One-day Synthetic Solar Irradiance Sequences

Introduction

This software uses an stochastic and bootstrap method to generate one-day synthetic solar irradiance data at a minimum 60-minute time resolution.

The software requires to define eight parameters to execute:

1. DF: Input dataset that contains the measured solar irradiance. Must be a pd.Dataframe.
2. COL: Column of the pd.DataFrame of the solar irradiance (input data).
3. YEAR: Year of study. Must exist in the input dataset.
4. MONTH: Month of study (1 to 12). Must exist in the input dataset.
5. SC: Sky condition of study. Must be 'sc1', 'sc2', 'sc3', 'sc4' or 'sc5'. Corresponds to fully covered, mostly covered, partially covered, mostly clear and totally clear, respectively.
6. METHOD: Method to generate the synthetic data. Must be 'stochastic' or 'bootstrap'.
7. IC: Confidence interval. Must be a value between 0 and 1. The suggested value is 0.95.
8. RUNS: Quantity of one-day synthetic sequences to generate.

The output of the software is a one-day synthetic solar irradiance time series with the same time resolution of the input dataset. It will be generated as many sequences as defined in the RUNS parameter.

Overview of the repository structure

.
|-- figs                   -- figures
|-- src                    -- source codes
|-- validations
|   |-- data               -- input datasets
|   |-- distributions      -- goodness of fit assessment
|   |-- metrics
|       |-- energy         -- energy production validation
|       |-- statistical    -- goodness of fit validation
|       |-- variability    -- dynamic behavior validation
|   |-- notebooks          -- validations for each input dataset
|-- environment.yml        -- python environment creation
|-- requirements.txt       -- dependencies
|-- setup.py               -- python library configuration
`-- tutorial.ipynb         -- jupyter notebook tutorial for execution

Instalation

A specific python environment called synthetic will be created to run the methods and the necessary packages will be installed. For this, open the terminal and from the (base) environment move to the directory where you have downloaded the folder of this repository. For instance, if you have downloaded the folder to your desktop:

cd /user/Desktop/synthetic_irradiance

Then, create the environment with the command:

conda env create --file environment.yml

Next, activate the environment:

conda activate synthetic

After executing the previous command, you must be in the corresponding environment, in this case denoted by (synthetic). Now the respective kernel of the newly created environment is installed with the following command:

python -m ipykernel install --user --name synthetic --display-name "synthetic"

Finally, run the following command to start the Jupyter Notebooks.

jupyter notebook

Uninstall

If you want to delete the synthetic environment, run the following command:

conda env remove -n synthetic

Finally, to remove the created synthetic kernel, run the following command:

jupyter kernelspec remove synthetic

Basic Usage

In the root folder of the repository, the tutorial.ipynb file contains the description for the basic usage of the code. This same file is suggested to be used as a template to generate the synthetic data. It is just needed to adaptate the input parameters.

An example of the code pipeline is:

import src
import pvlib
import numpy as np
import pandas as pd

First, upload the input dataset.

df = pd.read_csv(filepath_or_buffer='./validations/data/bogota-5.csv',
                 sep=',',
                 decimal='.',
                 index_col='timestamp',
                 parse_dates=True)

Next, define the input parameters.

DF = df
COL = 'ghi_wm2'
YEAR = 2023
MONTH = 1
SC = 'sc5'
METHOD = 'bootstrap'
IC = 0.95
RUNS = 5

Finally, generate the one-day synthetic solar irradiance time series with the following function:

synthetic = src.methods.sequential(data=DF,
                                   irradiance_column=COL,
                                   year=YEAR,
                                   month=MONTH,
                                   sky_condition=SC,
                                   method=METHOD,
                                   confidence_interval=IC,
                                   resolution=RESOLUTION,
                                   runs=RUNS)

The function will return a pd.DataFrame for the specified sky condition (i.e., fully covered, mostly covered, partially covered, mostly clear or totally clear) with a number of columns given by the RUNS parameters and the timestamps as an index.

Citation

The original paper describing the methods implemented is:

Salazar-Peña, N., Tabares, A., Gonzalez-Mancera, A., 2023. Sequential stochastic and bootstrap methods to generate synthetic solar irradiance time series of high temporal resolution based on historical observations. Solar Energy, Vol. 264, 112030. URL: https://www.sciencedirect.com/science/article/pii/S0038092X23006643, doi: https://doi.org/10.1016/j.solener.2023.112030.

The BibTex entry:

@article{SalazarPena2023,
title = "Sequential stochastic and bootstrap methods to generate synthetic solar irradiance time series of high temporal resolution based on historical observations",
journal = "Solar Energy",
volume = "264",
pages = "112030",
year = "2023",
issn = "0038-092X",
doi = "https://doi.org/10.1016/j.solener.2023.112030",
url = "https://www.sciencedirect.com/science/article/pii/S0038092X23006643",
author = "Nelson Salazar-Peña and Alejandra Tabares and Andrés González-Mancera",
keywords = "synthetic data, solar radiation models, irradiance generation, stochastic modeling, clear-sky index, sky condition",
}

Licence

GNU AFFERO GENERAL PUBLIC LICENSE v3.0, dispuesta en el archivo LICENSE.