DuraMat Data Analytics

DuraMat Data Analytics

• 1.
  DuraMAT Capability 1: Data Analytics Anubhav Jain Lawrence Berkeley National Laboratory August 28, 2018
• 2.
  Goals of the data analytics thrust Field deployment Data HubData analytics Materials Forensics & Characterization Predictive simulationModule testing Techno-economic analysis Data analytics is a cross- cutting thrust that works with DuraMat partners to provide software, visualization, and data mining capabilities. This thrust does not produce data itself!
• 3.
  Project 1: Clear Sky Detection (background) Jordan, D. C.,Deline, C., Kurtz, S. R., Kimball, G. M. & Anderson, M. Robust PV Degradation Methodology and Application. IEEE J. Photovoltaics 8, 525–531 (2018). In many analyses, data filtering / preprocessing greatly affects final results! D. Jordan et al. showed that clear sky filtering produces degradation rates closer to expectation and 20% different than without filtering! Generally clear Scattered cloudsPersistent clouds Can we design an algorithm that automatically and reliably distinguishes clear sky periods based on GHI measurements – across locations and data frequencies? Approach: use satellite data to modify a published clear sky detection technique (Reno and Hansen, 2016) to generalize across locations and data frequencies.
• 4.
  Project 1: Clear Sky Detection (results) Using satellite clearsky labels as a guide, we can design an “optimized” clear sky detection algorithm with no parameters that works better than existing pvlib across sites and data frequencies! Heatmaps plot F0.5 scores, or classification accuracy, of clear sky algorithms Visual inspection confirms that clear sky classifications from the optimized algorithm are more relevant and correct. Next: integration into pvlib and/or rdtools default optimized Sample GHI data and clear sky classifications for BMS site, 30 minute data frequency
• 5.
  Project 2: Degradation dashboards (background) Backend tools likerdtools (NREL, kWh Analytics) expose many powerful functions for processing and analyzing data through programmatic (e.g., Python, MATLAB) APIs. How can we make some of these tools more visual, interactive, and exploratory? Jordan, D. C., Deline, C., Kurtz, S. R., Kimball, G. M. & Anderson, M. Robust PV Degradation Methodology and Application. IEEE J. Photovoltaics 8, 525–531 (2018). www.github.com/NREL/rdtools
• 6.
  Project 2: Degradation dashboards (results) We have builtinteractive dashboards that analyze data from PVOutput.org Goal: make the backend tools interactive!
• 7.
  Misc. - contact angle analysis, I-V curve analysis Automated contact angle analysisfor evaluating anti- soiling coatings E x t r a c t i o n o f I - V c u r v e parameters, detection of m i s m a t c h , a n d l e a r n i n g “features” of I-V curves
• 8.
  Future directions •  Relating combined accelerating stress testing (C-AST) and field measurements •  Working with other data analytics efforts in the field (e.g., Case Western) • Open to other project ideas! –  contact: [email protected] •  Collaborators: Benjamin Ellis, Mike Deceglie, Birk Jones, Josh Stein, Dirk Jordan, Laura Schelhas, Stephanie Moffitt, Margaret Gordon