time_series_data_shifts.zip

This data set contains 101 daily sensor-based time series, which were labeled by two solar experts for abrupt data shifts. The algorithm associated with this data set is available via the PVAnalytics package, and the work associated with this data set was presented at PVSC-49 (citation pending). The data set consists of CSV files containing daily summed normalized time series data, and associated HTML files visualizing each time series with labeled data shifts.

Each CSV file contains four columns, described as follows:

filename: the name of the data stream being analyzed, which could be an irradiance or a power data stream timestamp: the daily timestamp of the reading value: the normalized value of the reading label: whether or not there is a changepoint/data shift on that day. If no, the label is 0. If yes, the label is 1.