dc.contributor.author | Omar, Najiya | |
dc.date.accessioned | 2023-05-01T15:23:18Z | |
dc.date.available | 2023-05-01T15:23:18Z | |
dc.date.issued | 2023-04-28 | |
dc.identifier.uri | http://hdl.handle.net/10222/82558 | |
dc.description | The limited accessibility of solar irradiance data drives the need for robust Global
Horizontal Irradiance (GHI) prediction models. To date, numerous scholars have carried out research looking for ways to enhance the performance of a Long Short-Term Memory (LSTM ) model in terms of univariate and multivariate analyses. Although
high-dimensional heterogeneous weather data are desirable for enhancing forecasting accuracy, LSTM performance deteriorates when changing from univariate to multivariate analyses. As previous research stops short of conducting detailed explorations
on how interactions in high dimensional heterogeneous data represent critical elements in LSTM predictive model development, the present research aims to fill that gap. This work proposes two techniques to enhance predictive performance. | en_US |
dc.description.abstract | This work proposes two techniques to enhance predictive performance.
The first technique addresses implementation details regarding relevancy and redundancy measures, exploring how they may, respectively, be enhanced and mitigated. The proposed technique is a novel hybrid feature selection method built to optimize feature selection using a framework based on Least Redundant/Highest- Relevant, named Weather Recursive Feature Elimination (WRFE). The WRFE approach uses feature importance to measure reductions in variance in Random Forest Regression (RFR) in addition to data perturbation in LSTM. The training set’s optimal features demonstrate strong contributions to the prediction outcome, indicating the proposed WRFE’s generalizability for hourly GHI prediction. To lessen the seasonality effect, the second proposed technique employs a deep stack of the clustering connected layer with hybrid LSTM models. This novel Seasonal Clustering Forecasting Technique (SCFT) is then compared with other forecasting strategies, revealing its superiority. | en_US |
dc.language.iso | en | en_US |
dc.subject | Deep Long Short-Term Memory (LSTM) | en_US |
dc.subject | GHI Forecasting | en_US |
dc.subject | Clustering Approach | en_US |
dc.subject | Correlation Analyses | en_US |
dc.subject | Hybrid Feature Importance | en_US |
dc.title | ENHANCED PERFORMANCE OF SOLAR IRRADIANCE PREDICTION USING DEEP LEARNING AND DATA MINING TECHNIQUES | en_US |
dc.date.defence | 2023-04-19 | |
dc.contributor.department | Department of Electrical & Computer Engineering | en_US |
dc.contributor.degree | Doctor of Philosophy | en_US |
dc.contributor.external-examiner | Dr. Mohamed Darwish | en_US |
dc.contributor.graduate-coordinator | Dr. Vincent Sieben | en_US |
dc.contributor.thesis-reader | Dr. Jason Gu | en_US |
dc.contributor.thesis-reader | Dr. William Phillips | en_US |
dc.contributor.thesis-supervisor | Dr. Timothy Little | en_US |
dc.contributor.thesis-supervisor | Dr. Hamed Aly | en_US |
dc.contributor.ethics-approval | Not Applicable | en_US |
dc.contributor.copyright-release | Yes | en_US |