Analyzing the Impacts of Variable Renewable Resources on

NREL is a national laboratory of the U.S. Department of Energy

Office of Energy Efficiency & Renewable Energy

Operated by the Alliance for Sustainable Energy, LLC

This report is available at no cost from the National Renewable Energy

Laboratory (NREL) at www.nrel.gov/publications.

Contract No. DE-AC36-08GO28308

Analyzing the Impacts of

Variable Renewable Resources

on California Net

-Load

Ramp Events

Preprint

Bing Huang,

Venkat Krishnan,

and

Bri-Mathias Hodge

National Renewable Energy Laboratory

resented at the IEEE Power and Energy Society General Meeting

Portland, Oregon

August 5

–10, 2018

IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all

other uses, in any current or future media, including reprinting/republishing this material for advertising

or promotional purposes, creating new collective works, for resale or redistribution to servers or lists,

or reuse of any copyrighted component of this work in other works.

Conference Paper

NREL/CP-5D00-70486

August 2018

NOTICE

This work was authored by the National Renewable Energy Laboratory, operated by Alliance for Sustainable

Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. Funding

provided by U.S. Department of Energy Office of Energy Efficiency and Renewable Energy Wind Energy

Technologies Office (Grid Modernization Laboratory Consortium Category 2). The views expressed in the article do

not necessarily represent the views of the DOE or the U.S. Government. The U.S. Government retains and the

publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive,

paid-up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to

do so, for U.S. Government purposes.

This report is available at no cost from the National Renewable

Energy Laboratory (NREL) at www.nrel.gov/publications

U.S. Department of Energy (DOE) reports produced after 1991

and a growing number of pre-1991 documents are available

free via www.OSTI.gov

Cover Photos by Dennis Schroeder: (left to right) NREL 26173, NREL 18302, NREL 19758, NREL 29642, NREL 19795.

NREL prints on paper that contains recycled content.

This report is available at no cost from the National Renewable Energy Laboratory (NREL) at www.nrel.gov/publications.

Analyzing the Impacts of Variable Renewable

Resources on California Net-Load Ramp Events

Bing Huang, IEEE St. Member, Venkat Krishnan, IEEE Member, and Bri-Mathias Hodge, IEEE Senior Member

Abstract—This paper characterizes the ramping features of

multiple renewable resources, load and net-load in California

using 2016 measured data. The goal of this analysis is to

understand the interactions between variable renewable resource

ramps and net-load ramps, and delineate the positive and negative

impacts. The findings indicated that although the frequency and

uncertainty in the system level net-load ramp occurrences

increased due to variable renewable integration, there was also a

decrease in the average ramp up and down magnitudes by about

20% and 30% respectively, along with decreases in their

respective standard deviations by 40% and 20%. At the zonal

level, the analyses also indicated a 4-5% decrease in the net-load

standard deviations due to wind integration alone at various

market scheduling periods. The paper finally concludes with

future research needs to better integrate variable renewable ramp

forecasts into system operations and planning for economic

utilization of resources.

Keywords—Ramping, Renewables, Net-load Ramps, Flexible

Ramping Product, CAISO, Ramp Forecasts, Reserves

I. INTRODUCTION

ramp event can be defined as large or rapid changes in

power and it can appear at both generation as well as for

load in the power system. Such variations in either

direction (i.e., up ramp when there is an increase and down

ramp when there is a decrease) must be compensated in real

time using load following resources to keep the generation-load

balance and the consequent system frequency response in order.

In situations when serious ramp events cause generation-load

imbalances, especially of the generation scarcity types with

insufficient ramp capability, power markets send out high real-

time prices to quickly mitigate the scarcity situation by tapping

into reserves or unplanned quick start generation [1]. With

increases in renewable generation penetration such as wind and

solar, that induce higher amount of variability and uncertainty

on the system net-load (i.e., total load minus renewable

generation), it becomes challenging for the system operators to

maintain the generation-load balance efficiently. To address

this issue, market-based flexible ramping products (FRPs) have

been proposed in the industry, which embeds foresight about

the anticipated net-load ramps into the Independent System

Operator’s (ISO) market clearing procedures that co-optimize

energy, reserves, and ramping. This in turn allows ISOs to

utilize the existing ramping capabilities better by timely

procurement of ramping resources in the current dispatch

Bing Huang (binghuang@utexas.edu) is a graduate student at the University of

Texas, Austin, and was an intern at National Renewable Energy Laboratory

(NREL) during 2017 summer. Venkat Krishnan, Senior Engineer

(

Venkat.kri[email protected]) and Bri-Mathias Hodge, Manager, Power System

intervals for future needs [1]. In 2016, California ISO (CAISO)

implemented FRP in its fifteen-minute real-time market (FMM)

[2] and the Midcontinent ISO (MISO) in both its day-ahead and

5-min real-time markets [3].

Under the current design, FRPs are provided by fast

responding peaking units and fast start units, typically gas-fired.

Cui et al. [4] and Chen et al. [5] studied the feasibility and

impacts of wind resources offering ramping capacity into the

market clearing processes. In order for this concept to mature,

it is important to understand the ramp events in the renewable

generation and their interactions with the system net-load ramps

better. Sevlian and Rajagopal et. al. [6] proposed an optimal

ramp detection technique and analyzed the distribution of

different wind power ramp features using data from the

Bonneville Power Authority (BPA) and NREL Western

Integration Dataset. Kamath [7] showed the effectiveness of a

simple statistical analysis on the characterizations of wind

generation ramp events, including distributions of severity

levels, start time within the day and occurrence by month from

a BPA dataset. Based on systematic studies on wind power

changes at different time steps, Wan [8] studied the statistics of

wind power ramp events from ERCOT data and inspected their

impacts on the system net-load ramps. However, all of these

works only focused on detection and characterization of wind

ramp events, and there is a dearth of works that can relate the

impact of variable renewable resource ramps on the system net-

load ramps and understand the changes in their summary

statistics.

Thus, the goal of this paper is to analyze the ramp events

in the California system’s load and net-load data, and quantify

the impact of variable renewable ramps on net-load ramps from

both negative and positive perspectives. For this purpose, the

optimized swinging door algorithm (OpSDA) developed in [9]

was used in this study to detect and assess the ramp events and

their features such as start times, magnitudes, duration and

rates. Based on the ramp characterizations shown in Sections II

(load and net-load) and III (renewable resources), the

possibilities of future power market innovations for variable

renewable ramp forecast integration are discussed in Section

IV. Finally, Section V presents conclusions.

Design and Studies Group (bri.mathias.[email protected]), are with the Power

Systems Engineering Center at NREL.

This report is available at no cost from the National Renewable Energy Laboratory (NREL) at www.nrel.gov/publications.

II. L

OAD AND NET-LOAD RAMPS CHARACTERIZATION

The 2016 California system measured data, at 1 min

resolution [10], was used to analyze the ramp up and ramp

down events for load and net-load. The net-load is defined as

system load minus renewable generation, including distributed

generation, solar PVs, solar thermal, and wind power in

California. Table I shows the number of ramp events detected

in the load and net-load data (using OpSDA [9], with a

qualifying ramp event defined as 5% of the total load), where

we observe an increase in the net-load ramp events influenced

by the addition of renewable generation. Fig. 1 shows the

distributions of ramp event start times across the 24-hour

period, where we observe that the ramp event occurrences are

relatively less predictable in the net-load data (i.e., occurring

throughout the day) compared to the load data which had ramps

mostly happening during morning and evening hours.

TABLE I NUMBER OF RAMP EVENTS IN 2016 CAISO LOAD & NETLOAD DATA

Load

Net Load

Ramp Up

313

731

Ramp Down

245

622

Fig. 1 Daily Distributions of CAISO Ramp Up and Down Event Start Times

Fig. 2 Distributions of CAISO Ramp Up and Down Event Magnitudes

However, it is also interesting to note from Fig. 1 that there

are less ramp events in the net-load data during early morning

(5-7 am) and evening peak hours (16-18 pm). Upon analyzing

ramp magnitudes, Fig. 2 shows that the probability of certain

magnitudes (around 8-14 GW, 18-20 GW) decreases in the net-

load ramp data compared to the load data. The summary

statistics in Table II for ramp magnitudes and rates further

corroborate this finding, where we can observe that the ramps

in the net-load have lower maximum and average magnitudes

and rates, along with a smaller standard deviations compared to

the load ramps. This indicates that the ramp magnitude and rates

in net-load is distributed in lower values with a smaller range.

Therefore, though the integration of renewables has increased

the number of ramp events and their uncertainties, there are

times when they seem to alleviate the net-load ramps in either

direction, thereby implicitly providing ramping capability to the

grid for reliable operation.

TABLE II 2016 CAISO LOAD AND NET-LOAD RAMP STATISTICS

Load (w/o renewables)

Net load (% change)

Down

Ramp magnitude (MW)

Max

21,704

21,058

18,400

16,594

Average

5,406

8,075

4,317(-

20%)

5,694(-

30%)

Std. Dev.

5,077

4,227

2,894(-

43%)

3,349(-

20%)

Ramp rate (MW/min)

Max

3,540

3,692

3,537

3,134

Average

104

118

80(-28%)

69(-41%)

Std. Dev.

370

439

252(-32%)

264(-40%)

III. RENEWABLE GENERATION RAMPS CHARACTERIZATION

igure 3 shows the ramp up (left) and down (right) events

start time distributions for different renewable generation, that

include distributed generation (DG), PV Solar (PV), Thermal

Solar (TS) and Wind. One can draw the following inferences

about the aggregated impact of solar-dependent resources on

net-load ramps from Figs 1 and 3:

• Negative impacts: The worsening of ramp up

requirements in the net-load data during the evening 15-

19 hours in Fig. 1 may be primarily due to solar-

dependent resources as they tend to ramp down during

those hours as seen in Fig. 3 (right).

• Positive impacts: However, reduction in the ramp up a

wn magnitudes in the net-load data compared to loa

ata (Fig. 1) during morning and evening hours

respectively may be primarily attributed to the solar-

based variable renewable resources, which ramp up in

the morning and ramp down in the afternoon (Fig. 3).

Wind resources, however, are different from the rest of the

renewable resources, in that the majority of ramp up start times

is in the evening, and ramp down start times is in the early

morning. The following inferences can be drawn about their

aggregated impact on net-load ramping events:

• Negative impacts: Comparing Figs. 1 and 3, during

morning hours when load is expected to ramp u

aggregate wind resources ramping down may have an

acerbating effect. Additionally, the spread out nature of

wind ramping events in Fig. 3 throughout the day may

This report is available at no cost from the National Renewable Energy Laboratory (NREL) at www.nrel.gov/publications.

contribute to the similar observation in the net-load ramp

events and increase uncertainties

• Positive impacts: Comparing Figs. 1 and 3, load and wind

ramp up events coincide to some extent during evening

hours, when the system could possibly face ramping

capability limitations.

Fig. 3 Distributions of Renewable Ramp Start Times: Ramp up (left) and Ramp down (right)

To further confirm the above observations, investigations

at the zonal level were performed. In this analysis, load ramp

features were compared with wind affected net-load ramp

features (i.e., Net-load

or Netload (wind) defined as load

minus wind) at three different reserve zones in California,

pertaining to the Pacific Gas & Electric (PG&E) bay area (with

~3% wind by capacity), the Sacramento Municipal Utility

District (SMUD) (with no wind) and the San Diego Gas &

Electric (SDGE) (with ~9% wind by capacity) utility regions.

Based on the work in [11], yearly load and renewable resource

data at 1 min-resolution for the three CAISO zones were used

to estimate the load and wind resource distribution factors

among the three zones, which was applied to the 2016

California system time series data [10]. To study the impact of

wind generation on net-load ramps, three different maximum

instantaneous wind energy penetration levels (estimated as the

largest wind to load ratio in the time series data) were chosen,

i.e., 15% (2016 CAISO system data), 30%, and 50%.

Fig. 4 California Zone 3 Ramp Start Time Distribution of Load and Net-

load(wind) at Different Wind Penetration Levels

Figure 4 shows the ramp start time distribution of load and

net-load

at different wind penetration levels in zone 3. When

the penetration level increases from 0% indicated by blue lines

(load data) to 50% indicated by green lines, we observe that

although wind resources do spread the ramp events across the

day (8-15 hours), they do contribute to the decrease in the

frequencies of ramping events, especially during early morning

(4-7) and peak evening hours (16-18) for ramp up events, and

late evening/night hours (20-23) for ramp down events. Table

III shows the statistics of ramp up and ramp down magnitudes

for load and net-load

at different wind penetrations in zone 3.

Though the number of net load ramps does increase with the

increase in wind penetration level (doubles at 30% wind and

almost triples at 50% wind), the values of ramp magnitude

statistics including the average and standard deviations

decrease (decreases by 40% and 50% for ramp up, and 28% and

33% for ramp down under 30% and 50% wind penetrations

respectively). These are further evidences that wind energy in

zone 3 is indeed helping the system by reducing ramps,

including during peak hours.

ABLE III ZONE 3 RAMP MAGNITUDE STATISTICS AT DIFFERENT WIND

PENETRATIONS

(% CHANGE FROM LOAD ALSO SHOWN)

Load

(MW)

Netload

15% (MW)

Netload

30%

(MW)

Netload

50% (MW)

Ramp Up

Average

1,514.2

1,283.7(-15%)

1,204.7(-20%)

1,194.6(-

21%)

Std. Dev.

1,062.3

770.8 (-27%)

613.6 (-42%)

520.6 (-51%)

Number

731

1,018

1,423

1,966

Ramp Down

Average

1,271.4

1,209.5 (-5%)

1,172.8 (-8%)

1,179.9 (-7%)

Std. Dev.

743.7

656.9 (-12%)

531.7 (-28%)

495.8 (-33%)

Number

777

1,001

1,433

1,963

IV. DISCUSSION ON FUTURE NEEDS: BETTER INTEGRATION

VARIABLE RENEWABLE RAMP FORECASTS

The analyses in Sections II and III indicate a need for better

integration of variable renewable ramp forecast information

into the market operations and planning tools. Currently,

average forecasts of variable renewable generation at different

time intervals suitable for different market clearing processes

are accounted, i.e., day-ahead forecasts for day-ahead market, a

few hours-ahead forecasts for look-ahead reliability unit

commitments, and intra-hour ahead (15-min and 5-min)

This report is available at no cost from the National Renewable Energy Laboratory (NREL) at www.nrel.gov/publications.

forecasts for real-time markets. Though average forecasts of

renewable generation will provide an estimation of the expected

ramps, nevertheless, markets have to explicitly take into

account the accurate ramp forecasts of renewables in order to

1) better cope with their anticipated negative influences on the

net-load, and 2) be in a position to take advantage of their

positive impacts on net-load ramps that could alleviate the

reserve needs. In this context, we envision at least two kinds of

integrations of renewable ramp forecasts.

A. Better FRP procurements using net-load ramp forecasts

ISOs that have implemented FRPs estimate the ramping

capability procurements for their day-ahead and real-time

markets based on day-ahead and real-time net-load forecasts,

respectively. In addition to the net-load ramp forecast, the

ramping requirements are also informed by ramp uncertainties

(typically a function of standard deviations in historical net-

load ramp data). Therefore, any improvement in the renewable

generation ramp forecasts will directly impact the net-load

ramp forecasts and consequently the FRP procurement in the

markets. This can be observed from the standard deviations of

the 5-min, 15-min, and 1-hour net-load ramps shown in Table

IV. The standard deviations at various time-intervals decrease

in the net-load

data under 50% wind penetration compared to

15% wind penetration data (~1% reduction in zone 1, ~4-5%

reduction in zone 3 and ~2% reduction in system (zone 2 has

no wind)), which will reduce the ramp procurement needs in the

respective markets and consequently reduce the reserve costs

from conventional generation and the system production costs

[4]. Therefore, precise information of renewable ramp forecasts

and their uncertainties through advanced probabilistic forecasts

[12], and the integration of such information to extract accurate

net-load ramp forecasts will likely benefit system economics.

TABLE IV NET-LOAD

RAMP STANDARD DEVIATION: 15% VS. 50% WIND

Region

5-min

15-min

1-hour

CAIS

141.70 vs.

138.99

213.81 vs. 209.96

392.30 vs.

386.84

Zone 1

102.19 vs.

101.22

126.86 vs. 125.64

200.16 vs.

198.49

Zone 2

66.63 vs. 66.63

75.73 vs. 75.73

107.31 vs.

107.31

Zone 3

77.49 vs. 74.39

92.70 vs. 87.99

135.98 vs.

129.07

B. Plant-level ramp forecasts for situational awareness and

ramp products from variable renewable generation

The previous discussion was from the system perspective,

where aggregated wind and net-load ramp forecasts are to be

used. Under increasing levels of variable renewable

penetrations, plant-level ramp forecasts will also be important.

Already, the Electric Reliability Council of Texas (ERCOT) has

implemented programs to gain visibility into renewable

resources in their energy management system (EMS) that will

inform the operators about impending wind ramping events

along with their probabilities [13]. This feature provides

necessary situational awareness of variable renewables for the

operators to take timely corrective actions including calling on

unplanned quick start units to offset any ramp up deficits.

However, as discussed before, there are scenarios when certain

renewable resource plants also provide positive impacts on

system net-load ramps. Therefore, future EMS applications will

benefit from having precise plant-level ramp forecast

information, along with associated uncertainties in the form of

probabilistic forecasts [14], which will enable system operators

to gain accurate foresight into such situations. Such abilities

could also enable operators in the real-time EMS environment

to use variable renewable ramps better, including strategically

curtailing certain plants in order to offset down-ramp deficits

(instead of cycling a conventional plant), and then compensate

the respective renewable plant with lost opportunity cost or FRP

clearing prices.

Additionally, integrating plant-level ramp forecast

information in the planning and market dispatch processes

could enable variable renewables to offer their services for the

flexible ramping product explicitly, unlike the conventional

idea of implicitly accounting for their impacts on net-load

ramping requirements (as discussed in Section IV.A). Figure 5

shows probability distributions of wind ramp start times,

magnitudes and duration based on 2013 simulated wind power

data at 5-min resolution from NREL’s WIND toolkit [15] over

24 hrs. The selected wind site is from a city named Felicity, on

the border between California and Mexico, with a site ID 14960

in the WIND toolkit, a capacity of 16 MW, a wind speed of

6.36m/s, and a capacity factor of 0.33. We can observe that the

peaks of the distribution curve of ramp up events are in the early

morning (around 4 am to 8 am) while the peaks of the

distribution curve of ramp down events are in the afternoon (13-

15 pm), except for in the Summer. The ramp up events from

this plant in the morning are beneficial to the system since the

system load also ramps up at the same time, as seen in Fig. 1.

While looking at the aggregated wind ramp up start times in

Fig. 3, wind resources contributions to ramping up events in the

morning hours seemed trivial, but in looking at this plant-level

information, one can see opportunities for this plant to offer its

ramping capacity explicitly into the FRP. Fig. 5 shows that this

wind plant can provide a ramping capability of 4-6 MW (~32%

of 16 MW plant) with high probability, and that some of these

ramps could also last for hours, thereby creating opportunity for

this and any other wind plants with correlated outputs to offer

their ramping capability explicitly for FRP services, thereby

having a chance to gain additional revenues from FRP marginal

clearing prices. Apart from aiding system ramps when a

particular wind plant’s ramp is forecasted to be in sync with the

load ramps, wind generation can also consider controlling their

outputs or ramp rates under situations when a wind resource’s

ramp is forecasted to exacerbate the system ramping needs (i.e.,

wind ramps in opposite direction to net-load). One such

example is when wind resources can consider strategically

This report is available at no cost from the National Renewable Energy Laboratory (NREL) at www.nrel.gov/publications.

curtailing the wind power and thereby provide down-ramp

under situations when system down-ramp capability is limited,

under situation when it is economically more attractive than

providing power to the energy market.

Therefore, having accurate forecasts of plant-level ramps

and their uncertainties will open the door for better

understanding of the synergies between variable renewable

generation and load ramps, and even utilizing the renewable

ramps for grid services. Consequently, ramping provisions by

individual renewable plants will reduce the FRP needs from

conventional units and reduce total system production costs.

Fig. 5 Distributions of Wind Site Ramp: 1) Start Time, 2) Magnitude, and 3) Duration

V. CONCLUSIONS

This paper presented an analysis of the impacts of

variable renewable generation ramps on the system net-load

ramps, taking California 2016 measured data as an example

and extracting ramp features using the optimized swinging

door algorithm. The analysis showed that integration of

variable renewables definitely increased the overall annual

frequencies of ramp events and uncertainties related to their

occurrences throughout the day. Nevertheless, during certain

ramping challenged situations like early morning or evening

hours, renewable generation ramps seemed to implicitly

provide support to net-load ramps and decrease the

magnitudes and ramp rates of net load. The summary

statistics showed the average of system aggregate net-load

ramps to decrease by 20% in magnitude for ramp ups and

30% for ramp downs, with their standard deviations

decreasing by 40% and 20% respectively.

Upon analyzing the solar and wind resources separately,

it was observed that solar ramps had positive influences on

overall ne-load ramps during morning and early evening

hours when their trend coincided with diurnal load, except

during late evening hours when net-load ramp up was

exacerbated due to solar ramp down. On the other hand, wind

[1] N. Navid and G. Rosenwald, "Market Solutions for Managing Ramp

Flexibility With High Penetration of Renewable Resource," in IEEE

Transactions on Sustainable Energy, vol. 3, no. 4, pp. 784-790, Oct. 2012.

[2] California ISO, “Flexible Ramping Product- Revised Draft Final

Proposal,” Dec 2015

[3] N. Navid and G. Rosenwald, “Ramp Capability Product Design for MISO

Markets,” Market Development and Analysis, July 2013

[4] M. Cui, J. Zhang and B.-M. Hodge, “Wind-Friendly Flexible Ramping

Product Design in Multi-Timescale Power System Operations,” IEEE Trans.

Sustain. Energy, vol. 8, no. 3, pp. 1064–1075, Jan. 2017.

resources seemed to contribute to the decrease in the

frequencies of ramping events, especially during early

morning and late evening hours for ramp up events, and late

evening/night hours for ramp down events, thereby providing

implicit support to the grid when solar ramped down. A

further zonal level investigation of wind ramp impacts on

California net-load showed that in zone 3 (depicting SDGE

with higher wind capacity) wind ramps could decrease ramp

standard deviations by up to 4-5% at 5-min, 15-min, and 1-

hour intervals, which would mean the required ramping

reserve procurements will be reduced.

All of these analyses indicated a need for better

integration of variable renewable ramp forecasts information,

including their uncertainties via probabilistic forecasts, into

the market operations and planning tools, both at the

individual plant level, as well as aggregated regional or

system level. This will enable better utilization of existing

ramping capabilities in the grid, including from the variable

renewable resources, reducing unplanned quick starts,

reducing system production costs, and finally achieving more

efficient markets.

REFERENCES

[5] R. Chen et. al., “Wind Power Providing Flexible Ramp Product,” IEEE

Trans. Power Syst., vol. 32, no. 3, pp. 2049-2061, May 2013.

[6] R. Sevlian and R. Rajagopal, “Detection and statistics of wind power

ramps,” IEEE Trans. Power Syst., vol. 28, no. 4, pp. 3610–3620, 2013.

[7] C. Kamath, “Understanding wind ramp events through analysis of

historical data,” in Proc., Transmission Distribution Conf. Expo., New

Orleans, LA, USA, pp. 1–6, 2010

[8] Y. Wan, “Analysis of Wind Power Ramping Behavior in ERCOT,”

NREL, Golden, CO. Mar 2011.

This report is available at no cost from the National Renewable Energy Laboratory (NREL) at www.nrel.gov/publications.

[9] M. Cui et. al., “An optimized swinging door algorithm for identifying

wind ramping events,” IEEE Trans. Sustain. Energy, vol. 7, no. 1, pp. 150–

162, Jan. 2016.

[10] California ISO OASIS [Online]. Available FTP:

http://oasis.caiso.com/mrioasis/logon.do

[11] I. Pena; C. Brancucci; B. M. Hodge, "An Extended IEEE 118-bus Test

System with High Renewable Penetration," IEEE Transactions on Power

Systems , vol.PP, no.99, pp.1-1, April 2017

[12] M. Cui et. al., "Wind Power Ramp Event Forecasting Using a Stochastic

Scenario Generation Method," IEEE Transactions on Sustainable Energy,

vol. 6, no. 2, pp. 422-433, April 2015

[13] ERCOT operating procedure manual, Power Operations Bulletin # 794,

http://www.ercot.com/mktrules/guides/procedures/index.html

(also seen in

the news release: http://www.ercot.com/news/releases/show/326)

[14] M. Cui et. al., “A Data-Driven Methodology for Probabilistic Wind

Power Ramp Forecasting,” to appear in IEEE Trans. Smart Grid, Oct 2017

[15] Draxl, C et. al., "The Wind Integration National Dataset (WIND)

Toolkit." Applied Energy 151: 355366, 2015