Enhancing deep learning-based slope stability classification using a novel metaheuristic optimization algorithm for feature selection.

The evaluation of slope stability is of crucial importance in geotechnical engineering and has significant implications for infrastructure safety, natural hazard mitigation, and environmental protection. This study aimed to identify the most influential factors affecting slope stability and evaluate the performance of various machine learning models for classifying slope stability. Through correlation analysis and feature importance evaluation using a random forest regressor, cohesion, unit weight, slope height, and friction angle were identified as the most critical parameters influencing slope stability. This research assessed the effectiveness of machine learning techniques combined with modern feature selection algorithms and conventional feature analysis methods. The performance of deep learning models, including recurrent neural networks (RNNs), long short-term memory (LSTM) networks, and generative adversarial networks (GANs), in slope stability classification was evaluated. The GAN model demonstrated superior performance, achieving the highest overall accuracy of 0.913 and the highest area under the ROC curve (AUC) of 0.9285. Integration of the binary bGGO technique for feature selection with the GAN model led to significant improvements in classification performance, with the bGGO-GAN model showing enhanced sensitivity, positive predictive value, negative predictive value, and F1 score compared to the classical GAN model. The bGGO-GAN model achieved 95% accuracy on a substantial dataset of 627 samples, demonstrating competitive performance against other models in the literature while offering strong generalizability. This study highlights the potential of advanced machine learning techniques and feature selection methods for improving slope stability classification and provides valuable insights for geotechnical engineering applications.

An analysis of case studies for advancing photovoltaic power forecasting through multi-scale fusion techniques.

Integration renewable energy sources into current power generation systems necessitates accurate forecasting to optimize and preserve supply-demand restrictions in the electrical grids. Due to the highly random nature of environmental conditions, accurate prediction of PV power has limitations, particularly on long and short periods. Thus, this research provides a new hybrid model for forecasting short PV power based on the fusing of multi-frequency information of different decomposition techniques that will allow a forecaster to provide reliable forecasts. We evaluate and provide insights into the performance of five multi-scale decomposition algorithms combined with a deep convolution neural network (CNN). Additionally, we compare the suggested combination approach's performance to that of existing forecast models. An exhaustive assessment is carried out using three grid-connected PV power plants in Algeria with a total installed capacity of 73.1 MW. The developed fusing strategy displayed an outstanding forecasting performance. The comparative analysis of the proposed combination method with the stand-alone forecast model and other hybridization techniques proves its superiority in terms of forecasting precision, with an RMSE varying in the range of [0.454-1.54] for the three studied PV stations.

Hybrid attention-based deep neural networks for short-term wind power forecasting using meteorological data in desert regions.

This study introduces an optimized hybrid deep learning approach that leverages meteorological data to improve short-term wind energy forecasting in desert regions. Over a year, various machine learning and deep learning models have been tested across different wind speed categories, with multiple performance metrics used for evaluation. Hyperparameter optimization for the LSTM and Conv-Dual Attention Long Short-Term Memory (Conv-DA-LSTM) architectures was performed. A comparison of the techniques indicates that the deep learning methods consistently outperform the classical techniques, with Conv-DA-LSTM yielding the best overall performance with a clear margin. This method obtained the lowest error rates (RMSE: 71.866) and the highest level of accuracy (R2: 0.93). The optimization clearly works for higher wind speeds, achieving a remarkable improvement of 22.9%. When we look at the monthly performance, all the months presented at least some level of consistent enhancement (RRMSE reductions from 1.6 to 10.2%). These findings highlight the potential of advanced deep learning techniques in enhancing wind energy forecasting accuracy, particularly in challenging desert environments. The hybrid method developed in this study presents a promising direction for improving renewable energy management. This allows for more efficient resource allocation and improves wind resource predictability.