Overview of DuraMat software tool development
0 likes176 views
The document discusses software tools being developed by researchers for photovoltaic (PV) applications. It summarizes several software projects funded by DuraMat that address different aspects of PV including: (1) PV system modeling and analysis, (2) operation and degradation modeling, and (3) planning and reducing levelized cost of energy. The software aims to solve a range of PV problems, are open source, and developed collaboratively on GitHub to be reusable and sustainable resources for the community.
More Related Content
Similar to Overview of DuraMat software tool development (20)
Overview of DuraMat software tool development
- 1. Overview of DuraMat software tool development Anubhav Jain Lawrence Berkeley National Laboratory w/contributions from: Todd Karin (LBL), Xin Chen (LBL), Thushara Gunda (Sandia), Cliff Hansen (Sandia), Bennet Meyers (Stanford), Brittany Smith (NREL) and their respective teams
- 2. Why a talk on software development? • The “end product” of most conventional funded research is a paper or report – This is typically how projects are evaluated • However, software can also be an invaluable end product! – Sometimes, more valuable than text • Data can also be an end product! – See poster on DuraMat Data Hub by R. White. 2 Given a data set, up to 10 analysts were asked to calculate the degradation rate of 5 PV systems using either (i) any approach they wanted, (ii) a documented standard, or (iii) a documented codebase (RdTools). Consistent results came only with the codebase, highlighting the role software can play in reproducibility and reusability. Jordan, D. C., Luo, W., Jain, A., Saleh, M. U., von Korff, H., Hu, Y., Jaubert, J.-N., Mavromatakis, F., Deline, C., Deceglie, M. G., Nag, A., Kimball, G. M., Shinn, A. B., John, J. J., Alnuaimi, A. A. & Elnosh, A. B. A. Reducing Interanalyst Variability in Photovoltaic Degradation Rate Assessments. IEEE J. Photovoltaics 10, 206–212 (2020).
- 3. DuraMat-funded projects are developing software to solve a spectrum of PV problems 3 Core functions common to many PV analyses Operation and degradation in the field Planning and reduction of LCOE pvanalytics pv-pro data preprocessor pv-pro pvOps pvARC pv-vision simple LCOE calculator vocmax
- 4. DuraMat software projects share some DNA • Open-source licenses – Typically MIT / BSD which are commercial-friendly • All based on Python / pydata stack for interoperability – not mixing MATLAB, Python, Java, Excel macros, etc. • Make use of large data sets and machine learning / statistical learning – e.g., neural networks for vision, natural language processing for text • Collaborative and sustainable development via the Github platform – i.e., NOT a single developer sending around “script_v34” by email • More integration forthcoming – e.g., https://duramat.github.io/pv-terms/ to unify common variable names 4
- 5. pvlib/pvanalytics / Cliff Hansen, Will Vining, Matt Muller 5 • python library for automating analysis of data from PV systems • Workflow independent, built up from base functions • v0.1 released 20 November github.com/pvlib/pvanalytics • Fully compatible with pvlib/pvlib-python for PV system modeling Quality control functions • Plausibility of irradiance and weather measurements • Identification of missing, interpolated, or stale data • outlier detection • Identification of timestamp problems such as daylight savings shifts Feature identification functions • inverter clipping • clear-sky periods • day/night detection from power or irradiance Identification of system properties • tilt and azimuth from power data • differentiation between fixed and tracking PV systems Metrics • NREL weather corrected performance ratio v0.1 Content Core
- 6. PV-Pro Data Preprocessor / Bennet Meyers, Todd Karin 6 PV-Pro can be found at: https://github.com/DuraMAT/pvpro • 0: System at maximum power point. • 1: System at open circuit conditions. • 2: Clipped or curtailed. • -1: No power/inverter off • -2: Other (errors, corrupted data) def run_preprocess(self, correct_tz=True, data_sampling=None, correct_dst=False, fix_shifts=True, max_val=None): Core
- 7. PVPRO/ Todd Karin, Anubhav Jain, Bruce King, Michael Deceglie, Bennet Meyers, Dirk Jordan, Cliff Hanen, Laura Schelhas 7 PVPRO be found at: https://github.com/DuraMAT/pvpro • Use operation data (DC current, DC voltage, module temperature and plane-of- array irradiance) to determine module single diode model parameters vs. time. PRODUCTION DATA Time series database PVPRO • Filter power data • Meteorological data • Circuit model • Parameter estimation • Uncertainty analysis I-V Parameters Rseries, Voc, Isc, Bvoc, … Degradation Mode Estimates Soiling, PID, Encpasulant discoloration, solder bond failure Output BIG data analysis Application Technology Degradation mode Solder bond damage detected Methods Actionable analytics More accurate power predictions Operation & Degradation
- 8. pvOps / Thushara Gunda, Michael Hopwood, Hector Mendoza 8 pvOps is currently underdoing review and will be published on GitHub as part of pvlib in April 2021 Module 1: Text • Date Extraction • Consistent Labels • Issue Summary Module 2: Text2Time • Date Alignments • Quantify Impact • Visualizations { “Date_EventStart”: XXX “Date_EventEnd”:XXX “Asset”:XXX “ProductionImpact”:XXX “Response”:XXX “IssueDescription”:XXX } • Although text-based records can contain valuable contextual information to support data processing and site assessment activities, significant diversity in these records makes it challenging to ascertain needed insights. • The pvOps package contains two modules to support: 1) processing of text data and 2) fuse text with timeseries data • Similar to other pvlib packages, pvOps is written as a series of individual functions that could be integrated using class wrappers to support individual workflows Operation & Degradation
- 9. PVARC/ Todd Karin, David Miller, Anubhav Jain 9 PVARC be found at: https://github.com/DuraMAT/pvarc • Extract coating parameters from spectral reflectance data using thin-film interferometry model “Non-destructive Characterization of Anti-reflection Coatings on PV Modules” Posted arXiv:2101.05446. (in- press at JPV). Spectral Reflectance Measurement of PV module glass Quantify anti reflection properties Operation & Degradation
- 10. PV-vision/ Xin Chen, Todd Karin, Anubhav Jain 10 PV-vision be found at: https://github.com/hackingmaterials/pv-vision Solar Cell Crack Detection Algorithm & Data Analytics Automatic Classification of Fire damaged Solar Modules • Goal: automatically extract features of cracks on solar modules and quantify the relationship between features and module power loss • Tool developed for cropping out single cells from solar modules with accuracy of 90% • Goal: automatically classify the solar modules based on the categories of defects which are induced by conflagration • New tool assisted with CNN developed to do perspective transform and cell cropping. Accuracy improved and will be applied to all dataset crop • UNet model trained to segment the cracks with overall F1 metric of 0.89 • Initial tool developed to calculate crack features (length, orientation, etc.) Fig.4 Left: original image. Right: predicted masks. F1 metric=0.882 vectorized Fig.5 Left: crack vectors. Middle: poly-fitted cracks. Right: filter short cracks(length<30) UNet Fig.1 Solar module Fig.2 Cropped solar cells Fig.3 Masks predicted by UNet model. Left: original image. Right: masks are highlighted in different colors. Purple (crack), Green (power loss area), brown (busbar). F1 metric=0.862 Fig.6 Original, perspective- transformed solar module and cropped cells transform (CNN assisted) crop • Yolo model trained to detect defective cells in the solar modules • Solar modules classified based on the number of different defective cells with accuracy of 98% Fig.7 Defective cells detected by Yolo model. Green: crack defect. Blue: Solder defect. Red: Oxygen induced defect. Purple: Intra-cell defect. Classification criterion If no defective cell or ‘crack’ <= 1: Category 1 If ‘crack’ >= 2 or ‘oxygen’ >= 1: Category 2 If ‘intra’ >= 1: Category 3 Operation & Degradation
- 11. Simplified PV LCOE Calculator* / Brittany Smith (NREL) 11 The simplified PV LCOE calculator can be found at: www.github.com/NREL/PVLCOE Currently, the calculator offers only two default degradation rates (based on cell technology selection) and assumes module degradation approximates system-level degradation. DuraMAT analysis of PV Fleets data will provide updated system-level degradation rates and could potentially populate different default degradation rate values for: • Cell technology • Package type • System type • Install location The calculator lets you select from a set of system options: Default values pre-populate based on selections Results include LCOE, module price, system cost *calculator was originally developed outside of DuraMAT Planning & LCOE
- 12. Vocmax – a string length calculator for reducing LCOE / Todd Karin, Anubhav Jain 12 Conventional string length calculations are unnecessarily conservative • Most common practice is to use minimum historical ambient temperature and 1000 W/m2 for calculating maximum string Voc, but these conditions do not co-occur in practice. • Longer strings lower system costs – get more power through the same wires. Need a more realistic method for calculating string length • We developed and validated a method for calculating the string length by modeling the system VOC over time at the location of interest. • Method is consistent with NEC 690.7(A)(3) standard. Impact • In the US, string lengths increased by 10% on average using site—specific modeling, leading to a 1.2% reduction in LCOE ! https://github.com/toddkarin/vocmax https://pvtools.lbl.gov/string-length-calculator https://ieeexplore.ieee.org/document/9000497 Approach: model the distribution of operating and open-circuit voltage over time to determine threshold Planning & LCOE
- 13. Conclusions • In addition to papers and reports, software and data represent important outputs for research • DuraMat has a diverse portfolio of software development projects across the chain of PV operation • The software is available open-source and is the product of collaborative development • We would be very happy to hear from you! 13