report EPRI-WP-012.bpa. Power Syst. vol. 2009. scholarship. “Where the wind blows. 503–515.” Oceanography. Feb. 1. A. 23. Nov. Ozkan-Haller of Oregon biente. Varley.. DeMeo. Nov. Denholm.xls in energy systems at Oregon State University. of the Wallace Energy Systems and Renewables Facility (WESRF). “Up with wind. A. “A methodology for quantifying from the University of Arizona in 1989. 2.xls [18] Wind generation capacity in the BPA balancing area authority Bonneville Power Administration. and 2005 respectively. plication of satellite-derived solar resource data. Ross. Topics Appl. [15] D. fileName=WI-09-E-BPA-02_Testimony.gov/corporate/rate- sign at Postech in Pohang. B. pp. and L. “Locational capacity credit evaluation. Schuerger. wind N. Parks. Billinton. Tech. pp. He is codirector [14] A. in energy systems at Oregon State University.gov/public/WWIS/ digital control techniques applied to renewable energy systems. 3623–3628. pp. University of Science and Technology (NTNU) in and B. Nov. B. Bedard. 106–116. “Ocean renewable energy’s potential role Forecasting for Grid Connected Wave Energy Projects in Washington in supplying future electrical energy needs. no. pp. no.. Jordan. vol.” Oceanography. A. Tech.. Ancell. 2010. Available: http:// terns and relationship to electricity demand. Zack. 2. vol. MN. from the tion inputs study Bonneville Power Administration. Koehler. 7. he re- [9] R. University of Minnesota in 1999. He case/2008/2010_BPA_Rate_Case/docs/WP-10-FS-BPA-08_Web. specifically wind. 4. M... . Hoff. Available: http://www.S. power myths debunked.S and Foreign patents. no.” in Proc. He received his B. Taylor..” IEEE Trans. 25. pp. L. 4th IEEE Conf. Perez. He studied electric vehicle motor de- Jul. “An [26] Reserve capacity forecast for wind generation within-hour balancing overview of ocean renewable energy technologies. 2009 (ICIEA 2009). Thresher and W. [7] M. vol. Feb. O’Sullivan. G. Ferreira and P. pp. Jun.. Available: http://www. 89–99. “Ocean wave power data genera. Rep.” Renewable Energy. Rep. a systems engineer on the P-8A Poseidon at Boeing IEEE. designer. [13] Western wind and solar integration study [Online]. Van Hull. P.. grammes of ENEA. Jan. Borup. K. pp. Sel. Power and Energy Society trol. R. Rep. 47–58. 6. McArthur. Zavadil. service Bonneville Power Administration. cusing on energy systems. Norway in 2004–2005 on a Fulbright 6.E.gov/ratecase/openfile. Kirby.. J. P.” IEEE tion%2fpdf Power Energy Mag. [10] T. Available: http://solardat.E. Carvalho. 1. 7. Miller. Industrial Electronics and Applica- tions. [Online]. and R.transmission. Grant. Brekken. no. Halamay. He is an inventor of record on over Authority Area) for 2008 Bonneville Power Administration. vol. 31. Atlanta. 7. 25. Shrivastava. 35. [25] R. vol. Feb. 2008 50 U. 2008 vol. Finley. Mar. and Asher Simmons (A’09–M’10) is a nontraditional ocean wave power generation. His research interests include con- tion for grid integration studies. M. uate student in energy systems at Oregon State Uni- vol. 2002. and Oregon Electric Power Research Institute (EPRI).S. A. da Silva. pp. Ahlstrom.S. demand. J. Dumas. 2005 with his B. vol. 2010.. P. 1. O’Malley. and R. A. Power Syst. W. Porter. Ernst. McArthur and T. Oct. Lerner. and M. 2008. vol. Earth Observ. tions/Wind/WIND_InstalledCapacity_current. pp. 1. Li. 2009. [22] Y.” in Proc. Brekken. Orths. Corbus. He graduated from Gonzaga University in [8] L. and S. [24] H. H. and a “Long-term probabilistic evaluation of operating reserve requirements recipient of the NSF CAREER award. G. M. no. Manobianco. pp.” in Proc. JULY 2011 ACKNOWLEDGMENT [20] University of Oregon solar radiation monitoring laboratory [Online]. GA. pp. May 2009. Sinden.gov/Business/Oper- ations/Wind/TotalWindLoad%_5Min_08./Dec. Rep. W. State University for her assistance with the wave data genera. 2008. and optimization. K. physics from the University of Utah in 2006. Jacobson. Amelin. “Reaching consensus in the turned to graduate school to focus on power systems definition of photovoltaics capacity credit in the USA: A practical ap- and renewable energy integration. D. versity. 36–46. Minneapolis.D.” IEEE J. T. [6] G. 471–475. Feasibility of Using Wavewatch III For Days-Ahead Output [3] R. Halamay (S’02–M’05–S’08) is a grad- conventional power plants and wind power. no. Available: https://secure..

Master your semester with Scribd & The New York Times

Special offer for students: Only $4.99/month.

Master your semester with Scribd & The New York Times

Cancel anytime.