Environmental Data: Monitoring data sets of alpine photovoltaic power-...

Description

On this repository you can find the monitoring data of the 3 PV power-plants analysed in the article _"Confirmation of the power gain for solar photovoltai...

Citation

Frischholz, Y. (2024). Monitoring data sets of alpine photovoltaic power-plants. EnviDat. https://www.doi.org/10.16904/envidat.469.

Resources

• A1.zip

  Title: Monitoring data set at A1 (test-site of Totalp), years 2018-2022 Provider: ZHAW - EKZ Edition: Y.Frischholz, Laboratory of Cryospheric Sciences, EPFL Last edition: 17.01.2024 File name structure: The test site contains 6 segments named from A to F. The six files provided in the data/ subfolder follow the same naming convention (e.g. for segment A, the file name is: “A1_A.csv”). Refer to the publication for further details on the setups of the segments. File format: All files are provided in the CSV format. Column headings: • timestamp (datetime64): time stamp in format “YYYY-MM-DD HH:mm:ss”, time-zone: UTC • poa_front (double): front plane-of-array irradiation (W/m**2) • poa_back (double): rear plane-of-array irradiation (W/m**2) • mod_temp (double): module’s temperature (°C) • P_AC (double): output AC power (W) Preprocessing: Outliers: ⁃ Quantiles Q1 and Q99 are computed for each variable ⁃ Values below Q1 and above Q99 are considered outliers ⁃ Outliers are replaced by NaNs. Missing data: ⁃ Data holes of less than 3 hours are interpolated linearly for all variables ⁃ Data holes of more than 3 hours are replaced using the monthly nan-mean of the considered hour if more than 50% of the qualifying data points are available. For instance, if the data point 2022-12-12 12:12:00, is missing for a given variable and is part of a data hole larger than 3 hours, then it is replaced by the mean of all data points with hour = 12 of the month of December 2022, if 50% of all the qualifying data points are not NaN. In case more than 50% of the qualifying points are missing, the data point remains as missing (NaN). Wiring losses : ⁃ As described in [1] , the test-site setup caused abnormally large wiring losses, mainly due to the distance to optimisers and to the low installed power per DC-wire. ⁃ The wiring losses comprised between 10% and 15%, are averaged hourly and added to the measured DC power. Misc.: ⁃ P_AC is computed for each segment based on the measured DC power and the average inverter’s efficiency (96.1%). ⁃ As described in [1], the peak power of the installed module, used to compute the final yield is slightly different from the values announced by the manufacturer. Mean uptime: Raw: ⁃ 95.2% Processed: ⁃ 100% References: [1] D. Anderegg, S. Strebel, and J. Rohrer, “Alpine Photovoltaik Versuchsanlage Davos Totalp : Erkenntnisse aus 5 Jahren Betrieb,” de, Sep. 2023, Accepted: 2023-09-29T09:13:16Z Publisher: ZHAW Zürcher Hochschule für Angewandte Wissenschaften.DOI: 10.21256/ zhaw-2524. [Online]. Available: https://digitalcollection.zhaw.ch/handle/11475/28797 (visited on 12/15/2023).

  A1.zip

• A2.zip

  Title: Monitoring data set at A2 (utility-scale power-plant Muttsee), months 10.2022-09.2023 Provider: AXPO - IWB Industrielle Werke Basel Last edition: 17.01.2024 Editor: Y.Frischholz, Laboratory of Cryospheric Sciences, EPFL File name structure: The power-plant is made of 23 string boxes (GAK) and inverters (WR) named after their given number. One file per device is provided in the data/ subfolder, following the same naming convention (e.g. for the string box number 1, the file name is: “A2_GAK01.csv” and for the corresponding inverter, the file name is “A2_WR01.csv”). Refer to the publication for further details on the setup of the power plant. File format: All files are provided in the CSV format. Column headings: String boxes and inverters (GAK & WR): ⁃ device_name (string): name of the device (same as file name) ⁃ timestamp (datetime64): time stamp in format “YYYY-MM-DD HH:mm:ss”, time-zone: UTC String boxes (GAK): ⁃ I_# (double): input current from the substring number # (W) ⁃ I_SUM (double): output current of string box (A) ⁃ U_DC (double): output voltage of string box (V) ⁃ P_DC_# (double): input DC power from the substring number # (W) ⁃ P_DC (double): output DC power of string box (W) ⁃ $S$_T#_$T$ (double): temperature from the sensor connected to the string box (empty if not existing), placed on the “front” or “back” side (S) of table T. Two types of temperature measurements (#) are available: 1=standard cell temperature (°C) , 2=module temperature on the rear side (°C) . ⁃ $S$_SRAD_T (double): poa irradiation (W/m**2) from the sensor connected to the string box (empty if not existing), placed on the “front” or “back” side (S) of table T. Inverters (WR): ⁃ I_DC (double): input DC current(A) ⁃ U_DC(double): input DC voltage(V) ⁃ P_DC (double): input DC power (W) ⁃ P_AC (double): output AC power (W) Preprocessing: 1. Outliers: ⁃ Quantiles Q1 and Q99 are computed for each variable ⁃ Values below Q1 and above Q99 are considered outliers ⁃ Outliers are replaced by NaNs. 2. Missing data: ⁃ Data holes of less or equal to 3 hours are interpolated linearly for all variables ⁃ Data holes of more than 3 hours are replaced using the monthly nan-mean of the considered hour if more than 50% of the qualifying data points are available. For instance, if the data point 2022-12-12 12:12:00, is missing for a given variable and is part of a data hole larger than 3 hours, then it is replaced by the mean of all data points with hour = 12 of the month of December 2022, if 50% of all the qualifying data points are not NaN. In case more than 50% of the qualifying points are missing, the data point remains as missing (NaN). 3. Wiring losses: ⁃ To be consistent with A1, wiring losses are corrected at A2. The wiring loss between the string boxes and the inverters is removed by multiplying the value of input P_DC at the inverter level (P_DC,inv) by the mean ratio P_DC,inv / P_DC,arr. ⁃ The computed mean wiring loss is 2.3\%. It is computed from 2023-02-01 to 2023-09-30, after the long data hole in the data sets of the string boxes. Mean uptime: Raw: ⁃ WR: 98.2% ⁃ GAK*: 38.5% Processed: ⁃ WR: 99.6% ⁃ GAK*: 56.4% *The uptime at the string box level (GAK) is low because PMUs at this level shut down during steady and low signal leading to unavailable data at night. In addition, the 23 days downtime of January 2023 could not be interpolated nor replaced.

  A2.zip