Scholarlink

<p>The development of intelligent power distribution, demand-side management, and smart building operations depends on precise forecasting of electrical energy consumption at the appliance level. This study proposes a novel hybrid predictive framework that incorporates multi-source sensor data from actual residential and commercial settings. A dual-phase metaheuristic optimization process using the Non-Monopolize Search (NMS) and Black-Winged Kite Algorithm (BWKA) further enhances the framework’s core, which combines two cutting-edge ensemble learning models: Extreme Gradient Boosting Regressor (XGBR) and Light Gradient Boosting Regressor (LGBR). Significant gains in forecasting accuracy and model generalizability are shown by the suggested hybrid models. Notably, the optimized XGBW configuration reduced the Root Mean Squared Error (RMSE) by 54.3% and obtained a coefficient of determination (R<sup>2</sup>) of 0.9865 when compared to baseline models. These improvements highlight how well the hybridization approach captures intricate, nonlinear patterns of energy consumption under various operational and environmental circumstances. The framework offers high predictive accuracy along with intelligent system features like adaptive energy control, peak load estimation, and anomaly detection. To increase model transparency and facilitate well-informed decision-making, a sensitivity analysis based on SHAP is performed to determine and interpret the impact of individual sensor features. This work represents an important step toward robust, interpretable, and deployable energy forecasting systems despite ongoing issues with computational scalability in large-scale deployments. The suggested framework has great potential for integration into IoT-enabled smart energy infrastructures, encouraging effective resource allocation and improved system reliability.</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

A Novel Hybrid Framework for Forecasting Appliance-Level Electrical Energy Consumption Using Multi-Sensor Data in Smart Systems

  • 期刊论文

  • 发表日期：

    2025年8月21日

查看全文
  • Anupam Yadav,
  • Hardik Doshi,
  • B. Jayaprakash,
  • Sarraa Ahmad Qahtan,
  • Laith Saheb,
  • Mayank Kundlas,
  • B. Bharathi,
  • Prabhat Kumar Sahu,
  • Satvik Vats

摘要

The development of intelligent power distribution, demand-side management, and smart building operations depends on precise forecasting of electrical energy consumption at the appliance level. This study proposes a novel hybrid predictive framework that incorporates multi-source sensor data from actual residential and commercial settings. A dual-phase metaheuristic optimization process using the Non-Monopolize Search (NMS) and Black-Winged Kite Algorithm (BWKA) further enhances the framework’s core, which combines two cutting-edge ensemble learning models: Extreme Gradient Boosting Regressor (XGBR) and Light Gradient Boosting Regressor (LGBR). Significant gains in forecasting accuracy and model generalizability are shown by the suggested hybrid models. Notably, the optimized XGBW configuration reduced the Root Mean Squared Error (RMSE) by 54.3% and obtained a coefficient of determination (R2) of 0.9865 when compared to baseline models. These improvements highlight how well the hybridization approach captures intricate, nonlinear patterns of energy consumption under various operational and environmental circumstances. The framework offers high predictive accuracy along with intelligent system features like adaptive energy control, peak load estimation, and anomaly detection. To increase model transparency and facilitate well-informed decision-making, a sensitivity analysis based on SHAP is performed to determine and interpret the impact of individual sensor features. This work represents an important step toward robust, interpretable, and deployable energy forecasting systems despite ongoing issues with computational scalability in large-scale deployments. The suggested framework has great potential for integration into IoT-enabled smart energy infrastructures, encouraging effective resource allocation and improved system reliability.