AIN311 Machine learning in Sustainability project proposal
Title: Predictive Analysis of Global Horizontal Irradiance for Enhanced Solar Energy Integration
Research Topic to be Investigated: The efficient utilization of solar energy hinges on the accurate prediction of Global Horizontal Irradiance (GHI). This research intends to delve into the complexities of forecasting GHI to facilitate the optimization of photovoltaic (PV) system performance. The problem under inspection involves assessing the quantity of solar radiation that reaches the Earth’s surface and processing this data to predict future trends and potential energy yields. Through this, the project aims to support the strategic planning for solar energy integration into the electrical grid, paving the way for a transition towards more sustainable energy sources and a reduction in the reliance on non-renewable fossil fuels.
Sustainability Aspect: A core component of this research is its sustainability angle. Accurate predictions of GHI are vital for solar energy to become more reliable and widely adopted. By increasing the efficiency and predictability of solar power, this study seeks to contribute to the creation of a sustainable energy landscape. The expected outcomes will be instrumental in managing energy resources, optimizing the use of natural sunlight, and consequently reducing carbon footprints.
Data Utilization Strategy: The initial phase of this project will employ a dataset available on Kaggle which was featured in “Wipro’s sustainability machine learning challenge”. If the Kaggle dataset prove insufficient for certain predictive models, an expansion or integration with the National Solar Radiation Database’s data can be made. This dataset offers extensive and granular data which may enhance the robustness of our predictive analytics
Dataset Links:
https://www.kaggle.com/datasets/vickeytomer/wipros-sustainability-machine-learning-challenge/data
Related work:
Many machine learning models have been tried to solve GHI prediction problem. In [1], researchers used K-nearest neighbors, Support Vector Machines, Logistic Regression, and Random Forest models to compare their performance for GHI prediction. In another study [2], researchers evaluated three models on the same task. These models are extreme learning machine, random forests, and neural networks. Since the time information in the data is important, LSTM models have also been used. [3][4] mainly focus on LSTM models while [4] uses CNN along with LSTM to capture spatial features in the data.
1- Alghamdi, H.A. A Time Series Forecasting of Global Horizontal Irradiance on Geographical Data of Najran Saudi Arabia. Energies 2022, 15, 928. https://doi.org/10.3390/en15030928 (Access Date: 3/11/2023)
2- https://www.sciencedirect.com/science/article/pii/S0959652619307826?via%3Dihub (Access Date: 3/11/2023)
3- Malakar, S., Goswami, S., Ganguli, B. et al. Designing a long short-term network for short-term forecasting of global horizontal irradiance. SN Appl. Sci. 3, 477 (2021). https://doi.org/10.1007/s42452-021-04421-x (Access Date: 4/11/2023)
4- https://www.sciencedirect.com/science/article/pii/S0960148120308557 (Access Date: 4/11/2023)
