Cloud Enhancement Of Global Horizontal Irradiance In .

1y ago
5 Views
1 Downloads
1.23 MB
11 Pages
Last View : Today
Last Download : 2m ago
Upload by : Milo Davies
Transcription

Available online at www.sciencedirect.comScienceDirectSolar Energy 130 (2016) 128–138www.elsevier.com/locate/solenerCloud enhancement of global horizontal irradiance in Californiaand HawaiiRich H. Inman, Yinghao Chu, Carlos F.M. Coimbra Department of Mechanical and Aerospace Engineering, Jacobs School of Engineering, Center for Renewable Resource Integration and Center forEnergy Research, University of California San Diego, La Jolla, CA 93093, USAReceived 22 August 2015; received in revised form 5 January 2016; accepted 6 February 2016Communicated by: Associate Editor David RenneAbstractClouds significantly attenuate ground-level solar irradiance causing substantial reduction in photovoltaic power output capacity.However, partly cloudy skies may lead to temporary enhancement of local Global Horizontal Irradiance (GHI) above the clear-sky ceiling and, at times, the extraterrestrial irradiance. Such enhancements are referred to here as Cloud Enhancement Events (CEEs). In thiswork we study these CEEs and assess quantitatively the occurrence of resulting coherent Ramp Rates (RRs). We analyze a full year ofground irradiance data recorded at the University of California, Merced, as well as nearly five months of irradiance data recorded at theUniversity of California, San Diego, and Ewa Beach, Hawaii. Our analysis shows that approximately 4% of all the data points qualify aspotential CEEs, which corresponds to nearly 3.5 full-days of such events per year if considered sequentially. The surplus irradianceenhancements range from 18 W m 2 day 1 to 73 W m 2 day 1. The maximum recorded GHI of 1400 W m 2 occurred in San Diegoon May 25, 2012, which was nearly 43% higher than the modeled clear-sky ceiling. Wavelet decomposition coupled with fluctuationpower index analysis shed light on the time-scales on which cloud induced variability and CEEs operate. Results suggest that whilecloud-fields tend to induce variability most strongly at the 30 min time-scale, they have the potential to cause CEEs that induce variability on time-scales of several minutes. This analysis clearly demonstrates that CEEs are an indicator for periods of high variability andtherefore provide useful information for solar forecasting and integration.Ó 2016 Elsevier Ltd. All rights reserved.Keywords: Global horizontal irradiance; Cloud enhancement; Ramp rate; Wavelet1. IntroductionThe total extraterrestrial beam irradiance incident onthe earth’s atmosphere I 0 fluctuates about an average valueof approximately 1360 W m 2 (Kopp and Lean, 2011).This incident radiation is attenuated as it negotiates itsway to ground level through a complex series ofmultiple reflections, absorptions and re-emissions due to Corresponding author.E-mail address: ccoimbra@ucsd.edu (C.F.M. 02.0110038-092X/Ó 2016 Elsevier Ltd. All rights reserved.interactions with atmospheric constituents (Goody andYung, 1995). This results in the division of the incidentextraterrestrial beam radiation into two distinct components; Direct Normal Irradiance (DNI) and Diffuse Horizontal Irradiance (DHI), the geometric sum of which isthe Global Horizontal Irradiance (GHI), defined by theclosure equation,GHI ¼ DNI cos hz þ DHIð1Þwhere hz is the solar zenith angle, see Fig. 1.In addition to the partitioning of the radiation,atmospheric cloud-formation is typically associated with

R.H. Inman et al. / Solar Energy 130 (2016) 128–138DHIDNICSMz161 156800144 139600127 122400111 10620094 8907277 60 55z [Degrees]1,000Solar Zenith Angle,Irradiance [W m-2]GHI12943 386:00 AM9:00 AM12:00 PM3:00 PM6:00 PMTime [PDT]Fig. 1. Components of solar irradiance sampled every 30 s at the University of California, Merced, on March 21, 2011. Global Horizontal Irradiance(GHI) was measured with a Precision Spectral Pyranometer (PSP), manufactured by the Eppley Laboratory, Inc. Direct Normal Irradiance (DNI) wasmeasured using a Normal Incidence Pyrheliometer (NIP) and 2-axis automatic solar tracker (SMT-3), both of which are also available from the EppleyLaboratory, Inc. Diffuse Horizontal Irradiance (DHI) was measured using an additional PSP and SMT-3 with accompanying Shade Disk Kit (SDK).Solar zenith angle is also plotted and observations with cos hz 6 0:3 ðhz P 72:5 Þ, shown outside solid red-lines, being excluded from the study. The ClearSky Model (CSM) used in this figure is explained in detail in Section 3.2. (For interpretation of the references to color in this figure legend, the reader isreferred to the web version of this article.)a pronounced decrease in the intensity of solar irradiancecomponents. In fact, the attenuation of incoming solarradiation by clouds is routinely larger than any other atmospheric component (Inman et al., 2013). Furthermore, thedriving effects of clouds on radiative energy budgetsinclude short wave cooling, as a result of absorption ofincoming solar radiation, and long wave heating, due toreduced emission of thermal radiation by relatively coolcloud tops (Schade et al., 2007). However, partly cloudyskies may lead to the reverse; i.e., multiple scatteringsand reflections of short wave radiation by cloud fieldsmay lead to increased irradiance from a portion of thesky above the corresponding cloud-free value(Franceschini, 1968; Wen et al., 2001; Wyser et al., 2002;Pfister et al., 2003; Emck and Richter, 2008; Berg et al.,2011). In rare occasions, these enhancements can causethe local GHI to instantaneously exceed the extraterrestrialsolar constant I 0 (McCormick and Suehrcke, 1990; Tapakisand Charalambides, 2014). Such enhancement of GHIabove the corresponding clear-sky value are referred tohere as Cloud Enhancement Events (CEEs).With the exception of two studies, one by Luoma et al.(2012) and a second by Tapakis and Charalambides (2014),little work has been done in the realm of CEEs with respectto PhotoVoltaic (PV) power generation. In this work we donot intend to suggest a new mechanism by which CEEsoccur, but rather investigate the coherent Ramp Rates(RRs) associated with CEEs and their potential impacton the quality of PV power generation. To clarify, a coherent CEE RR is defined as a series of monotonically increasing or decreasing GHI observations whose maximum valueexceeds the expected clear-sky value by a given threshold.These CEEs and their associated RRs are of interest forseveral reasons: current models typically do not considerthe ability of clouds to increase the local availableirradiance, these events commonly precede or follow periods of lower than normal irradiance associated with thepresence of passing opaque clouds leading to relativelylarge RRs, large RRs can cause voltage flicker that in turntrigger tap chargers on distribution feeders increasing operations cost for utilities, and therefore the successful forecasting of these events could lead to an effective controlscheme to reduce the cost associated with high levels ofvariability in photovoltaic power generation.Analysis of the amplitude, persistence, and frequencyoccurrence of ground-level irradiance fluctuation requiresa decomposition of the input time-series into a set oforthogonal sub-signals each representing a specified timescale of fluctuation. Due to the stochastic and nonperiodic nature of the atmospheric processes that driveground-level fluctuations in clearness, Fourier analysis istypically not suitable. Alternatively, spectral analysis ofhigh frequency (e.g., 1 s to 1 min) irradiance time-serieson a wavelet basis rather than a periodic basis can be foundin the literature. Kawasaki et al. (2006) decomposed 2years of 1-min irradiance data from 9 sites in a 4 4 kmgrid using the Daubechies 4 wavelet (Hazewinkel, 2001).Woyte et al. (2007) applied the Haar (1911) wavelet toclearness index time-series and defined a fluctuation powerindex (fpi) that quantified the amplitude and frequencyoccurrence of variability on specified time-scales.Perpiñán and Lorenzo (2011) analyzed several days of 1 sirradiance time-series using the MOD-WT wavelet andlater used wavelet transform correlations to study fluctuations of the electrical power generated by an ensemble of 70DC/AC inverters from a 45.6 MW PV plant (Perpiñánet al., 2013). Lave et al. (2012) applied a wavelet transformto clear-sky index time-series from a single site to the average of six sites and showed a strong reduction in variabilityat short time-scales (i.e., shorter than 5-min) with lesser

130R.H. Inman et al. / Solar Energy 130 (2016) 128–138reductions at longer time-scales. Lave et al. (2013) alsodeveloped a wavelet-based variability model for PV plantoutput and Lave et al. (2013) employed the wavelet-basedmodel to study the impact of cloud speed on solar variability scaling. Similarly, Peled and Appelbaum (2013) usedstatistical tools and wavelet analysis to develop estimatorsfor the magnitude of voltage and power variations withinPV systems due to climatic conditions.The time-scales of these events have been shown torange from seconds (Schade et al., 2007), to 15–30 min(Segal and Davis, 1992; Tapakis and Charalambides,2014), and occasionally as long as hours (Cede et al.,2002). Here we employ a wavelet decomposition usingthe top-hat wavelet to show that while the CEE RRs typically only last a few minutes they are generally correlatedwith variability on longer time-scales associated with thepassing of cloud fields.This study makes use of a full year of irradiance datarecorded at the University of California, Merced, as wellas nearly five months of irradiance data recorded at theUniversity of California, San Diego, and on South Oahu,Hawaii, see Section 2. Methods are described in Section 3which includes descriptions of the clear-sky model, statistical analysis, ramp rate calculation, wavelet decomposition,and fluctuation power index analysis. The results of theseanalyses are covered in Section 4 and the conclusions inSection 5 respectively.and manifest in the global component of irradiance as aresult of Eq. (1). This is because CEEs result from cloudselevating short wave diffuse irradiance above that of thecorresponding clear-sky value with little or no change inDNI, which is illustrated through an analysis of DNIclear-sky indices,2. Data3. Methods2.1. Irradiance data3.1. Data qualityIrradiance data are collected at three locations characterized by different micro-climates: Merced California(continental), San Diego California (coastal), and EwaBeach Hawaii (island). A broadband (285–2800 nm) Eppley Precision Spectral Pyranometer (PSP) and Normal Incidence Pyrheliometer (NIP) are used in Merced (March 21,2011 to June 19, 2012) to collect GHI and DNI data,respectively. Two Multi-Filter Rotating ShadowbandRadiometers (Model MFR-7), available from YankeeEnvironmental Systems (YES), which measure both GHIand DNI are used to collect irradiance data in San Diegoand Ewa Beach during the period from January 27, 2012to June 17, 2012. Irradiance data are logged using Campbell Scientific CR-1000 data-loggers at a sampling rate of30 s. It is important to note that due to the 30 s samplingrate used in this study the observed RRs are likely lowerthan what would be observed with an increased samplingrate, see for example Yordanov et al. (2013a,b), andAlmeida et al. (2014).Data from the early morning and late afternoon areexcluded from the analysis. There are several motivationsfor their removal: primarily, the cosine response of pyranometer measurements are typically maximized forhz J 70 ; secondly, times when hz J 70 are associated witha relatively high airmass resulting in a large fraction ofGHI originating from the diffuse component; and finally,the high solar zenith angle in combination with an elevatedairmass results in rather low photovoltaic power production. To this end, a threshold is applied to the solar zenithangle hz according to,2.2. Selection of GHIThis work focuses on GHI rather than DNI due to thenature of the cloud enhancement process. CEEs are necessarily observable in the diffuse component of irradiancek D ðtÞ ¼DNIðtÞDNIclr ðtÞð2Þand GHI clear-sky indicesk G ðtÞ ¼GHIðtÞGHIclr ðtÞð3Þwhere GHIclr ðtÞ and DNIclr ðtÞ are generated from theClear-Sky Model (CSM) described in Section 3.2. Asshown in Fig. 2, k D generally increases with k G for nonCEE periods and typically results in correlation coefficientsgreater than 0.9, which is to be expected (see Table 1). Onthe other hand, during CEE periods, the correlationbetween k D and k G is low and in some cases negative (seeTable 1). It is also clear that during CEE periods, k G is typically in the range of 1.05 to 1.5 while k D tends to occupythe range from 0.5 to 1.0. It is also important to note thatthere are some occasions where k D did exceed 1.05, however, such enhancement was found to result from errorsin the DNI CSM at elevated zenith angles under cloudlessskies and, as a result, were excluded from the study.cos hz P 0:3;ð4Þwith data not satisfying Eq. (4), which corresponds tohz P 72:5 , being removed from the data set, see Fig. 1.Furthermore, in order to avoid erroneous observationof CEEs resulting from reflections from local surfaces, aquality filter based on the summation method to obtainGHI was used (Gueymard and Myers, 2009). This filtercalculates GHI from Eq. (1) and compares it against theindependently observed GHI. Observations where the ratioof calculated and observed GHI differ from 1.0 by morethan 0.2 where excluded as they typically correspond toan error in the observation of at least one irradiancecomponent.

R.H. Inman et al. / Solar Energy 130 (2016) 128–1382MercedSan DiegoNon-CEECEE131Ewa BeachNon-CEECEENon-CEECEE1.5kD 10.500.511.500.511.500.5kGkG11.52kGFig. 2. Scatter plots of k D versus k G for irradiance measurements in Merced, CA (left); San Diego, CA (center); and Ewa Beach, HI (right). Each markerstands for a single observation. Red markers represent GHI CEE cases and the dash lines represent the 1.05 threshold for both k D and k G . (Forinterpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)3.2. Clear-sky modelIn addition to the measured irradiance data, a Clear-SkyModel (CSM) is used to characterize events that exceed theclear-sky ceiling. The clear-sky GHI is modeled using theCSM developed by Ineichen and Perez (2002), whichrequires Linke Turbidity as an input. Maps of monthlyaverage Linke Turbidity developed by Remund et al.(2003) were used. These maps account for seasonal changesof aerosol content in the atmosphere and their performancehas been confirmed internally by Sandia National Laboratories (Reno et al., 2012). The CSM was calibrated againstseveral clear days and a linear fit is used in order to ensureagreement between measured and modeled values.3.3. Statistical analysisA statistical analysis was carried out in order to characterize the individual events that exceed the CSM into boundedranges. Only GHI measurements that exceed the CSM by atleast 5% and satisfy Eq. (4) are considered potential CEEs.Potential CEEs are then grouped by the degree to which theyexceed the CSM (5–10%, 10–15%, etc.). Statistical analysisof these subsets of CEEs provides insight into the distribution and probability of GHI measurements as a functionof the degree to which they exceed the CSM.violating the quality filter being removed. As was done inSection 3.3, only events which exceed the CSM by a threshold of at least 5% are considered potential CEEs. However,rather than grouping the individual measurements by therange with which they exceed the clear-sky model (5–10%, 10–15%, etc.), the present study employs a slidinglower bound which classifies coherent ramps by theamount with which their maximum value exceeds theCSM ( 5%, 10%, etc.), see Fig. 5.In order to study the coherent CEE RRs, characteristicevents must be located and quantified. This was accomplished through the identification of successive measurements which lie on opposite sides of the CSM. Thesepoints were then extrapolated to their respective local maxima and minima. This is illustrated in Fig. 3 in which theramps associated with CEEs, along with their local maximaand minima, have been identified in the Eppley PSP datafrom March 21, 2011 in Merced. It is important to notethat while raw GHI data was used to locate potential CEEs(and the degree to which they exceed the CSM) the calculation of the RRs themselves ignores deterministic diurnalvariations through the removal of the CSM values asdescribed below,jRRj ¼ðGHImax CSMmax Þ ðGHImin CSMmin Þ:Dtð5Þ3.5. Wavelet analysis3.4. Ramp rate analysisAnalysis of the RRs uses the same set of GHI data fromthe statistical study with data not satisfying Eq. (4) orTable 1Correlation coefficients between DNI clear-sky indices and GHI clear-skyindices for overall, non-CEE, and CEE periods at different locations.Correlation coefficientMercedSan DiegoEwa BeachOverallNon-CEECEE0.90.90.940.920.910.950.07 0.16 0.03Wavelet analysis allows for decomposition of nonperiodic time-series into sets of orthogonal sub-signals representing fluctuations on specific time-scales, see for example Mallat (2009). For the purposes of the present study awavelet similar to the top-hat wavelet employed in Laveet al. (2012, 2013) is also used here. The top-hat waveletused here is centered at zero and is defined as,8 1 4 t 1 4 1ð6ÞwðtÞ ¼ 1 1 2 t 1 4 k 1 4 t 1 2 :0elseso that the dictionary of top-hat atoms can be written,

132R.H. Inman et al. / Solar Energy 130 (2016) 128–138GHIIneichenRamps1,000Irradiance [W m 2]80060040020007:37AM8:26AM9:16AM 10:06AM 10:56AM 11:46AM 12:36PM 1:26PM2:16PM3:06PM3:56PMTime [PDT]Fig. 3. Ineichen clear-sky model as well as GHI data from the Eppley PSP located in Merced for the portion of March 21, 2011 satisfying Eq. (4). Thelocal maxima and minima of the coherent ramps associated with CEEs are highlighted in red. Coherent CEE ramps are defined as a set of monotonicallyincreasing or decreasing irradiance observations whose maximum exceed the CSM by a specified threshold. (For interpretation of the references to color inthis figure legend, the reader is referred to the web version of this article.) D¼ 1t sws;2 j ðtÞ ¼ re the wavelet has been offset by s and scaled bypffiffiffiffiffiffiffiffi1 2jþ1 in order to compensate for the total length ofthe wavelet. The corresponding wavelet transform iswritten Z þ11t sjWf ðs; 2 Þ ¼ pffiffiffiffiffiffiffiffif ðtÞwdtð8Þ2j2jþ1 1Fig. 4 shows the w0;22 ðtÞ top-hat wavelet centered at zeroas well asntwo scaled, dilated,o and translated versions of thewaveletw8;23 ðtÞ; w24;24 ðtÞ . These wavelets, which have alength of 2 j , when applied to a stationary time-series ofirradiance measurements would capture clear periods oflength 2j 1 surrounded by overcast periods or vice versa.In order to determine the power contained in each of thetime-series Wf ðs; 2 j Þ, a wavelet periodogram I j ðsÞ is calculated. In analogy to the Fourier periodogram as well as thework by Woyte et al. (2007) and Lave et al. (2012, 2013) thewavelet periodogram is defined as the square of the wavelettransform, normalized by the period of the wavelet,I j ðsÞ ¼2½Wf ðs; 2 Þ 2jþ1jð9ÞHowever, information regarding the variability from thewavelet periodograms is inconvenient to examine periodslonger than a day. To remedy this, a fluctuation powerindex as described by Woyte et al. (2007) and employedin Lave et al. (2012) is used here to characterize the amountFig. 4. Illustration of the w0;22 ðtÞ top-hat wavelet centered at zero as wellas two scaled,no dilated and translated versions of the waveletw8;23 ðtÞ; w24;24 .of power included in each time-scale. The fluctuationpower index (fpi) is defined as,Z Tj21½Wf ðs; 2 j Þ fpiðjÞ ¼dsð10ÞTj 02jþ1where T j is the total duration of the wavelet periodogramI j ðsÞ. It should be noted that while all of the definitionsconcerning wavelet transforms have been written in thecontinuous sense, our dataset is discrete in nature.

R.H. Inman et al. / Solar Energy 130 (2016) 128–13855%1.4No CEE5% to 10%10% to 15%15% to 20%20% to 25%25% to 30% 30%1.3Measured GHI (kW/m2)1.21.1 96.04% 1.68% 0.94% 0.56% 0.34% 0.22% 0.22%I j ðsÞ ¼50% 2Wf ðs; 2 j Þ2jþ1133;ð13Þand45%"#TXj 1Dtpffiffiffiffiffiffiffiffi I j ð0Þ þ I j ðT j Þ þ 2 I j ðtÞ ;fpiðjÞ ¼2T j 2jþ1t¼140%35%1.0ð14Þrespectively, where N represents integers between 2j 1 andN 2j 1 .30%0.925%0.820%4. Results0.715%4.1. Statistics0.610%0.55%0.40.40.50.60.70.80.9Clear GHI (kW/m2)Fig. 5. Plot of the entire year of GHI data from Merced as a function ofthe expected clear-sky value illustrating the set of potential CEEs.Therefore, discrete forms of Eqs. (7)–(10) must beemployed. The discrete forms are, D¼ 1t sws;2 j ðtÞ ¼ tWf ðs; 2 j Þ ¼ pffiffiffiffiffiffiffiffi2 2jþ1" #NX sN st sf ð0Þw j þ f ðN Þwþ 2 f ðtÞw;ð12Þ22j2jt¼1A plot of the entire year of GHI data from Mercedalong with suspected cloud enhancement events where themeasured GHI exceeds the modeled clear-sky ceiling areshown in Fig. 5. Allowing for errors in the measurementof GHI and modeling of clear-sky irradiance ceiling, potential enhancement events are associated with measured values which exceed the clear-sky model by at least 5%. Ouranalysis shows that 10,228 of the 262,281 data points, corresponding to approximately 4% of the data, qualify aspotential CEEs. Multiplying the number of potential datapoints by 30 s time-steps corresponds to over 3.5 fulldays of these events per year if considered sequentially.The surplus irradiance enhancements range from18 W m 2 day 1 to 73 W m 2 day 1. The maximum GHIin Merced was measured on May 8th, 2011 at 12:08 p.m.ðhz ¼ 20:5 Þ with a value of 1365 W m 2, or approximatelyequal to the extraterrestrial beam irradiance I 0 . This valueis nearly 40% higher than the modeled clear-sky ceiling suggesting substantial gains may result from these events.Table 2Statistical results for CEE analysis for an entire time period (456 days) in Merced, %P30%TotalMeanStandard deviationOccurrenceProbabilitySurplus energy(W m 2)(W m 2)(–)(%)(W h m 64.413.382.572.041.73955810,2283.9830018.19(W h m 2 day 1)Table 3Statistical results of CEEs at three observatories for the time period when all data from all sites was available (142 days).LocationMercedSan DiegoHawaiiRangeMeanStandard deviationProbabilitySurplus energyFraction of total GHI(–)(W m 2)(W m 2)(%)(W h m 2)(W h m 2 day 1)(%)January 29–June 19, 2012January 29–June 17, 2012January 27–June 17, 8.472.91.21.01.3

134R.H. Inman et al. / Solar Energy 130 (2016) 128–138Table 4Data used in the calculation of P ðRRÞ separated into both up rampingevents and down ramping events.PPPi Dtii Dt ii Dti(Up RRs)(Down RRs)(CEE RRs)(h)(h)(h)15%10%15%20%25%30%Day Values0.90.800100.6100.51010-1P(RR)0.40.3P(Ramp Rate)P( RR .3911.886.60134.289.6058.6536.4023.9813.370.20.1RR [W m-2 min-1]1000-100000200800600-600-200 0 200Ramp Rate [W m 2 min 1]40020002004002 600 1-2-1RR RR[W]0 [W m mini ]400800600600100010008001000Fig. 6. Cumulative Distribution Functions (CDFs) of the coherent RampRates (RRs) whose maximum values exceed the CSM by a increasingthreshold and are associated with Cloud Enhancement Events (CEEs).The CDF of the entire year of data that satisfy Eq. (4) is also shown forcomparison and labeled as ‘Day Values’. Intersection of the P 0.95 lineand the Day Values CDF occurs at 76 W m 2 min 1 and an average valueof 633 W m 2 min 1 for the CEE RRs. Inset: Probability DistributionFunction (PDF) of the Day Values as well the 10% threshold distributionshowing the symmetry of RR associated with the passing of opaque cloudsas well as the increased probability of elevated RRs.Statistical results from the data in Merced as well as similar results for data sets collected in San Diego and EwaBeach, are shown in Tables 2 and 3 respectively, all ofwhich recorded maximum values of GHI P I 0 . The maximum recorded GHI in San Diego of 1396 W m 2, which is4.5 x 10332.52j 761.5jj 54105%10%15%20%25%30%25fpiCEE/fpiNoCEE3.5fpi305% CEE10% CEE15% CEE20% CEE25% CEE30% CEE5% no CEE10% no CEE15% no CEE20% no CEE25% no CEE30% no CEE40.542.7% higher than the modeled clear-sky ceiling, occurredon May 25, 2012 at 12:45 p.m. (hz ¼ 17:6 ). Similarly, amaximum recorded value of 1380 W m 2, which is32.95% higher than the modeled clear-sky ceiling, occurredon April 29, 2012 in Hawaii at 12:18 p.m. (hz ¼ 7:2 ). It isbeneficial to note that the magnitude, mean and deviationof these potential enhancement events across all threeobservatories do not vary greatly. In addition, cloudenhancements in Merced are seasonally dependent andare more likely to occur in the spring months. Moreover,there are very few potential cloud enhancement events during the summer months when the skies are relatively cloudless over Merced. Due to the difference in local climatologybetween Hawaii and California, there are less clear skiesper year in Hawaii. Consequently, cloud enhancementevents are more likely to occur and provide higher surpluslocal energy than the other two observatories during thesame season, see Table 3. 56jj jj 43jj 78201510j 895j 23j 21j 12102103Timescale [s]0j 23j 34j 452j 56j 67j 78j 8931010Timescale [s]Fig. 7. (a) Average annual fluctuation power indices as a function of time-scale and CEE threshold. Days without CEEs tend to possess fluctuation powerindices that are typically 5–15 times lower in magnitude. Regardless of the time-scale (mode) all of the fluctuation power indices for a particular thresholdare maximum at a time-scale of approximately 30 min (j ¼ 6), suggesting that this is the dominant time-scale of clouds that are in turn responsible forCEEs. (b) Ratios of annual average fluctuation power indices of CEE to non-CEE days as a function of CEE threshold. Unlike (a), which show a peak atthe j ¼ 6 mode associated with approximately 30 min, the ratios show a peak at the j ¼ 3 mode which is associated with a time-scale of approximately4 min, suggesting that while the clouds that are responsible for creating CEEs tend to introduce variability on the order 30 min, the CEEs themselves tendto operate on time-scales about one eighth as long ( 4 min).

3,840 secClear Sky IndexR.H. Inman et al. / Solar Energy 130 (2016) 128–1381351.510.5jj 785001,920 sec5050jj 67050960 sec50jj 56050480 sec30jj 450240 sec3020j 34020120 sec20j 2302060 sec10j 120107:00AM8:00AM9:00AM10:00AM 11:00AM 12:00PM1:00PM2:00PM3:00PM4:00PM5:00PMFig. 8. The wavelet transforms from the clear-sky index for Merced at each time-scale, j ¼ 1 to 7 for the entire day of March 21, 2011 satisfying Eq. (4).4.2. Ramp rate probabilitiesCumulative Distribution Functions (CDFs) of the previously defined coherent CEE RRs are shown in Fig. 6. TheCDF for the entire data set satisfying Eq. (4) is also shownfor reference and labeled as ‘Day Values’. It should benoted that, as a result of varying climates, cloud inducedRRs vary by location. Cloud induced RRs resulting in achange of at least 5% in k G for Merced, San Diego andHawaii account for 6.86%, 15.6% and 44.3% of the data,respectively. It is clear from Fig. 6 that the CEE RR distributions corresponding to increasing thresholds are quitesimilar. In addition, one can see that CEEs are associatedwith elevated RRs. Fig. 6 also includes a horizontal linecorresponding to P ¼ 0:95. The intersection of this linewith the CDFs represent the RR magnitude which isexceeded 5% of the time. This intersection occurs atapproximately 76 W m 2 min 1 for the Day Values andan average value of 633 W m 2 min 1 for the CEE RRdistributions. This suggests a significant increase in the

R.H. Inman et al. / Solar Energy 130 (2016) 128–1383,840 secClear Sky Index1361.510.5jj 7820101,920 sec0jj 6720100960 secjj 562010480 sec010240 sec0120 secjj 3420100jj 232010060 secjj 4520jj 12201007:00AM8:00AM9:00AM10:00AM 11:00AM 12:00PM1:00PM2:00PM3:00PM4:00PM5:00PMFig. 9. The wavelet periodogram from the clear-sky index for Merced at each time-scale, j ¼ 1 to 7 for the entire day of March 21, 2011 satisfying Eq. (4).probability of inflated RRs associated with CEEs. Theinset of Fig. 6 shows the PDF of the 10% CEE RR caseas well as the PDF for the Day Values. The PDF isincluded to show the symmetry of the CEE RR distribution. This symmetry indicates that each ramp is associatedwith a corresponding ramp of similar magnitude and opposite sign. This is to be expected from the passing of a discrete and opaque cloud field. This behavior is also clearfrom a careful examination of Fig. 3 and is summarizedin Table 4. Only the 10% PDF case is shown for clarity,however it should be noted that the remaining distributionsare quite similar.4.3. Wavelet decompositionThe wavelet transforms and periodograms from the clearsky index for Merced at each time-scale, j ¼ 1–7 for the entireyear of data separated into days with and without CEEs of

R.H. Inman et al. / Solar Energy 130 (2016) 128–138increasing thresholds were calculated. As mentioned in Section 3.5, it is important to note that when the top-hat waveletof time-scale j is applied to a stationary time-series of irradiance measurements, it captures clear periods of length 2j 1surrounded by overcast periods or vice versa. The wavelettransform and periodogram for March 21, 2011 over modesj ¼ 1 (60 s) to j ¼ 7 (about an hour) are shown in Figs. 8 and9 respectively. This day was chosen not only because it wasused to illustrate the previous concepts but also because ithas transitions between clear and cloudy periods at severaltime-scales which would be detected by the top-hat wavelettransform. Upon inspection, it is clear from Fig. 8 that thewavelet transforms tend to be zero for tim

to PhotoVoltaic (PV) power generation. In this work we do not intend to suggest a new mechanism by which CEEs occur, but rather investigate the coherent Ramp Rates (RRs) associated with CEEs and their potential impact on the quality of PV power generation. To clarify, a coher-ent CEE RR is defined as a series of monotonically increas-

Related Documents:

The enhancement itself is performed in two steps: auto-enhancement, and personalized enhancement. The auto-enhancement step (Section 4.3) is necessary to handle bad quality photos that the system is not trained to handle. This step generates some kind of a baseline image that is then further adjusted using personalized enhancement.

sites cloud mobile cloud social network iot cloud developer cloud java cloud node.js cloud app builder cloud cloud ng cloud cs oud database cloudinfrastructureexadata cloud database backup cloud block storage object storage compute nosql

Speech enhancement based on deep neural network s SE-DNN: background DNN baseline and enhancement Noise-universal SE-DNN Zaragoza, 27/05/14 3 Speech Enhancement Enhancing Speech enhancement aims at improving the intelligibility and/or overall perceptual quality of degraded speech signals using audio signal processing techniques

FlexPod Hybrid Cloud for Google Cloud Platform with NetApp Cloud Volumes ONTAP and Cisco Intersight TR-4939: FlexPod Hybrid Cloud for Google Cloud Platform with NetApp Cloud Volumes ONTAP and Cisco Intersight Ruchika Lahoti, NetApp Introduction Protecting data with disaster recovery (DR) is a critical goal for businesses continuity. DR allows .

Multi-position, Upflow, Horizontal Left, or Horizontal Right, Modular Air Handlers Black Epoxy Coil GAF2A0A18S11EE GAF2A0A24S21EE GAF2A0A30S21EE GAF2A0A36S31EE. 2 Pub. No. 22-1858-10 Features and Benefits . Horizontal Left Horizontal Right 1 1 Upflow . Trane .

The number of workloads per installed cloud server will increase from 3.5 in 2010 to 7.8 in 2015. By 2014, more than 50 percent of all workloads will be processed in the cloud. Global cloud traffic: Annual global cloud IP traffic will reach 1.6 zettabytes by the end of 2015. In 2015, global cloud IP traffic will reach 133 exabytes per month.

Cloud Foundry Foundation Going Cloud Native with Cloud Foundry. Why does Cloud Native matter? Since 2000, 52% of the Fortune . Continuous Innovation. There is a rough consensus on many Cloud Native traits. Containers as an atomic unit, for example. Micro-services as the means of both construction and communication. Platform independence .

Cloud bursting is the simplest and most common hybrid/multi-cloud cloud model scenario, in which an application that is executing in a private cloud bursts into a public cloud when the demand for computing capacity spikes. The advantage of such a hybrid cloud deployment from a cloud