A recent study led by Abdulelah Alharbi from the Department of Electrical Engineering at Qassim University has introduced an innovative approach to short-term forecasting of solar energy potential in Saudi Arabia. Published in ‘IEEE Access’, the research focuses on predicting Global Horizontal Irradiance (GHI), a key factor in determining solar energy generation.
As the world shifts towards renewable energy sources to address climate change and rising energy demands, solar power stands out due to its abundance and environmental benefits. However, one of the significant challenges in harnessing solar energy is its variability, which can complicate the integration of solar power into existing energy systems. Traditional deep-learning techniques have been employed to address this issue, but they often come with high computational costs and energy requirements.
Alharbi’s team has proposed a novel deep-learning model called the NeuroSpike network, which utilizes Leaky Integrated and Fire (LIF) spiking neurons. This model combines a Recurrent Neural Network (RNN) layer with a Long Short-Term Memory (LSTM) layer to enhance forecasting accuracy while reducing computational demands. The research utilizes historical GHI data from three different locations in Saudi Arabia—Al-Jouf, Qassim, and the King Abdulaziz City for Science and Technology (K.A.CARE).
A key part of the study involved a data preprocessing technique known as Recursive Feature Elimination with Categorical Boosting (RFE-CatBoost). This method helps in selecting the most relevant features from the dataset, ultimately improving the model’s accuracy. The results were promising, showing that the NeuroSpike network achieved lower forecasting errors compared to existing techniques. Specifically, the model recorded improvements of 30.33%, 43.12%, and 23.4% in the Mean Absolute Error (MAE) for the datasets from Al-Jouf, Qassim, and K.A.CARE, respectively.
“The integration of spiking neurons and the proposed RFE-CatBoost feature selection technique makes the training of the NeuroSpike network more effective and computationally less demanding,” stated Alharbi. This advancement could have significant implications for the energy sector, particularly in regions like Saudi Arabia where solar energy potential is high.
The commercial opportunities arising from this research are substantial. Energy companies could leverage the NeuroSpike network to optimize solar energy generation forecasts, leading to better integration of solar power into the grid. This could enhance grid reliability and efficiency, reduce reliance on fossil fuels, and ultimately contribute to a more sustainable energy future.
As countries continue to pursue renewable energy goals, innovations like the NeuroSpike network could play a crucial role in making solar energy more viable and efficient. The research not only highlights the potential of advanced computing techniques in energy forecasting but also underscores the importance of adapting to the challenges posed by renewable energy sources.
For more information about the lead author’s affiliation, visit Qassim University.
