Data Cleaning for Degradation Analyses
DuraMAT will significantly enhance data preparation capabilities by providing open-source to automatically filter time-series irradiance and photovoltaic (PV) power data. We will also create methods to automatically translate textual O&M records to time-series indicators of PV system availability.
Data filtering is currently a manual, time-intensive task. Current filtering often relies on user-specified bounding values and often does not account for non-time series data such as O&M records.
Analysis of PV performance data to determine degradation requires careful and reliable filtering of the performance data. Filtering should remove data suspected to be erroneous, but also should isolate times when the PV system is operating without external constraints, for example, when portions of the array are shaded or when DC output is limited due to inverter curtailment. Without appropriate filtering, the degradation signal can be swamped by the noise of incorrect data.
Sandia National Laboratories
Capabilities will be shared in open-source python libraries. For time-series filtering and feature recognition, see GitHub.
Webinar: "Degradation Analysis with Real-World PV Power Plant Data" presented by Todd Karin, LBL, and Cliff Hansen, Sandia.
To learn more about this project, contact Clifford Hansen.