Scholarlink

La Guajira, Colombia, is highlighted as one of the most solar-rich regions in Latin America, with Global Horizontal Irradiance (GHI) values higher than \(6.0~\text {kWh/m}^2/\text {day}\) and annual mean temperatures around \(28\,^{\circ }\text {C}\) . However, indigenous communities from this region still face high levels of energy poverty. In this chapter, a holistic approach is presented to combine NASA Prediction of Worldwide Energy Resources (POWER) with state-of-the-art machine learning algorithms in order to improve solar resource assessment. The workflow is embedded in the Google Colab, using Python libraries such as scikit-learn, TensorFlow, pandas, and matplotlib. Long Short-Term Memory (LSTM) networks are adopted for time-series forecasting of GHI and temperature, as well as K-Means clustering to partition microclimatic zones with similar solar irradiance patterns. Finally, Principal Component Analysis (PCA) is used for feature selection and dimensionality reduction, which can enhance the accuracy of the model.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Time-Series Forecasting of Solar Irradiance for Resilient Microgrids in Wayuu Communities Using NASA POWER Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Maria C. Moreno,
  • Brayan Daniel Sarmiento,
  • John C. Moreno,
  • Oscar J. Suarez

摘要

La Guajira, Colombia, is highlighted as one of the most solar-rich regions in Latin America, with Global Horizontal Irradiance (GHI) values higher than \(6.0~\text {kWh/m}^2/\text {day}\) and annual mean temperatures around \(28\,^{\circ }\text {C}\) . However, indigenous communities from this region still face high levels of energy poverty. In this chapter, a holistic approach is presented to combine NASA Prediction of Worldwide Energy Resources (POWER) with state-of-the-art machine learning algorithms in order to improve solar resource assessment. The workflow is embedded in the Google Colab, using Python libraries such as scikit-learn, TensorFlow, pandas, and matplotlib. Long Short-Term Memory (LSTM) networks are adopted for time-series forecasting of GHI and temperature, as well as K-Means clustering to partition microclimatic zones with similar solar irradiance patterns. Finally, Principal Component Analysis (PCA) is used for feature selection and dimensionality reduction, which can enhance the accuracy of the model.