Predictive Modeling For Solar-Thermal Power Plant Output: Leveraging Machine And Deep Learning Techniques.

Saved in:

Bibliographic Details
Title:	Predictive Modeling For Solar-Thermal Power Plant Output: Leveraging Machine And Deep Learning Techniques.
Authors:	Sharma, Vivek¹, Panda, Mohit Ranjan¹, Pattanayak, Binod Kumar²
Source:	International Journal of Special Education. 2026 Special Issue, Vol. 41, p25-41. 17p.
Subject Terms:	*Machine learning, Random forest algorithms, Long short-term memory, Time, Prediction models, Deep learning, Energy industry forecasting, Energy consumption
Abstract:	It is vital to correctly forecast the output of a solar power plant to manage the energy efficiency and stability of the grid. The current study examines multiple machine and deep learning ideas for forecasting the power output of a plant using data already available to us. Key aspects such as temporal variables, solar irradiance, etc. that are important for setting up the relation between climate conditions and power output are non-negotiable parts of the available data set. The study considers 5 models currently: Linear Regression, Decision Tree Regression, Random Forest, Support Vector Regressor (SVR), convolutional neural networks (CNN), and bidirectional long short-term memory (BiLSTM). Highest accuracy was shown by the random forest model, that of around 72.39% with a standard deviation of 4.36%, over and above other methods. 53.42% accuracy was displayed by the Decision Tree Regressor, but high variance limited its usage. At 27.87%, the linear regression method could not handle the non-linear factors in the data. The least useful was the SVR with an accuracy of 0.79%. Supposedly, the BiLSTM method shows promising results while using temporal dependencies - a R² score of 0.6793, a mean absolute error (MAE) of 28,519.93, and a mean squared error (MSE) of 1,594,774,513.30, thus properly utilizing the sequential data available. It performed better than the CNN model, which showed a test loss of 2,618,489,344.0 and an MAE of 39,056.19. To summarize the findings, we can confidently state that while BiLSTM is useful for sequential dependencies for time-series forecasting, the random forest method is appropriate for nonlinear forecasting. Thus, it becomes vital to select the right predictive model based on pre-available data to get the desired result for output forecasting. [ABSTRACT FROM AUTHOR]
	Copyright of International Journal of Special Education is the property of International Journal of Special Education and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Education Research Complete

Description
Abstract:	It is vital to correctly forecast the output of a solar power plant to manage the energy efficiency and stability of the grid. The current study examines multiple machine and deep learning ideas for forecasting the power output of a plant using data already available to us. Key aspects such as temporal variables, solar irradiance, etc. that are important for setting up the relation between climate conditions and power output are non-negotiable parts of the available data set. The study considers 5 models currently: Linear Regression, Decision Tree Regression, Random Forest, Support Vector Regressor (SVR), convolutional neural networks (CNN), and bidirectional long short-term memory (BiLSTM). Highest accuracy was shown by the random forest model, that of around 72.39% with a standard deviation of 4.36%, over and above other methods. 53.42% accuracy was displayed by the Decision Tree Regressor, but high variance limited its usage. At 27.87%, the linear regression method could not handle the non-linear factors in the data. The least useful was the SVR with an accuracy of 0.79%. Supposedly, the BiLSTM method shows promising results while using temporal dependencies - a R² score of 0.6793, a mean absolute error (MAE) of 28,519.93, and a mean squared error (MSE) of 1,594,774,513.30, thus properly utilizing the sequential data available. It performed better than the CNN model, which showed a test loss of 2,618,489,344.0 and an MAE of 39,056.19. To summarize the findings, we can confidently state that while BiLSTM is useful for sequential dependencies for time-series forecasting, the random forest method is appropriate for nonlinear forecasting. Thus, it becomes vital to select the right predictive model based on pre-available data to get the desired result for output forecasting. [ABSTRACT FROM AUTHOR]
ISSN:	08273383