Scholarlink

<p>Electrical power consumption is a priority area in the optimization of domestic and industrial system performance and sustainability. Despite the application of various machine learning (ML) models in predicting power consumption, it is difficult to improve prediction accuracy and avoid overfitting because of intricate, nonlinear relationships. This study fills these gaps by developing novel hybrid ML models by integrating adaptive boosting regression (ADA) and least square support vector regression (LSSVR) with sooty tern optimization algorithm (STOA) and horned lizard optimization algorithm (HLOH). Hybrid models ADST (ADA + STOA) and ADHL (ADA + HLOH) were developed to achieve highest prediction accuracy and lowest error. Experimental results illustrate that the ADST model performed better with an RMSE of 966.90 and an <i>R</i><sup>2</sup> value of 0.984, an RMSE decrease by 33.1% compared to the baseline ADA (RMSE: 1445.26, <i>R</i><sup>2</sup>: 0.962). Similarly, the ADHL model decreased the RMSE by 13.3% (RMSE: 1252.89, <i>R</i><sup>2</sup>: 0.972). Hybrid models of both outperformed their respective LSSVR-based models. The findings validate that the integration of optimization methods with ML models greatly elevates the accuracy of predictions. From the literature presented, the application of hybrid ML-optimization models like ADST for accurate forecasting of energy consumption is recommended to facilitate sustainable energy management.</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

Integrating machine learning with metaheuristic optimization for power consumption prediction

  • 期刊论文

  • 发表日期：

    2026年2月17日

查看全文
  • Decai Tang

摘要

Electrical power consumption is a priority area in the optimization of domestic and industrial system performance and sustainability. Despite the application of various machine learning (ML) models in predicting power consumption, it is difficult to improve prediction accuracy and avoid overfitting because of intricate, nonlinear relationships. This study fills these gaps by developing novel hybrid ML models by integrating adaptive boosting regression (ADA) and least square support vector regression (LSSVR) with sooty tern optimization algorithm (STOA) and horned lizard optimization algorithm (HLOH). Hybrid models ADST (ADA + STOA) and ADHL (ADA + HLOH) were developed to achieve highest prediction accuracy and lowest error. Experimental results illustrate that the ADST model performed better with an RMSE of 966.90 and an R2 value of 0.984, an RMSE decrease by 33.1% compared to the baseline ADA (RMSE: 1445.26, R2: 0.962). Similarly, the ADHL model decreased the RMSE by 13.3% (RMSE: 1252.89, R2: 0.972). Hybrid models of both outperformed their respective LSSVR-based models. The findings validate that the integration of optimization methods with ML models greatly elevates the accuracy of predictions. From the literature presented, the application of hybrid ML-optimization models like ADST for accurate forecasting of energy consumption is recommended to facilitate sustainable energy management.