DOE/VT/EPRI Hi-Pen PV Project, Phase III

Transcription

1 DOE/VT/EPRI Hi-Pen PV Project, Phase III Smart Inverter Modeling Results, Variability Analysis, and Hosting Capacity Beyond Thresholds Matt Rylander Senior Project Engineer Wes Sunderman, Senior Project Engineer Jeff Smith, Manager, System Studies Chris Trueblood, Senior Project Engineer Steven Coley, Project Engineer Tom Key, Program Manager May 8, 2013 Blacksburg, VA

2 Overview Task 11: Simulation of higher penetration levels with smart inverter, grid support functions Task 15: Feeder PV Monitoring and Modeling Subtask : Continue feeder and PV system monitoring at existing sites (from Phase 2) and develop statistics from site data that describe variability at the distribution level Subtask 15.3: Compare simplified cloud modeling approach developed and implemented by Baylor in Phase 2 with EPRI s method of utilizing multi-point sensors for representing cloud variability within a feeder Subtask 15.4: Evaluate feeder hosting capacity if the feeder were allowed to operate outside normal operating criteria for varying percentages of time 2

3 Task 11: Simulation of Smart Inverters and Impact on Feeder Hosting Capacity for PV Objective: Demonstrate, via simulation, higher hosting capacity via use of advanced, grid-support functions Approach Utilize detailed feeder model of voltage-constrained feeder developed in Phase II Apply autonomous, advanced inverter control functions to PV Re-evaluate hosting capacity Compare new hosting capacity with results obtained from Phase II 3

4 Contents Outline Phase 3 Task 11 details Method for analysis Seven advanced inverter functions Four hosting capacity metrics Two PV types Two load levels Small-Scale PV results Summary of findings 4

5 Phase 3 Task 11 Details Phase 1 Developed PV system model in OpenDSS Phase 2 Utilized the Phase 1 model and determined the PV Hosting Capacity on two different feeders Phase 3 Utilized one of the feeders modeled in Phase 2 and determined the impact advanced inverter controls have on PV Hosting Capacity 5

6 Phase 3 Feeder Selection Feeder J1 Feeder K1 KUB Voltage Profile 6

7 Method to Determine Advanced Inverter Function Impact Step 1: Create thousands of potential Small-Scale PV deployments Step 2: Run PV from zero to full output to determine worst case response Voltage Magnitude Primary voltage nodes should be < 1.05 Vpu Secondary voltage nodes should be < 1.05 Vpu Voltage Deviation Primary voltage nodes should be < 0.03 Vpu Secondary voltage nodes should be < 0.05 Vpu Step 3: Repeat step 2 for all advanced inverter functions Step 4: Repeat steps 1-3 for Large-Scale PV deployments 7

8 Step 2: Calculate Feeder PV Hosting Capacity Region A No Issue Region B Issue dependent on PV deployment Region C Issue due to total PV on feeder Maximum Feeder Voltage (pu) Minimum Hosting Capacity Maximum Hosting Capacity Region A Region B Region C Total PV (kw) Primary voltage magnitude metric 8

9 Step 3: Advanced Inverter Control Absorbing reactive power at constant power factor Unity power factor (Base case, control off) 98% power factor 95% power factor Inverter voltage determines reactive power output Volt-Var 1 Volt-Var 2 9

10 Step 3: Advanced Inverter Control, cont. Inverter voltage determines active power output Volt-Watt 1 Volt-Watt Watts Generated (%) Local Voltage (pu) Watts Generated (%) Local Voltage (pu) Dynamic control of reactive power output based on voltage deviation 50 Reactive Current (%) Local Voltage Change (pu) 10

11 Results 11

12 Small-Scale PV with Maximum Load Primary Voltage Magnitude Dyn-Var and Volt-Watt 2 control helps too late Volt-Watt 1, both Volt-Var, and both pf controls reasonably help Volt-Var is a better control than respective pf controls at higher penetration levels Based on right amount of reactive power absorbed at the right location 95% pf appears the best option while voltages are below 1.05 Vpu Lines indicate average voltage from all PV deployments 12

13 Small-Scale PV with Maximum Load Voltage Magnitude Based Hosting Capacity Primary nodes benefit most from 95% pf Secondary nodes benefit most from Volt-Var 1 PV systems are directly controlling those nodes Volt-Watt 1 appears beneficial, but Primary Nodes Secondary Nodes - Maximum hosting capacity greater than 5.2 MW 13

14 Advanced Inverter Control on Feeder Head Demand Volt-Watt controls limit PV output All other controls only impact the reactive power demand Bulk power system to provide or local capacitor banks 95% pf and Volt-Var 1 require the most reactive power 14

15 Small-Scale PV Top Control Summary Max Load Min Load Voltage Deviation Voltage Magnitude Primary Nodes Secondary Nodes Primary Nodes Secondary Nodes Control Off Volt- Var 1 95% pf Control Off Volt- Var 1 95% pf Minimum Median Maximum Minimum Median Maximum Minimum Median Maximum Minimum Median Maximum

16 Small-Scale PV Overall Control Summary 16

17 Large-Scale PV Overall Control Summary - Maximum hosting capacity greater than 10 MW 17

18 Summary Volt-Var 1 and 95% pf provide the highest maximum and minimum hosting capacities Benefit from these controls requires significant reactive power sources Requiring significant reactive power can be negative Volt-Var 1 requires slightly less reactive power due to its optimal control curve Volt-Var 2 is the best option if reactive power sources are not available 18

19 Task 11: Backup Slides 19

20 Large-Scale PV Top Control Summary Max Load Min Load Voltage Deviation Voltage Magnitude Primary Nodes Secondary Nodes Primary Nodes Secondary Nodes Control Off Volt- Var 1 95% pf Control Off Volt- Var 1 95% pf Minimum Median Maximum Minimum Median Maximum Minimum Median Maximum Minimum Median Maximum

21 Task 15: Integrate Results of Field Monitoring with Cloud Prediction and Feeder Modeling 21

22 Task 15: Integrate Results of Field Monitoring with Cloud Prediction and Feeder Modeling Subtask 15.1: Continue feeder and PV system monitoring at existing sites (from Phase 2) and include measurement data and simulation results from another site currently being monitored in the Western US Subtask 15.2: Develop statistics from site data that describe variability at the distribution level Subtask 15.3: Compare simplified cloud modeling approach developed and implemented by UT Austin in Phase 2 with EPRI s method of utilizing multi-point sensors for representing cloud variability within a feeder Subtask 15.4: Evaluate feeder hosting capacity if the feeder were allowed to operate outside normal operating criteria for varying percentages of time 22

23 Subtask Feeder Monitoring and Variability of PV on Distribution Systems Chris Trueblood Steven Coley 23

24 Approach Feeders Considered for Analysis Feeder US Location Feeder Footprint (km 2 ) Single- Module PV Systems J Northeast Large Scale PV Monitored 1.7 MW (4 plants) K Southeast MW M Southwest (none) 24

25 Instrumentation for solar resource & PV output Data logger Up to 1-second recording period Onboard memory for temporary storage Outbound transfers push data to EPRI Internet time synchronization Power meter ac output Real energy & power Voltage, current Reactive energy & power PV module/array Solar irradiance, plane-of-array Back surface temperature Optional: dc voltage, current Instrumentation Enclosure Pyranometer 25

26 Quantifying Solar Variability Benefit How often and when significant ramping events occur AC Power (kw) Factors to Consider Rate Magnitude Frequency of Occurrence Scope Geographic Location PV Plant Size Spatial Diversity AC Power (kw) (Sep 29, 2011)... 12:00 15:00 18:00 X: 29-Sep :09:55 Y: kW in 25 sec (2.9% per sec) X: 29-Sep :10:20 Y: X: 29-Sep :11:40 Y: kW in 25 sec X: 29-Sep :11:15 Y: (3.0% per sec) (Sep 29, 2011)... 13:10 13:11 13:12 13:13 Local Date & Time (Eastern) 26

27 Categories for Daily Variability Conditions Sandia s variability index (VI) and clearness index (CI) to classify days VI > 10 Clear Sky POA Irradiance Measured POA Irradiance High VI < 2 CI 0.5 Clear Moderate 5 VI < 10 VI < 2 CI 0.5 Overcast Mild 2 VI < 5 27

28 Geographic Diversity of Solar Variability Daily Variability Conditions as a Percentage of Days per Season Variability Conditions: Tennessee Site Variability Conditions: New Jersey Site Percentage of Days (%) Jan-Mar Apr-Jun Jul-Sep Percentage of Days (%) Apr-Jun Jul-Sep Variability Conditions: Arizona Site High Moderate Mild Clear Overcast Percentage of Days (%) Jan-Mar Apr-Jun Jul-Sep Annual Insolation Latitude Tilt 28

29 Example Power Output PV Plant, Aggregation of Single Modules, and a Single Module Plant (1.4 MW) Feeder J Single Module Power (% of Rating) (Aug 11, 2012)... 09:00 12:00 15:00 18:00 Local Time 29

30 Example Power Output A Closer Look Single Module and PV Plant have larger ramps than the Feeder Aggregation Plant (1.4 MW) Feeder J Single Module Power (% of Rating) (Aug 11, 2012)... 12:05 12:10 12:15 12:20 12:25 12:30 Local Time 30

31 Instantaneous Changes in Power Looking at a single day for a Plant of Feeder K (1.0MW) on August 1, 2012 Normalized Change Normalized Power AC Power 09:00 12:00 15:00 18:00 Local Time 10-Second Interval 1-Minute Interval 09:00 12:00 15:00 18:00 Local Time 31

32 Histogram of Changes in Power Aggregation of 5 single-modules on Feeder M (Jul-Sep 2012), Daytime only 5 Relative Frequency (%) min 1 min 30 sec 10 sec 5 sec Change in Power (% of System Rating) 32

33 Percentiles of Changes in Power Aggregation of 5 single-module systems on Feeder M (Oct 2011-Sep 2012) 100 Ramp Rate Interval 5 sec 10 sec 30 sec 1 min 5 min Change in Power (% of Rating) th 99.9th 99th 95th 90th -100 FaWiSpSu Fa WiSpSu FaWiSpSu Fa WiSpSu FaWiSpSu Season 33

34 Percentiles of Change: Feeder K and Feeder M Using instantaneous method for 5 ramp rate intervals (Jul-Oct 2012) 100 Feeder J (12.2 km 2 ) th 99.9th % Change th 95th 90th.01% of ramps are 40% or less Feeder M (0.48 km 2 ) 80 % Change sec 10 sec 30 sec 1 min 5 min Ramp Rate Percentile 34

35 Effect of Spatial Diversity on Variability Results from Distributed PV Modules (0.2kW each) in NJ from Jul-Oct, 2012 Poly c-si, fixed 30 tilt Largest Ramps 10 sec Ramps 0.01% of Occurrences Area 0.1 km² 12 km² th 99.9th 99th 95th 90th Change in Power (% of Rating) Rate (%/sec) 10 Second Ramps 40% 26% 4%/sec 1 Minute Ramps 2.6%/sec Ramp Rate (% of Rated Output) min Ramps Change in Power (% of Rating) 61% 40% Rate (%/sec) 1%/sec 0.6%/sec Area (km 2 ) 35

36 Correlation Coefficient of Changes in Power Correlation calculated using differences in block averages of power, (Jul-Sep 2012) 1 Feeder J 0.8 Correlation Coefficient Feeder K 1 hour 30 min 10 min 5 min 1 min 30 sec 10 sec 5 sec Distance (km) 36

37 Effect of PV Plant Size on Variability Results from PV Systems in NJ from Jul-Oct 2012, Poly c-si, fixed 20 tilt Largest Ramps 0.1% of Occurrences 1.4MW PV Plant System Size 0.2 kw 1.4 MW Vs. 10 Second Ramps 1 Minute Ramps Change in Power (% of Rating) Rate (%/sec) Change in Power (% of Rating) 50% 24% 5%/sec 2%/sec 70% 55% Rate (%/sec) 1.2%/sec 0.9%/sec Change in Power (% of Rating) 0.2kW single module System Size (kw) 5 sec 10 sec 30 sec 1 min 5 min 37

38 Key Findings Of the data examined, 99.9% of large scale PV ramps were less than 65% of rated output at second time frames Solar variability throughout the feeder footprints examined (areas less than 12 km²) were uncorrelated at smaller time scales of less than 1- minute A slight reduction, 20% of rating, occurred in maximum PV ramping across increasing footprints (from ~0 to 12 km²) There is a modest reduction in ramping (20% of rating for 99.9 th percentile ramps) when comparing single-module PV systems to 1MW PV plants at small time intervals (5 to 10-seconds) 38

39 EPRI-Hosted Website for Data Access Website for parties to access data from monitoring sites Access will be open, but EPRI will screen data for integrity and erroneous values and may blacklist unacceptable subsets of data so that a clean dataset is available Users can specify a site or group of sites, select a date range, time resolution, and type of measurement Debug alpha release for testing available in June First full beta release planned in July 39

40 Subtask 15.4 Evaluating Hosting Capacity Beyond Thresholds 40

41 Task Statement of Work Evaluate feeder hosting capacity if the feeder were allowed to operate outside normal operating criteria for varying percentages of time. Using the longer term, 8-12 months, time-series site monitoring data, this task will involve calculating feeder hosting capacity for different cases where limits are exceeded for certain percentages of time and where some small number and/or short period of operation outside limits is allowed. For the feeder analyzed in Phase 2 where operating limits can be exceeded by high penetrations of PV, evaluate hosting capacity as a function of the time where operating limits are exceeded. 41

42 Feeder Hosting Capacity Results (from Phase II) Regulator Voltage Deviation Secondary Overvoltage Primary Over Voltage PV Hosting Capacity (kw) Max Load Solar Max Load Solar Min Load Min Load 1st violation % scenarios with violation All scenarios with violation st violation % scenarios with violation All scenarios with violation st violation % scenarios with violation All scenarios with violation Maximum Feeder Voltage (pu) Minimum Hosting Capacity Maximum Hosting Capacity Region A Region B Region C Total PV (kw) 42

43 Simulating 330 Days at 5 Second Resolution Simulations in OpenDSS 16X16 GRID 256 unique solar inputs x 5-second resolution x 330 days 1.5 BILLION, unique input points!!!! DPV Pole-Mount masurements OpenDSS 43

44 Feeder Overvoltages and Net Load - Weekly Percentages (Small-Scale PV) 3500 Weekly Average of Net Daily Feeder Peak kw and Percentage of Time in Over Voltage Condition Percentage of Time in OV 35% Feeder Net kw Net kw 30% 25% 20% 15% 10% 5% Percentage of Time in Over Voltage (>=1.05 pu) 0 0% Apr '12 May '12 Jun '12 Jul '12 Aug '12 Aug '12 Oct '12 Oct '12 Dec '12 Dec '12 Jan '13 Mar '13 44

45 Feeder Overvoltages and Weekly Total PV Energy Small-Scale PV Percentage of Time in OV Condition 35% 30% 25% 20% 15% 10% 5% 0% Apr '12 Percentage of Time in OV and Per Unit Energy Production, by Week May '12 Percentage Time in OV Per unit Energy Production (by week) 0 Jun '12 Jul '12 Aug '12 Sep '12 Oct '12 Nov '12 Dec '12 Jan '13 Feb '13 Mar ' Per Unit Energy Production (by Week) Normalized by Peak Week 45

46 Regulator Tap Counts and Average Solar Variability Index, by Week 1000 Regulator Tap Counts by Week and Average of Daily Variability Index (by Week) Reg Tap Counts (total by week) Average Variability Index (by week) 12 Regulator Total Tap Counts by Week Variability Index (Avg by Week) 0 Apr '12 May '12 Jun '12 Jul '12 Aug '12 Sep '12 Oct '12 Nov '12 Dec '12 Jan '13 Feb '

47 Line Regulator Tapping Weekly Totals 1400 Base Case Tap Counts and with 1.17 MW Small Scale PV by Week Weekly Tap Count (across 9 LTC/reg) Base Case Tap Count Small Scale PV Tap Count PV Tap Counts Difference, from Base 1.1 MW distributed evenly through feeder, slight reduction in regulator tap counts Apr '12 May '12 Jun '12 Jul '12 Aug '12 Aug '12 Oct '12 Oct '12 Dec '12 Dec '12 Jan '13 Mar '13 47

48 Capacitor Switching Operations Weekly Totals 20 Capacitor Operation Counts for Base Case and for Small Scale PV by Week Cap Count by wk for base case Cap Count by wk for small scale PV 1.1 MW distributed evenly through feeder, reduction in capacitor switching operations Apr '12 May '12 Jun '12 Jul '12 Aug '12 Aug '12 Oct '12 Oct '12 Dec '12 Dec '12 Jan '13 Mar '13 48

49 Is there a correlation between percentage of time operating outside of limits and the hosting capacity? 49

50 Results of Hosting Capacity versus Out-of- Limits Operations Cases Small-Scale PV Only Relationship between Small-Scale PV Hosting Level and Percent of Time in Over-Voltage Condition Percent of Time in Over Voltage Condition versus Peak PV Output (Small Scale PV Only) Maximum Feeder Voltage (pu) Minimum Hosting Capacity Maximum Hosting Capacity Region A Region B Region C Total PV (kw) Percent of Time in Over Voltage Condition 20% 18% 16% 14% 12% 10% 8% 6% 4% 2% Percent of Time in OV Violation Linear Regression Best Fit y = x R² = % Amount of Installed PV (Peak AC Output) Small Scale (MW) 50

51 Conclusion Correlation between solar PV and grid impacts Overvoltage conditions not well correlated with seasons of high solar output Regulator operations correlate very well with variability indices Direct linear relationship exists between increasing hosting capacity and time exceeding thresholds Additional findings Line regulator operations 1.1 MW distributed, decreased line regulator operations Higher, more concentrated PV increased operations Capacitor switching 1.1 MW distributed, decreased switching Time series analysis further confirmed that hosting capacity methodology accurately predicted overvoltage occurrences 51

52 Further Efforts Perform 8760 simulations of advanced inverter functions to evaluate impact on voltage Losses Reactive power consumption Evaluate additional feeders and performance with advanced inverter functions Further evaluate linear relationship between hosting capacity and % above threshold limits P o w er (% of Ratin g ) Evaluate hosting capacity approach modification to consider local weather patterns and feeder footprint 0-100% assumed presently 0-X% could yield more accurate results for voltage deviations at regulator timeframes Plant (1.4 MW) Feeder J Single Module -0.2 (Aug 11, 2012)... 09:00 12:00 15:00 18:00 Local Time 0-100% or 0-X% 52

53 Backup Slides 53

54 Subtask 15.4 Evaluating Hosting Capacity Beyond Thresholds 54

55 Test Circuit Feeder J1 Will look at the circuit in two configurations of PV for this subtask (plus base case): Small-scale PV with one deployment scenario, at three penetration levels Small-scale PV with one deployment scenario + 4 Large-scale PV, three penetration levels of small-scale PV Deployment scenario chosen from stochastic analysis as first scenario and penetration to cause over-voltage violation (V >=1.05 pu, or 105%) Base case, without any PV on circuit to compare other cases against base-line conditions as needed Total of seven cases 55

56 Input Data Collection and Preparation from EPRI DPV Monitoring Initiative EPRI DPV Monitoring Initiative is collecting data on Feeder J1: Eight locations along the circuit Small pole-top panels with pyranometer and metering package At 4 Large-scale PV locations Pyranometer and metering package Collecting irradiance, panel temperature, AC output power, voltage, and current at each location Data collected at one-second intervals 56

57 Time Step and Set of Poles Five-second time step was chosen as the simulation interval Balance volume of data with capturing variability aspects of PV Take into account simulation time required Chose to query five-second average irradiance and average panel temperature from 6 of 8 pole top monitors Two DPV monitoring sites have been out of communication for periods of time, so excluded from input data set Chose to query five-second average AC active and reactive power output from four large-scale PV plants, for use in three of the simulations 57

58 Irradiance and Temperature Inputs to OpenDSS PVSystem Model and Output Utilized irradiance, panel temperature, panel characteristics, inverter characteristics as input into the OpenDSS PVSystem model to obtain AC output power for each of the six locations Sample of correspondence between AC power output from PV monitoring and AC power from OpenDSS model for a day 58

59 Invalid and Missing Data Handling Collected data had some missing data points as well as invalid irradiance data points Invalid irradiance on one pole monitor only, for short periods of time For irradiance and temperature DPV data If at least one valid data point is available for a given five-second interval, then accepted data point for simulation If no valid data points available for a given five-second interval, then use last available valid data point 59

60 Invalid and Missing Data Handling Where missing/invalid contiguous data duration was less than or equal to one minute, we go with last valid data point Where missing/invalid contiguous data duration was longer than one minute, we average data from other poles on circuit Total of about 0.044% of the 34.2 million five-second intervals were found to be bad across the six pole-mounted monitoring locations, and were handled as indicated above For Large-scale PV plants, same criteria were used for bad data detection Missing/invalid data were left as gaps to be linearly interpolated from last known good data point to next known good data point by OpenDSS 60

61 Input Data Preparation from EPRI DPV Monitoring Initiative Interpolation performed in Matlab was used to derive PV generation shapes for small-scale PV located on the circuit Circuit divided into 16 x 16 grid (256 grid sections) Developed 256 generation shapes from original six pole-monitor locations Each shape contains roughly 5.7 million points, spaced at five-second intervals Roughly 1.5 billion PV generation data points OpenDSS for small-scale PV Roughly 45 million PV generation data points OpenDSS for large-scale PV Date range for input data is from evening of April 4, 2012, to near midnight on February 28, 2013 Approximately 330 days covered in input data and simulated in OpenDSS for each case 61

62 Modeling and Simulation Seven cases were simulated, for 330 day range, at 5-second intervals: Base-case: No PV (one case) Small-scale PV Only (three cases) Three penetration levels corresponding to minimum hosting capacity from prior analysis on circuit, maximum hosting capacity from analysis, mid-point between minimum and maximum hosting capacity Small-scale PV + Existing Large-scale PV (three cases) Three penetration levels corresponding to minimum hosting capacity from prior analysis on circuit, maximum hosting capacity from analysis, mid-point between minimum and maximum hosting capacity 62

63 Results of Hosting Capacity versus Out-of- Limits Operations Used percentage of time on a weekly basis in over-voltage condition as the measure of out-of-limits operation Percentage of time in over-voltage condition is a percentage of total hours (daylight and night-time) Some over-voltages seen at night time during base case, so decided to use total number of hours, rather than only dawn to dusk For the base case about 0.006% of the time was spent in over-voltage condition out of total number of hours Is there a correlation between percentage of time operating outside of limits and the hosting capacity? 63

64 Modeling and Simulation Run-Times Using OpenDSS 64-bit On a Lenovo Thinkpad W520, with Intel Core i7 processor Approximately 26 to 28 hours to run 5.7 million time-steps per case Ran 5.7 million steps in 8 groups of about 715,000 steps each Peak memory usage was about 9 GB per group of 715,000 steps Large memory usage due to load-shapes, followed by in-memory monitors (fast) Other data written immediately to disk after each step (slow) 32 GB of physical RAM allowed running two cases in parallel Each OpenDSS instance uses one core on i7 processor Post-processing of data mainly performed in python (64-bit) and Microsoft Excel 2007 for stability, speed, and ease-of-use 64

65 Modeling and Simulation Three Penetration Levels for Small-Scale PV + Large-Scale PV Minimum Penetration 2.33 MW Middle Penetration 2.66 MW Maximum Penetration 2.96 MW 65

66 Results of Hosting Capacity versus Out-of- Limits Operations Cases Small-Scale Plus Large-Scale PV Relationship between Small-Scale+Large-Scale PV Hosting Level and Percent of Time in Over-Voltage Condition 20% Percent of Time in Over Voltage Condition versus Peak PV Output (Small Scale+Large Scale PV) Percent of Time in Over Voltage Condition 18% 16% 14% 12% 10% 8% 6% 4% 2% 0% Percent of Time in OV Violation Linear Regression Best Fit y = x R² = Amount of Installed PV (Peak AC Output) Small Scale+Large Scale (MW) 66

67 Comparison of Each Deployment 20% Percentage of Time At or Above 1.05 Per Unit Voltage Small Scale PV Only and Small Scale+Large Scale PV Percentage of Total Intervals At or Above 1.05 Per Unit Voltage 18% 16% 14% 12% 10% 8% 6% 4% 2% y = x R² = y = x R² = Small Scale+Large Scale Small Scale Only 0% MW of Peak PV Output During Simulation 67

68 Results of Hosting Capacity versus Out-of- Limits Operations Discussion/Conclusion For the range of hosting capacities reviewed, there is a linear relationship between installed PV and the percentage of time in out-oflimits operation A range of 10% variation in PV output (in terms of the peak loading) was simulated covering the minimum and maximum hosting capacity, and one additional level in between The deployment of PV (size and location) has a marked effect on the relationship between installed PV and time in out-of-limits operation A single line of best fit can not be derived for different deployments of PV It is size and location dependent, as well as circuit dependent Relationship may be non-linear if installed PV level is substantially above the maximum hosting level, or substantially below the minimum hosting level 68

69 Backup Slides Unless otherwise noted, these slides pertain to the Scenario #2 Maximum Penetration Small-Scale PV case Simulated at five second intervals (as are all cases) Intent is to validate simulations as well as to better understand the simulation results Variability and Clearness Indices and Day Classifications come from code developed in prior work performed by C. Trueblood and S. Coley with respect to variability analysis from the DPV Monitoring Project and collaboration with SANDIA National Labs Data analysis mainly performed with CPython 2.7.3, 64-bit; pandas high performance BSD-licensed data analysis library Plotting typically performed with Microsoft Excel from output from python-based analysis 69

70 Percentage of Time in OV Condition by Week Small-Scale PV (1.17 MW Peak Output) 70

71 Percentage of Buses in OV Condition by Week Small-Scale PV 71

72 Percentage of Time in OV Condition by Week versus Regulator Tap Counts Small-Scale PV 800 Percentage of Time in Over Voltage Condition versus Regulator Tap Counts by Week 35% 700 Percentage of Time in OV Weekly Regulators Tap Count 30% Total Regulators' Tap Counts by Week % 20% 15% 10% Percentage of Time in OV Condition 100 5% 0 Apr '12 May '12 Jun '12 Jul '12 Aug '12 Sep '12 Oct '12 Nov '12 Dec '12 Jan '13 Feb '13 Mar '13 0% 72

73 Feeder Overvoltages and PV Peak Power- Weekly Percentages (Small-Scale PV) 35% 30% Percentage of Time in OV and Per Unit Power Injected by Small Scale PV, by Week Percentage Time in OV Average Daily Peak PVSystem Power in Per Unit (by week) Percentage of Time in OV Condition 25% 20% 15% 10% Per Unit Power Injected, Weekly Avg of Daily Peak 5% 0.5 0% Apr '12 May ' Jun '12 Jul '12 Aug '12 Sep '12 Oct '12 Nov '12 Dec '12 Jan '13 Feb '13 Mar '13 73

74 Percentage of Time in OV Condition by Week versus Average Variability Index Small-Scale PV 35% Percentage of Time in OV and Average Variability Index, by Week Percentage Time in OV 14 30% Average Variability Index 12 Percentage of Time in OV Condition 25% 20% 15% 10% 5% Average Daily Variability Index, by Week 0% Apr '12 May '12 0 Jun '12 Jul '12 Aug '12 Sep '12 Oct '12 Nov '12 Dec '12 Jan '13 Feb '13 Mar '13 74

75 Percentage of Time in OV Condition by Week versus Average Clearness Index Small-Scale PV 35% Percentage of Time in OV and Average Clearness Index, by Week 1.2 Percentage of Time in OV Condition 30% 25% 20% 15% 10% 5% Percentage Time in OV Average Clearness Index Daily Clearness Index, Averaged by Week 0% Apr '12 May '12 0 Jun '12 Jul '12 Aug '12 Sep '12 Oct '12 Nov '12 Dec '12 Jan '13 Feb '13 Mar '13 75

76 Percentage of Time in OV Condition by Week versus Average Day Classification Index Small- Scale PV 76

77 Regulator Tap Counts Base Case versus Small-Scale+Large-Scale PV, by Week 2000 Base Case and 2.96 MW Small Scale+Large Scale PV Tap Counts by Week Base Case Tap Count Weekly Tap Count (across 9 LTC/reg) Small Scale+Large Scale PV Tap Count PV Tap Counts Difference, from Base 500 Apr '12 May '12 Jun '12 Jul '12 Aug '12 Aug '12 Oct '12 Oct '12 Dec '12 Dec '12 Jan '13 Mar '13 77

78 Together Shaping the Future of Electricity 78