Scholarlink

<p>The high growth in renewable energy systems has led to increased pressure on effective forecasting procedures to facilitate predictive maintenance as well as promote the reliability of operation of Conventional Hydroelectric Power (CHP) facilities. Even with deep learning, high-dimensional sensor data and sub-optimal tuning of hyperparameters are common challenges to models, resulting in poor forecasting performance. This work presents a new optimization framework that couples the Variable Attention Span Transformer (VAST) with the binary Swordfish Movement Optimization Algorithm (bSMOA) for feature selection and the Swordfish Movement Optimization Algorithm (SMOA) for hyperparameter optimization. The recommended model is compared to existing models, and VAST has achieved a baseline coefficient of determination (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^2\)</EquationSource> </InlineEquation>) of 0.8353 and a Mean Squared Error (MSE) of 0.0191. After selecting the features using the bSMOA, VAST increases to <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(R^2\)</EquationSource> </InlineEquation> = 0.8861 and MSE = 0.0077, which shows the importance of dimensionality reduction. Lastly, VAST reaches the state of the art when it optimizes using SMOA, with <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(R^2\)</EquationSource> </InlineEquation> = 0.9605 and MSE = 5.91 <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(\times 10^{-6}\)</EquationSource> </InlineEquation>, outperforming conventional optimization approaches, Particle Swarm Optimization (PSO) and Bat Algorithm (BA). These findings suggest that SMOA and bSMOA can help balance exploration and exploitation, reduce error rates, and enhance model generalization. The implications of these findings are substantial: the optimized VAST framework provides a predictive maintenance tool that enables the cost-effective, swift, and scalable forecasting of renewable energy infrastructures, allowing for more informed decisions in CHP forecasting and minimizing operational risks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Renewable energy consumption forecasting using the Swordfish movement optimization algorithm (SMOA) for feature selection and hyperparameter tuning

  • 期刊论文

  • 发表日期：

    2026年5月29日

查看全文
  • Marwa M. Eid

摘要

The high growth in renewable energy systems has led to increased pressure on effective forecasting procedures to facilitate predictive maintenance as well as promote the reliability of operation of Conventional Hydroelectric Power (CHP) facilities. Even with deep learning, high-dimensional sensor data and sub-optimal tuning of hyperparameters are common challenges to models, resulting in poor forecasting performance. This work presents a new optimization framework that couples the Variable Attention Span Transformer (VAST) with the binary Swordfish Movement Optimization Algorithm (bSMOA) for feature selection and the Swordfish Movement Optimization Algorithm (SMOA) for hyperparameter optimization. The recommended model is compared to existing models, and VAST has achieved a baseline coefficient of determination ( \(R^2\) ) of 0.8353 and a Mean Squared Error (MSE) of 0.0191. After selecting the features using the bSMOA, VAST increases to \(R^2\) = 0.8861 and MSE = 0.0077, which shows the importance of dimensionality reduction. Lastly, VAST reaches the state of the art when it optimizes using SMOA, with \(R^2\) = 0.9605 and MSE = 5.91 \(\times 10^{-6}\) , outperforming conventional optimization approaches, Particle Swarm Optimization (PSO) and Bat Algorithm (BA). These findings suggest that SMOA and bSMOA can help balance exploration and exploitation, reduce error rates, and enhance model generalization. The implications of these findings are substantial: the optimized VAST framework provides a predictive maintenance tool that enables the cost-effective, swift, and scalable forecasting of renewable energy infrastructures, allowing for more informed decisions in CHP forecasting and minimizing operational risks.