Analysis and Reporting of I-V Curve Data from Large PV Arrays

Transcription

1 Analysis and Reporting of I-V Curve Data from Large PV Arrays Solmetric Webinar December 12, 2013 Paul Hernday Senior Applications Engineer cell

2 Goals of I-V Curve Data Analysis External: Satisfy the terms of the contract & delight your client Internal: Develop core competencies around PV array performance To achieve these goals we need to: 1. Identify deviations caused by measurement technique, weather, and obstructions to sunlight 2. Provide clues for troubleshooting any issues

3 The Work of Data Analysis Reveal the real hardware performance as closely as possible PV hardware performance Follow-up may be needed for any of these categories. Population of measurement results Obstructions Shade, soiling, debris Weather conditions Low and/or variable irradiance Rapidly changing irrad or temp etc Measurement technique Irradiance sensor not in POA Thermocouple not attached etc In some cases re-testing, or repair and retesting, may be needed In other cases a physical explanation may be sufficient (for example, inter-row shading from measurements early in the day). In some cases the follow-up may be to simply watch the performance for potential degradation over time.

4 Topics PV Analyzer operation PV principles useful for data analysis How to use the I-V Data Analysis Tool Data analysis examples Summarizing your results Measurement tips

5 PVA1000 PV Analyzer & SolSensor Electronic tool (I-V curve tracer and sensors) for PV source performance measurements up to 1000V, 20A Compares results to built-in models Powerful PC-based user interface All wireless

6 SolSensor Wireless PV Reference Sensor Replaces the Wireless Sensor Kit as the curve tracer s companion sensor unit Measures irradiance, temperature, tilt Integrated design (sensors, rechargeable battery, common wireless unit) Uses same wireless USB adapter as the I-V measurement unit Self-aligns in plane of array. Clamp provided.

7 How It Works Irradiance (in plane of array) Temperature (on module backside) Module make & model Tilt Azimuth Irradiance Module temperature Latitude Longitude Date & time Built-in PV models 3 dots predict curve shape All wireless PV Module or string

8 PVA-1000 Instrument Block Diagram (simplified) Battery charging connector Control button with LED indicator Controller & Wireless C (1 of 3) I sense V sense PV Test Leads NEMA 4X FG Enclosure The load is a capacitor that charges from 0V to Voc Initial current is close to Isc, and final current is close to 0A The PVA software PVA automatically selects the capacitance and other circuit settings The selection is based on the first I-V curve measured

9 Current The Measured I-V Curve Straight from the curve tracer Isc Actual I-V curve. No adjustments for irradiance or temperature. Not affected by your performance model. Voltage Voc

10 Current The Predicted I-V Curve Modeled, including effects of irradiance & temperature Isc Imp, Vmp Expected I-V curve shape, based on the design details and the present irradiance and temperature. Voltage Voc

11 Current Measurement and Prediction Isc Imp, Vmp They are compared on-screen. This is why accurate irradiance and temperature measurements are important. Voltage Voc

12 Performance Factor = Imp x Vmp from the measured I-V curve PF = Performance Factor = Pmax (measured) Pmax (predicted) = Imp x Vmp from the predictive PV model If the measured curve passes through the predicted (Imp, Vmp), the Performance Factor is 100% Errors in the model inputs can push Performance Factor up or down

13 Typical Measurement Setup Courtesy of Chevron Energy Solutions 2011

14 Typical Measurement Setup

15 Test Process Example: Measuring strings at a combiner box Hardware setup (do once at each combiner box): 1. Move the sensors (if necessary to get wireless range) 2. Isolate the combiner box (open the DC disconnect) 3. De-energize the buss bars (lift the string fuses) 4. Clip test leads to the buss bars Electrical measurement (repeat for each string): 1. Insert a string fuse 2. Press Measure 3. View and save results 4. Lift the fuse 15 seconds, typically

16 Saving a Measurement 1 2

17 Live Demo of taking and saving a measurement

18 The Project File These 2 folders appear only if PVA software version 1.x was once installed on your PC xxxxxx.pvapx (v3.x) xxxxxx.pvap (v2.x) Contains array description, PV performance model, and I-V measurement data Easy to share between offices, and with Solmetric for support purposes Allows development of advanced analysis and reporting capabilities in the future

19 Exporting I-V Curve Data

20 Exported Data in Folder Tree This tree is imported into the I-V Data Analysis Tool for analysis The DAT can be directed to any level of the tree For very large systems, there are tradeoffs between analyzing data at the system vs. inverter levels.

21 Topics PV Analyzer operation PV principles useful for data analysis How to use the I-V Data Analysis Tool Data analysis examples Summarizing your results Measurement tips

22 Solar Cell Physics (shown in cross-section) Photon 1 2 Charge generation Charge collection Fingers 2 1 Semiconductor top layer Random drift Semiconductor bottom layer Backside metal A performance problem in one cell affects the operating point of cells in series with it.

23 Current Power I-V and P-V* Curves Expect this shape for healthy cells, modules, strings, arrays Isc Imp I-V curve Pmax P-V curve Voltage Vmp Voc *P-V curve is calculated from the measured I-V curve

24 Current (A) Irradiance Effects Conventional crystalline silicon module W/m 2 Isc doubles when irradiance doubles, but this rule does not apply at all points along the curve Below 400 W/m 2, and especially below 200, cell voltages drop rapidly. Low-light measurements do not accurately predict performance at high irradiance! That s true of ANY performance testing method, not just curve tracing Voltage (V)

25 Current (A) Temperature Effects Conventional crystalline silicon module 9 Vmp and Voc drop %/C. 8 7 Smaller effect for irradiance, but still important C The PV model accounts for these temperature effects The modeling is more accurate if the temperature measurement is accurate Temperature affects voltage more strongly than the current Voltage (V)

26 Current Stringing Modules for Higher Voltage Voltages add at each level of current. Location of building blocks in I-V graph does not correspond to locations of modules in the string! I-V Building Blocks Series Voltage

27 Current Paralleling Modules for Higher Current Currents add at each level of voltage. Location of building blocks in I-V graph does not correspond to locations of modules in the string! I-V Building Blocks Voltage

28 Current Series/Parallel Combination Array I-V curve will be smooth and regular if the modules have matching currents and voltages. Location of building blocks in I-V graph does not correspond to locations of modules in array. If we solidly shade any module in the array, we lose the upper right building block and the total I-V curve will have a step in its place. More on this topic later. Voltage

29 Current Square-ness of the I-V Curve Isc Increased square-ness means increased Pmax An important figure of merit of a PV source is the square-ness of its I-V curve. Squarer means higher Pmax for a given Isc and Voc. In an ideal world, the curve would be perfectly square and output power would be Isc x Voc. But this is not physically possible. Voltage Voc

30 Current Fill Factor An important figure of merit for PV source performance Isc Imp Current ratio Imp/Isc Max Power Point Voltage ratio Vmp/Voc Voltage Vmp Voc Area of green rectangle Fill Factor = = Area of blue rectangle Imp x Vmp (watts) Isc x Voc (watts) For xsi, the Fill Factor is normally in the 0.7 to 0.8 range.

31 Fill Factors of Field-Aged PV Modules Currents and voltages have been normalized to max values of 1.0 Blue I-V curves are from new PV modules. Red curves are from 20-year fieldaged PV modules. Fill Factor is an important troubleshooting tool because: Higher Fill Factor means higher output power Normalized I-V curves of field tested devices Fill factor dependence of bilinear I V curve translation accuracy, IEEE, Skoczek, 2008.pdf Fill Factor is less dependent on irradiance than Pmax or Isc, so it allows us to compare string measurements even if irradiance is varying

32 Deviations from Normal I-V Curve Each will be explained later in the webinar Conventional measurements do not reveal many of these effects.

33 Power Current I-V Curve of a Partially Shaded String Isc Shade causes steps in the I-V curve, in turn causing multiple peaks in the P-V curve. Changing conditions cause peaks to move around Inverter s job is to minimize disruption to the whole system. Bypass diode turns on here. This is the point at which the weak (or shaded) has reached it s maximum possible current. Voltage Voc

34 The Purpose of Bypass Diodes The primary purpose of bypass diodes is to prevent damage to the PV module and supporting structure In doing so, they also mitigate the impact of shade on energy production.

35 Temperature (C) Rapid Heating of an Unprotected Cell 85 Heating of a Single Shaded Cell Backside Temperature vs. Time Backside temperature with no bypass diodes Unshaded cell Shaded cell All 3 BP diodes removed Single cell shaded Output short circuited Backside thermocouples Shaded cell Unshaded cell Shade Duration (seconds)

36 Inverter dc input voltage Understanding Why Cells Need Protection A thought experiment 360v In the next slide we ll shade this one cell Imagine 4 strings of cell modules, or 720 cells per string. The inverter is operating at the array max power point. All the cells are matched in current and voltage and are operating at 5A and 0.5V. What happens if we shade a cell? Typical cell operating point: 5A, 0.5V Pmax Inverter (operating at Pmax) Pac 0v String 1 String 2 String 3 String 4

37 Inverter dc input voltage Understanding Why Cells Need Protection A thought experiment + 1. Shading the cell limits its current 2. All cells in series rise in voltage 360v Shaded cell becomes a load and sees enough reverse voltage to break down (avalanche) 4. Cell heats up and destroys module 5. BP diodes clamp the reverse voltage to a lower level, preventing reverse breakdown. 3A, 0.52v Inverter (operating at Pmax) Pac 0v String 1 String 2 String 3 String 4

38 Current Flow in Normal Operation Cell group Cell group Bypass Diodes Cell group + Bypass diode turns on when the shaded cell(s) can no longer pass as much current as the non-shaded cells.

39 Current Flow with One Shaded Cell Cell group Cell group Bypass Diodes Cell group + Bypass diodes start turning on when the shaded cell(s) can no longer pass as much current as the non-shaded cells.

40 Which shading hurts output the most? 1 2 +

41 Which soiling hurts output the most? 1 2 +

42 Rules of Thumb For shading or other current mismatch effects A bypass diode turns on when the most shaded cell in its cell group can no longer keep up with the rest of the module or string. Bypass diodes typically start turning on when the most shaded cell is about 20-30% shaded. The depth of the current step in the I-V curve tells us how heavily the most shaded cell is shaded. The width of the current step(s) in the I-V curve tells us about extent of the shading (or soiling, or debris) The location of the current step in the I-V curve does not tell us where the shading is located in the string under test. We always see their effects toward the right side of the curve, with the deepest steps at the far right.

43 Current Array With Full-Shaded Module The shaded module s bypass diodes turn on, removing its voltage (and power) from production. The result is seen in the upper right building block, causing a step in the I-V curve. Voltage

44 Current Array With Full-Shaded Cell Group The narrowest steps occur when a single cell group is shaded or its bypass diode fails short. In this example, we shade one of a modules three cell groups. Series Voltage

45 Current String with Shaded Cell Group The height of the step is related to the shading factor of the most shaded cell in the cell group. In this example, we shade one entire module with 33% shade cloth, reducing the irradiance to 2/3 of the level seen by the rest of the array. Series Voltage

46 Current (A) Visualizing Current Mismatch Losses Isc Shaded cell groups Unshaded cell group Normal I-V curve Expected max power point X X When a cell group can no longer sustain the current of the other cell groups, its bypass diode turns on (X s), allowing string current to increase. This gives the mismatch I-V curve its roller coaster shape. Voltage (V) Voc

47 Topics PV Analyzer operation PV principles useful for data analysis How to use the I-V Data Analysis Tool Data analysis examples Summarizing your results Measurement tips

48 I-V Data Analaysis Tool Home tab

49 1 I-V Data Analaysis Tool Home tab ( Front panel )

50 Choosing the PVA and Sensor Sources Before importing data to the DAT

51 Frequency Current (Amps) Analysis Tools Automatically generated String Table I-V Graphs (overlay) Voltage (Volts) Histograms 4 3 Three powerful, complementary tools Use them in combination Any of the three can be your starting point Pmax (Watts)

52 String Table Limits (user settable) Statistics (per column) Parameter values (per string)

53 Count Histograms Graphical displays of how data values are distributed Example: Histogram of 99 data values Bin (or bucket ) (10 wide in this example) Always whole numbers Value

54 Common Histogram Shapes Normal Fill Factor of healthy PV strings Right-skewed Left-skewed Double-peak Voc measured on cold morning and hot afternoon Plateau Isc values measured over a partially cloudy day

55 Outliers Any type of distribution can have outliers. These are points that just don t fit the distribution of the rest of the population of points. Here s an example of low-side and high-side outliers for the Normal distribution:

56 Standard Deviation The Standard Deviation describes the variability within a population of data, in other words, how widely is it spread out % 95.44% 99.74% -3 sd -2 sd -1 sd Mean or average value +1 sd +2 sd +3 sd

57 Data Within 1 Standard Deviation 68.26% -3 sd -2 sd -1 sd Mean or average value +1 sd +2 sd +3 sd

58 Data Within 2 Standard Deviations 95.44% -3 sd -2 sd -1 sd Mean or average value +1 sd +2 sd +3 sd

59 Data Within 3 Standard Deviations 99.74% -3 sd -2 sd -1 sd Mean or average value +1 sd +2 sd +3 sd

60 Data Analysis Process

61 Overview of Data Analysis 1. Export data from PVA software. This exports the most recent measurement for each location in the array tree. 2. Open the Data Analysis Tool 3. Use the DAT controls to import and crunch the numbers: 1. Import data 2. Create string table & histograms 3. Create Measured vs. Modeled table 4. Plot I-V curves (overlaid at the combiner level) 5. Translate Isc, Imp, Vmp, Voc to STC if desired 4. Review and interpret data 5. Generate punch list if needed 6. Generate report 7. Print report or save as a pdf file

62 Starting Points for Data Analysis The starting point for your analysis is a matter of personal preference, but if you like your information in graphical form, this is a good flow. I-V Curve Graphs Scan for outliers and identify those strings (hover with cursor) Histograms Scan for outliers and odd shapes Correlate shapes with variability of irradiance and temperature Table Check the statistics (rows 5-9) Enter limit values (blue fields) to identify outliers (shaded yellow)

63 Topics PV Analyzer operation PV principles useful for data analysis How to use the I-V Data Analysis Tool Data analysis examples Summarizing your results Measurement tips

64 Deviations from Normal I-V Curve Conventional measurements do not reveal many of these effects.

65 Understanding Variation in Voc In this set of curves from a combiner box, the shapes and levels are very consistent. Most likely, the irradiance and temperature were stable throughout and the strings were quite uniform.

66 Understanding Variation in Voc In this set of curves from another combiner box, the shapes are mostly consistent but the voltages are slightly spread - why? Here are several possibilities: 1.Strings are slightly mismatched in voltage 2.Temperature is rapidly changing due to wind or shifting clouds 3.The strings don t all get the same amount of ventilation behind the modules. 4.Voc changes at low irradiance, but that doesn t fit this situation.

67 Steps in the I-V Curve Typically caused by shade, soiling, debris, snow, or cracked cells 350 Clark i1c3 The small steps represent shaded cell groups within modules. The width of the step tells us how many cell groups are involved. The height of the step tells us about the extent of shading on the most shaded cell in the group; lower amps means it s more shaded. We can t tell from the I-V curve where the shaded cell groups are located in the string. Record the string ID (for example i3c4s7) for the punch list and/or report.

68 Partially shaded residential array Measured the lower string

69 Partially shaded residential array Measured the lower string Approximately 40% reduction in string s output power

70 Hockey Sticks BBB Florence InvA Phase 1 Hockey sticks often represent systematic shading over several adjacent cell groups or modules. In this case, the low current value of the hockey stick steps suggests that at least one cell in each of the cell groups is almost completely shaded. This type of pattern is unlikely to be caused by soiling or scattered shade because of the extent and uniformity of the obstruction and the fact that it happens on only a few of the strings.

71 Increased Slope in Horizontal Leg 350 Clark i2c3 Typically caused by tapered shading or tapered soiling. For a uniform slope, each cell group must be obstructed to a slightly different extent. Often slight steps will remain. Common causes are inter-row shading early or late in the day, or dirt dams that get progressively wider across a string of modules in portrait mode. Electrical shunts can cause slopes, but it s much less common. PID can also cause the slope, and may be accompanied by low Voc.

72 Fill Factor Representation of steps and slopes in the curve 350 Clark i2c3 The stepped and sloped I-V curves are represented as lowside outliers in the Fill Factor histogram. Fill Factor is a good diagnostic tool because it is not strongly affected by level of irradiance.

73 Random Non-uniform Soiling Seagull example Effect similar to partial shading Steps in the I-V curve Smallest steps correspond to individual cell groups

74 Light Snow Cover on Array

75 Heavier Snow Cover on Array

76 Potential Induced Degradation South string, west modules Fill Factor Histogram PID is driven by high voltage stress. It s more likely to occur at higher voltages and negative polarity, and in modules with less effective encapsulation. Electro-corrosion type is not reversible. Symptoms include reduced Voc and Fill Factor (more rounded knee). Can be seen at string or module levels.

77 Dropped Cell Groups FW Solar Field Voc Histogram If Voc is shifted downward by approximately a module Voc/N it may indicate a dropped cell group, likely caused by a shorted bypass diode. In this example at least two strings are likely to have one or more dropped cell groups. Validate dropped cell group by comparing the apparent Voc in the I-V curve with the true Voc value in the Table tab. Full shading of a PV cell causes a similar looking left-shift, but a tail is usually present where curve approaches x-axis.

78 Low Voc vs. Last Point Effect 350 Clark i2c2 The blue and orange traces (s12,13) do not reach all the way down to the x-axis. This is because the 100 I-V samples were all taken and the measurement stopped. This sometimes happens when ther is a low current tail on the curve. s12 s14 Voc s13 Voc 513 s11 Voc 498 Others (Avg) Voc 510 Note: the Voc measurement is performed before the I-V curve measurement, so this last point effect causes no harm. The green trace (s11) reaches the x- axis, but does so 12 volts lower than the average of the other strings. This is likely caused by a shorted bypass diode.

79 Rounding of the Knee Degraded fill factor, lower output power This corresponds to a lower than expected value for Fill Factor (less square curve)

80 Current - A Increased Series Resistance Reduced slope in vertical leg of curve Failed module Neighboring strings String 4B14 String 4B Voltage - V

81 Intermittent PV Source Circuit Example: Improper splicing technique in home run conductors An intermittent connection in the PV source circuit can cause the current to jump to 0A one or more times during the I-V sweep. This can also cause the curve tracer to use up the alloted trace points before reaching zero current (x-axis).

82 Low Current Due to Edge Soiling Dirt dams, common in low-tilt arrays Dirty The dirt dam hurt performance as much as all of the uniform soiling. Clean Clean 50% 50%

83 350 Clark i3 Strongly Irradiance-Dependent Parameters These tend to have irradiance-like distributions unless blurred by other issues Irradiance Isc Imp Pmax Histograms of the same population of measurements

84 Less Irradiance-Dependent Parameters (At high light levels. At low light levels their dependence increases.) 350 Clark i3 Irradiance Fill Factor Shade effects Voc Performance Factor Shade effects Shade effects Histograms of the same population of measurements

85 Creating Custom Graphs Easiest to do in the Table worksheet

86 Limitations of STC Translation Not unique to curve tracing! Translation is less accurate at low irradiance, especially <400W/m 2. (This is mitigated by the PVA-1000 with SolSensor because the model has lowlight corrections for many PV modules). If irradiance is unstable, there will be more ± scatter in the translation. (This is minimized by the PVA-1000 with SolSensor, because the two are triggered simultaneously.) Worst case is when irradiance is low and unstable Measured temperature may not track the strings under test (inconsistent placement of thermocouples, sub-arrays with different temperature profiles. Wind.

87 Live Demo of the I-V Data Analysis Tool (if time allows)

88 Topics PV Analyzer operation PV principles useful for data analysis How to use the I-V Data Analysis Tool Data analysis examples Summarizing your results Measurement tips

89 Creating a Summary of your data analysis in MS Excel Select Deviation and Follow-up items from drop-down lists, or enter your own text Data filtering allows sorting for particular cases Can send the worksheet to a printer or PDF file

90 Drop-down Lists Can be edited

91 Summarizing the Histograms in MS Excel

92 Creating a Summary of your data analysis in MS Word I-V Curve Graph Worksheets Inverter Combiner String(s) Comment Histogram Worksheets Histogram PF FF etc Comments Measured vs. Modeled Worksheet Parameter Isc Voc Imp Vmp Comments Table Worksheet Enter comments here Translated Parameters Table (In Table Worksheet) Enter comments here, if applicable Conclusion Enter comments here

93 Topics PV Analyzer operation PV principles useful for data analysis How to use the I-V Data Analysis Tool Data analysis examples Summarizing your results Measurement tips

94 Top 10 Measurement Mistakes (Not unique to curve tracing!) 1. Setting the irradiance sensor on a module under test, shading it 2. Not orienting the irradiance sensor in the plane of the array 3. Mounting the irradiance sensor in shade or strong reflections, or in diffuse light making measurements with a large portion of the sky blocked by trees. 4. Placing the thermocouple at the (cool) edges of the module or array 5. No firmly attaching the thermocouple in contact with module backside 6. Deploying the temperature sensor inconsistently across the array 7. Inadvertently measuring two strings in parallel, and saving results as a single string 8. Incorrectly setting the local time, time zone, and daylight savings status 9. Selecting SmartTemp and forgetting to deploy a backside thermocouple 10. Taking measurements at very low irradiance 11. Not noticing the first trace effect and saving the trace anyway rather than re-taking the measurement.

95 Time Zone Considerations Setting up to make measurements The PVA software date/time stamps each measurement. The date and time are used in the model to predict the values of the Performance Factor and other key parameters. Before measuring, be sure your PC is set to the correct local date, time, time zone, and DS status. Exporting Project data Before exporting Project data from PVA software 2.x or 3.0, set your PC s UTC/GMT offset to the value that was used when the measurements were actually taken. Starting with v3.1, you will not need to fake your time zone before exporting data.

96 Time Zone Considerations UTC/GMT Offset (hours) Pacific time Mountain time Central time Eastern time DST off DST on Check to look up the time zone and Daylight Savings details for your site.

97 The First Trace Effect The PVA uses it to optimize internal settings The PVA software uses the first trace to learn the voltage and current characteristics of the PV source. The PVA then selects internal circuit settings to optimize the measurement of that type of device. If you get a first trace that has long straight line segments, that s the learning trace. Just take the measurement over. All subsequent measurements will use those internal settings. If the type of device you are measuring changes in mid-session, you may see another trace like that, and need to retake that measurement too.

98 Recommended Conditions For Array Performance Testing High irradiance Ideally more than 800 W/m 2, and not lower than 400 W/m 2. The I-V curve changes shape at low light, making it a less useful predictor of performance at high irradiance. 4-5 hour window centered on solar noon For good irradiance level and reduced angle of incidence effects Low or no wind For more consistent module temperature measurements Width of I-V curve varies inversely with temperature Good conditions more meaningful data

99 Effects of Unstable Irradiance & Temp. Unstable irradiance introduces scatter in the predicted performance values, especially if there is a time delay between I-V and irradiance measurements. The greater the time delay, the greater the scatter. The steeper the irradiance ramp, the greater the scatter. The same thing happens with temperature measurement, but to lesser degree because temperature ramping is slower, and the dependence of performance on temperature is weaker.

100 Be Aware of Temperature Uncertainties It would be nice if: All of the strings had the same temperature profile. The thermocouple was mounted at a location that sampled the average temperature of a typical string. There was no wind. Real world: The strings have different temperature profiles. The thermocouple is mounted closer to the edge of the array, where the modules are cooler. There is a variable wind.

101 Consistency of Thermocouple Location Choose a good location and repeat it on each sub-array Photo courtesy of Sun Lion Energy Systems

102 Why Locate the Thermocouple Consistently? PV array performance is always evaluated with respect to a model For example: Performance Factor (the predicted value of Pmax is based on temperature) Translating to STC (uses difference between measured temperature and 25C) If you locate your thermocouples consistently, you ll tend to see less variation in your Performance Factor or STC translation, and your distributions will tend to look like: If you locate your thermocouples inconsistently, you ll tend to see more variation in your Performance Factor or STC translation, and your distributions will tend to look like: or

103 Temperature Profile Flush Mounted Array

104 Voc Profile Flush Mounted Array Roof String Voc

105 Mounting the Temperature Sensor Mount the thermocouple 2/3 of the way between the corner and center of a module. Use high-temperature tape (eg 1-3/4 inch Kapton dots**). Press TC into contact with backside. ** MOCAP MCD-PE 1.75 poly dot ~$80/roll of 1000 dots customerservice@mocap.com

106 Analysis and Reporting of I-V Curve Data from Large PV Arrays Solmetric Webinar December 12, 2013 Paul Hernday Senior Applications Engineer cell