基于混合机器学习模型的固废基回填材料导热性能预测研究

Prediction of Thermal Conductivity in Solid Waste-Based Backfill Materials Using a Hybrid Machine Learning Model

摘要: 本研究针对地源热泵固废基回填材料导热性能优化中传统实验方法存在的局限性，提出了融合双向长短期记忆网络与智能优化算法的混合机器学习模型。通过系统分析多种固废材料与石墨掺量的耦合影响，利用双向传播机制同步捕捉材料组分的导热衰减与补偿效应，结合智能算法实现模型参数自适应优化，建立了高精度的导热性能预测模型。研究表明，该模型在预测精度和泛化能力方面均表现出显著优势，同时通过可解释性分析清晰揭示了各组分对导热性能的影响规律，为固废基回填材料的性能预测与优化设计提供了有效的预测模型，对推动工业固废资源化利用和地源热泵技术发展具有重要价值。

Abstract: To address the limitations of traditional experimental methods in optimizing the thermal conductivity of solid waste-based backfill materials for ground source heat pumps, this study proposes a hybrid machine learning model integrating Bidirectional Long Short-Term Memory (BiLSTM) networks with intelligent optimization algorithms. By systematically analyzing the coupling effects of various solid waste materials and graphite content, the model utilizes bidirectional propagation to simultaneously capture the thermal attenuation induced by solid waste components and the compensation effects of graphite. Through intelligent optimization algorithms, adaptive tuning of model parameters is achieved, resulting in a high-precision predictive model for thermal performance. The results demonstrate that the proposed model exhibits significant advantages in both prediction accuracy and generalization capability. Furthermore, interpretability analysis clearly reveals the influence patterns of different components on thermal conductivity. This study provides an effective predictive model for the performance prediction and optimal design of solid waste-based backfill materials, contributing significantly to the resource utilization of industrial solid waste and the advancement of ground source heat pump technology.