1. CRITICAL TIMEFRAMES OF IMPORTANCE FOR PV FROM A UTILITY PERSPECTIVE
Obadiah Bartholomy
Yong Cai Pramod Krishnani
Belectric Inc.
Brad Dommer
8076 Central Avenue
Sacramento Municipal Utility District
Newark CA 94560
6201 S St
Sacramento, CA 95817 Email: krishnani.pramod@gmail.com
Email: Obadiah.Bartholomy@smud.org pramod.krishnani@belectric-usa.com
Yong.Cai@smud.org
Brad.Dommer@smud.org
ABSTRACT loads occur during approximately 15% of the 4,200 hours
that PV produces energy during year in Sacramento,
As solar penetrations increase, several resource attributes primarily in the springtime with a second but lower
and timeframes are becoming of increasing importance to concentration during the fall. Within an hour, a few dozen
utility and balancing authority operators. In particular, days a year appear to have high variability, potentially
output and variability of PV systems during summer peak creating issues for distribution system operators with
demand periods, during daytime minimum load changes exceeding 50% of rated capacity of a multi-MW
conditions in the spring and fall, and finally at very high system in a 1-minute period. Finally, over the longer term,
penetrations, during periods of significant load ramps for periods of significant ramps of PV in the morning and
the utility. Using historical solar resource data and load evening are likely to present challenges for existing grid
data, PV output is simulated for different configurations balancing resources during non-summer months.
and capacity levels to evaluate its potential to impact grid
operations during these types of timeframes. In this 1. INTRODUCTION
study, historical SMUD hourly load data is correlated to
validated modeled PV output to determine appropriate on- Anecdotal assessment of PV output and problems can
peak capacity values by orientation, minimum load – tend to receive much of the focus and may leave long-
maximum PV output correlations and frequency, and lasting impressions within a utility. Single events that
diurnal patterns of change on an hourly basis. cause problems for a utility may reinforce perceptions of
the resource as unreliable and problematic. For most
The assessment uses 9 years of SolarAnywhere(c) 10 km utilities, high penetrations of solar are relatively new and
grid hourly data correlated to SMUD system load data. up until recently, high resolution historical datasets were
More recent sub-hourly assessments are done for these not readily available to begin assessing the frequency that
values using recently installed ground-based solar problematic conditions may occur. This paper will begin
monitoring to evaluate sub-hourly variability that may to use these relatively recently available datasets to
significantly impact grid operations during low-load time answer questions related to timeframes that are important
periods of interest. to utility operators during which PV performance is of
heightened importance.
The results show that for SMUD’s service territory, peak
contribution from solar PV can be consistently expected The timeframe of primary importance to utility system
to contribute between 35% and 60% of its peak rating operators generally is during the hot summer months for
depending on orientation during the typical peak hour summer peaking utilities, and in particular, for SMUD,
ending 6PM. Periods of high PV output and low utility during hot afternoons between 2PM and 8PM. PV
1

2. performance during this timeframe varies primarily due to provides 1 km and half hour time resolution, however for
the steady pattern of a declining solar resource as the sun the purposes of this analysis those data were not used.
sets to the West. Late afternoon clouds are possible, but
generally also reduce peak demand, or occur on days that In addition to the SolarAnywhere© data, SMUD, working
are not likely to reach critical peaks. with NEO Virtus Engineering Inc. under a grant from the
CPUC California Solar Initiative RD&D program have
The second timeframe that has been considered to be of deployed 71 solar monitoring devices around the service
importance for operations have been days where utility territory providing 1 minute data with 5km spacing. The
load is low, and the solar resource is high and variable. network was deployed midway through 2011, so only data
These timeframes can be challenging at the distribution for the fall were used to evaluate minimum load high
level as the PV resource is most likely to back-feed the variability conditions. Both datasets are shown in Figure
distribution substation and/or create undesirable voltage 1, which provides a view of SMUD’s service territory.
fluctuations on a distribution feeder. Further, as solar
penetrations increase significantly, it is possible that solar
will push up against minimum load levels of hydro
systems in the spring and nuclear and combined cycle gas
units kept online as efficient baseload or load following
resources. In these instances, certain resources could be
curtailed, as has been the case with oversupply of wind
and hydro resources in multiple locations around the US
and the world1, or alternatively pricing in the market
could go to zero or even negative, as has been explored by
Baldrick for the ERCOT electricity market in Texas in
2008 and 20092.
Finally, resources for load following may be challenged
during times of significant slightly non-coincident diurnal
ramps between solar and the grid. Likely situations here
include slightly time delayed up and down ramps of the Fig. 1: Map of SMUD Service Territory,
solar resource relative to morning and evening utility SolarAnywhere© 10 kM Grid blocks (blue) and NDFD
ramps. For timeframes where both ramps are steep, Grid cells and primary and secondary ground monitoring
significant load following resources may be needed for sites
very short periods of time, implying increased operating
costs for these resources if it increases the starts and stops Solar data was modeled using the solar resource data
on thermal units in particular. These conditions, as well as based on a validated Excel based model for PV output
the potential curtailment conditions are not likely to be that uses inverter efficiency, module temperature effects,
problematic for a number of years until solar penetrations tilt, azimuth, and dust factors to estimate hourly solar
are 6 or 7 times greater than current levels, however given output based on measured Global Horizontal Irradiance
the rapid growth of solar PV, it is prudent to begin (GHI) and Direct Normal Irradiance (DNI).
exploring the conditions and planning for them today.
3. PV PERFORMANCE DURING PEAK DEMAND
By using 9 years of hourly 10 km resolution PERIODS
SolarAnywhere© data, this paper examines PV output and
correlated utility load during these timeframes to evaluate SMUD’s peak demand historically occurs during the
how challenging each may be for SMUD to address. summer months, most typically in July, but occasionally
during June and August. The peak is driven primarily by
2. SOLAR RESOURCE, PV MODELING AND LOAD residential air conditioning, which contributes
DATA approximately 30% of overall demand during peak
conditions. Overall, SMUD’s historic peak demand was
To perform the analysis in this study, 9 years of hourly set in 2006 at 3,299 MW. Most typically, this demand
data were downloaded from the SolarAnywhere© tool occurs later than the CAISO peak due to the heavily
covering the Sacramento region. The data collected had a residential component. As temperatures rise in the hot
10 km resolution, which is the standard resolution offered afternoon, residents come home from work and turn on
by the tool. Enhanced resolution is also available which their air conditioning as businesses are still continuing to
light and cool buildings, resulting in a peak that occurs
2

3. between 5 and 6 PM.
Typically, SMUD meets the majority of its peak demand
needs with internal natural gas and hydro generation, and
imports power from the Northwest and CAISO markets to
supplement our internal generation. However, as PV
penetrations increase, the utility will look to it as a
resource to be counted on during times of peak demand.
Because backup resources are most expensive during peak
demand periods, it will be of significant financial
importance to understand the solar PV resource during
these timeframes. Fig. 3: Comparison of Modeled Output Profiles for PV
systems on SMUD’s Peak Setting Day, 2006
By looking at historical loads and modeled PV output
coinciding with those peak demand periods, a clear 4. MINIMUM LOAD PERIODS WITH COINCIDENT
picture of the relative reliability of the solar resource MAXIMUM PV OUTPUT
during peak demand periods can be attained. For this
analysis, we examined days with peak demand above Beyond peak periods, utility operators and planners are
2,800 MW, which is approximately 500 MW below the most concerned with the operation of both distribution
all time peak. Over the 9 year period, there were only 142 systems and the overall transmission system during low
hours which exceeded 2,800 MW, or roughly 0.18% of demand periods that coincide with high PV output
the hours. Of these hours, the solar resource followed a periods. For SMUD, generally these periods occur during
consistent pattern during the peak hours, contributing spring and fall months, where there is limited need for
between 35% and 40% of rated output for a South facing cooling and/or electric heating, resulting in daytime loads
system during the Hour Ending 6 PM for all of the 37 that may be between 30 and 40% of the peak demand
peak load days examined. This peak load and modeled period. Similarly, at the distribution level, minimum load
generation, along with the curves for SMUD’s all time periods generally occur during these same months, though
peak day set in 2006 are shown in Figure 2. can vary a bit more due to the different types of customers
that may be connected to a given distribution feeder.
Compounding the challenges associated with these time
periods at the transmission level are high variability days
and the availability of the hydro system to serve as a
regulating and balancing resource. It is not uncommon
under certain springtime runoff conditions for the hydro
system to be running at full output to avoid spill
situations. During these times, the system is not able to
provide the regulation services that it is able to provide
Fig. 2: South facing PV Output during top 0.18% of Load during other times of the year. It is not able to ramp up or
Hours ramp down as it is running at full output. During fall
periods, the hydro system may also be constrained in its
One way to increase the amount of PV capacity available ability to provide regulation services, in particular in its
on peak is by orienting it towards the West or ability to reduce output in response to increased PV
implementing tracking PV systems. This can increase on- output. This constraint can occur when there is inadequate
peak capacity available from a given system up to 60-65% water in the reservoirs to support generation that could be
as shown in Figure 3, as compared to the 35-40% ramped downwards. While historically these conditions
available from a South facing system. have not been problematic for SMUD operators, as the
utility adds significant amounts of solar these conditions
are likely to increase in significance.
As PV penetrations increase, it will become increasingly
likely that coincident high PV, hydro, and wind
generation could create conditions conducive to
curtailment of resources during minimum load conditions.
Understanding when these conditions may occur, at what
3

4. penetrations, and for what frequency will help utility basis, suggesting voltage control on such a feeder could
planners identify optimal strategies for resource be challenging on such variable days. As the data has only
development, curtailment, energy storage, and been collected for a 6 month period, it is uncertain what
complementary grid resources that ensure flexibility and the total number of high variability days will be yet, but is
minimize overall costs. For this analysis, we examined the expected to be between 40 and 60 days per year based on
frequency of low-load, high-PV output conditions that the data collected so far.
would likely create challenges at the transmission level.
The data were filtered to identify load periods between 30
and 40% capacity factor that matched PV output periods
exceeding 50% capacity factor. The results are
summarized in Table 1.
TABLE 1: PV Output during periods of minimum load
(30% - 40% of peak), average values over 9 year study
period
PV Total % Hours % Energy
Capacity Hours Exceeding PV Exceeding PV
Factor CF CF
50% 636 15% 25%
60% 373 9% 16%
70% 149 4% 7% Fig. 4: Modeled 1 minute change in output for a 3 MW
80% 15 0% 1%
PV system for 4 variable days in October, 2011
90% 1 0% 0% 5. PERIODS OF SIGNIFICANT RAMPS
In addition to concerns at the transmission level, The last period of significance to a utility is generally
distribution system planners and operators have expressed much more broad than either of the first two. It is also
likely to be further into the future than either of the first
concern about PV driving voltage and backfeeding
two concerns. This period really reflects the up and down
substations which is not consistent with their design
ramps associated with the rising and falling sun and the
conditions. As noted above, many of these conditions will
be specific to the load characteristics of a given feeder, typical up and down of a daily utility load. While the
however generally the same time periods are likely to be trajectory of solar crudely approximates the trajectory of
of concern as air conditioning loads are at a minimum. utility load, in fact most days there is an offset between a
Rather than concern about specific resources and markets, utility demand ramp and the ramp up of solar generation.
distribution system concerns relate to the second to Similarly, as the solar resource dwindles, utility demand
second and minute to minute variability of PV systems on often drops several hours later. The result of this slight
a specific feeder. As some feeders may feature relatively mismatch is that as penetrations increase, resources
dense concentrations of PV systems, high variability in previously used to meet demand will be significantly
turned down and potentially off. While from a greenhouse
their combined output is possible, and could lead to
gas mitigation perspective this is ultimately the goal, for
reliability concerns and substandard voltages for other
utility operators this may mean significant increases in
customers on the feeder. SMUD is currently collecting
data from our high density solar monitoring network to operating costs of thermal units required to ramp up and
support evaluations of PV output at the distribution level, back down, and potentially off, followed by a restart and
and will be working with Clean Power Research to second ramp up during the evening. Adding one restart
evaluate variability at the feeder level using high per day could substantially increase O&M costs and
resolution SolarAnywhere© data. would undoubtedly impact plant emissions and
efficiencies. Alternatively, it could mean the deployment
In the meantime, analysis using SMUD’s high-density of significant storage resources, new targets for demand
solar monitoring network indicates that multi-MW response applications, and/or other zero emission flexible
systems may cause significant challenges during high- generation as a means of addressing the issue without
increasing emissions.
variability days on a given feeder. SMUD currently
allows PV to be sized to 100% of minimum daytime load
on a feeder. This data indicates changes that would be the
equivalent of 50% of this load occurring on a 1 minute
4

5. perceived by the non-PV resources on SMUD’s system.
The result would be two-fold. First, SMUD could be
forced to either curtail PV generation or sell generation
into what would likely be a low-priced electricity market,
Minimum load as evidenced by what is occurring in Germany currently
as a result of high PV penetrations3. Second, as SMUD
ramped up generation to meet load in the morning,
peaking or simple-cycle load following generation would
be required to be turned on to cope with the short duration
of need. This same type of generation would be looked to
again as the sun set to replace PV in the resource mix and
meet the evening load. Such operation would represent as
significant departure from current utility operations, as it
would rely much more heavily on simple cycle generation
and would incur significantly increased O&M costs
Minimum load
associated with these types of generation, given current
operations rarely require more than one start in a day, and
generally involve operation for 6 to 8 hours rather than a 2
to 3 hour operating mode as indicated in Figure 6.
The intent of this significantly simplified assessment is to
point out the challenges in operations that could occur as
a result of significant penetrations of PV and to begin to
identify mitigation approaches for dealing with these
challenges. It is likely that interaction with the broader
Minimum load
Northwest and California electricity markets may mitigate
some of the concerns with regard to resources, and further
that other resources on SMUD’s system, in particular our
Upper American River System, would be used for
meeting some of the ramping demands, alleviating much
of the concern regarding double-starts of thermal units.
However, as PV penetrations increase statewide,
availability of market units and or hydro assets for early
morning ramps may become increasingly unlikely as
Minimum load
these more flexible resources are deployed to deal with
intermittency issues and firming of solar and wind.
6. CONCLUSIONS
Utility operators have historically been distrustful of the
Fig. 5: Ramp timing implications for solar and load for ability of PV resources to meet their needs from a
March, June, September and December, modeling 1,000 reliability perspective. Concerns about consistency on
MW PV and average loads peak, low-load variability, and implications for
complementary grid resources generally have hamstrung
For this assessment, as described in Figure 6, it was widespread acceptance of PV as a valued grid asset.
assumed that the minimum load threshold in each night However, using tools that have only been recently
represented the minimum turndown for utility resources. developed, utilities today are much more able to model
It was also assumed that SMUD expanded its current output of significant penetrations over longer historical
installed PV capacity from an expected 140 MW at the periods to better understand how these generating units
end of 2012 to 1,000 MW, which is currently beyond perform during time periods of interest.
anything in the 20 year planning horizon, but is certainly
possible if PV prices continued to drop and carbon In this paper, three time periods of interest were explored
constraints drove significant additional renewables that could pose challenges to the utility depending on the
beyond 2020. In this scenario, a new minimum load performance of PV generating units. The first time period,
would be achieved during the daytime, at least as corresponding to times of peak demand, was found
5

6. historically to coincide with highly predictable solar ensure that the utility is best prepared for significant
resources. In fact the primary variables related to increases in solar generation in our mix.
predicting solar output on a peak day relate more to the
orientation of the system than the availability of the solar (1) Porter, K., Rogers, J., Wiser, R. “Update on Wind
resource. Based on examination of 37 days representing Curtailment in Europe and North America” June 16, 2011
the top 0.18% of peak demand hours, the modeled PV http://www.efchina.org/csepupfiles/report/201163041046
generation was found to perform within a very narrow 377.1820833464392.pdf/Update%20on%20Wind%20Cur
band between 35% and 40% capacity factor for a South
Facing system during the peak hour. As a result, it was
determined that PV generation could be counted on for a
consistent amount of output during these time periods, (2) Baldrick, Ross “Wind and Energy Markets: A Case
which will assist in planning for SMUD’s future Load Study of Texas” 2011, US Association for Energy
Serving Capability with large penetrations of solar on the Economics Dialogue
grid. http://dialogue.usaee.org/index.php?option=com_content
&view=article&id=109&Itemid=170
The second timeframe of significance, the coincidence of Accessed 3/2/2012 3:03 PM PST
low utility load and high PV output was also examined
from a macro and micro level. At a macro level, (3) Shahan, Zachary “Solar PV Reducing Price of
uneconomic conditions could occur leading to curtailment Electricity in Germany” February 9, 2012
or sale of generation into markets at rates below the long- CleanTechnica.com
term expected value of that generation. The frequency of http://cleantechnica.com/2012/02/09/solar-pv-reducing-
these types of conditions could represent as much as 7% price-of-electricity-in-germany/ Accessed 3/2/2012 5:37
of the hours of the year, where utility loads were between PM PST
30 and 40% of peak and PV output was greater than 50%
of rated capacity. In addition, at the distribution level,
extreme variability, including ramps of up to 50% of
capacity in a minute is likely to create issues on a few
dozen days of the year. These conditions could potentially
lead to local curtailment, changes in distribution system
design, or the inclusion of new distributed storage or other
grid assets to compensate for these extreme conditions.
Finally, over the longer term, scenarios may develop
similar to those in Germany, where daytime loads net of
PV are as low or lower than nighttime loads. With
significant PV penetrations throughout the state, and the
non-coincidence between load and solar ramps, existing
generation will be less useful in filling in short term needs
for morning and evening generation, leading to a
potentially significantly price differentiated market.
Simple cycle generation, storage, and demand response
will likely be better suited for meeting ramps than
combined cycle generation given the short lengths and
significant ramp magnitudes that may occur.
Understanding these conditions can help guide the types
of grid resources that utilities plan for over the next
decade.
This exercise was an initial attempt to begin to answer
some of the questions around the impacts of PV on
existing utility operations. Further work is expected to
refine these assessments, and develop more realistic long-
term scenarios against which to evaluate complementary
grid resources. However, the exercise has shown the
benefits of beginning to think about these timeframes to
6