Data Modelling and Aerial AnalsyisSeeking Synergies

Rob Andrews, PhD

CEO, Heliolytics

rob.andrews@heliolytics.com

Pramod Krishnani, M.S. [IEC TC – 82 PV Expert]

Manager, Performance Engineering sPower

pkrishnani@spower.com

Presentation goals

• Introduction of Aerial infrared imaging

• Compare Aerial infrared imaging to:

– Data analysis

– Manual inspection techniques

• Recommend optimal DC O&M scopes

• ROI considerations

• Provides aerial Module Level Thermal Audits (MLTA)

• C&I and Utility scale systems

• Commercial operation October 2015

• >1GW scanned to date

• Proprietary technology to collect and analyze high resolution infrared and visible imagery

• Based in Toronto, ON, operates internationally

MLTA- Data collection

Outputs

Comparison of IR data to Combiner data

• Key Question:

– Can module and string level defects be detected with combiner data alone?

• Approach:

– Generate multiple methodologies for analyzing system data and compare these outputs

to aggregated failures per combiner

Raw Data

Spatially Resolved Combiner-level data

Spatially Resolved Combiner-level data

Absolute performance Relative performance

Spatially Resolved Combiner-level data

9:15 am 12:15 pm 4:45 pm

Evaluation Methodologies

• Methodology #1

– Calculate sum of energy produced with the following restrictions:

• Non-clipping

• Within a 1 month window around the scan

• Normalized by Isc of strings attached to combiner

• Methodology #2

– Calculate Normalized DC Performance Index (DCPI) with the following restrictions:

• Non-clipping

• Curtailment data points removed

• Within identified clear sky day

• All data set removed below 200 W/m2 POA Irradiance

• Uncertain DCPI calculation removedDCPI =

𝑰𝒎𝒆𝒂𝒔𝒖𝒓𝒆𝒅𝑮𝑷𝑶𝑨

𝑰𝒏𝒐𝒎𝒊𝒏𝒂𝒍 (𝒂𝒔 𝒑𝒆𝒓 𝒅𝒆𝒔𝒊𝒈𝒏)

𝑮𝑺𝑻𝑪(Courtesy of formula Improvement by Tassos Golnas, PhD)

(1)

(1)

Spatially resolved performance index

Methodology #1

Performance Index

Methodology #2

Methodology #1 Results

Methodology #2 results

Conclusions from data analysis

There is too much uncertainty in the basic combiner-level data to detect

individual module faults

Manual Fault detection

• Traditionally, multimeter or I-V checks

at a combiner level are used to detect

faults at a combiner level

Manual Fault detection results

• Site was inspected using combiner box checks of Voc and Impp

– September 2015- 6 months prior to site inspection

• Hit rate for manual inspection: 22%

DC measurement campaign goals

• Find Broken modules/Strings

– Partial system checks can leave parts of the system offline for multiple years

– Manual techniques regularly miss DC outages (on the order of 0.5%-0.7% energy loss)

• Detect degredation

– Bulk I-V or other manual techniques have structural limitations to their accuracy

Increasing ‘efficiency’

Ideal DC measurement campaign

• Periodic MLTA to detect DC availability loss– Yearly scans if tracking an active fault mode, less frequent if site is proven to be low risk

• Regular I-V tracing of specified “golden” strings– Minimum one string per manufacturing batch.

– Initial 3rd party flash test of all modules. I-V immediately after re-install to identify field testing bias.

• Alt: Install “field flash test” persistent I-V tracers

– Annual or bi-annual module level I-V of “golden” string.

• Control for Irradiance level, Angle of Incidence, utilize multiple temperature probes per module

– Periodic re-flash (2-5 year)

• Alt: Periodic re-use of calibrated persistent I-V tracers

Value add of aerial infrared inspections

Asset Management Performance Engineering

Safety

Financial Management

Detect Module malfunctions

Warranty Implications

Dispatch Control

Site Cleaning

Performance Guarantee Optimization

Model Improvement

Performance model calibration

O&M Contractor Evaluation

Acceptance Testing

Field Optimization

Future Purchase Decisions

Asset Valuation

Evaluate Availability

ROI of aerial inspections

• Increased site output

– A large cost of remediating DC

issues is in accurately finding

them

– Accurate identification decreases

remediation costs and increases

system availability

• Decreased site labour costs

– NREL/SunSpec cost model

Energy weighted DC unavailability by asset class (800MW sample)

Procedure Cost

I-V, Voc or Current clamp checks

$814/MW

IR camera inspection $852/MW

Module electrical connection test and visual inspection

$278/MW

Questions?