由于卸料后带面清扫效果不稳定，矿用带式输送机在长期运行过程中，常面临带面物料粘结、扬尘和撒料等问题，直接影响了运输安全性能及能量高效利用。针对这一问题，提出了一种基于数字孪生技术（DT）的高效、环保的智能带式输送系统。在虚拟实体中，应用在线极限学习机（OS-ELM）构建清扫力实时预测模型，以物理反馈数据作为该模型的输入参数；在物理实体中，结合力学模型、工程领域的知识与经验和运行数据计算出实际清扫力。并基于当前时刻清扫力预测值制定下一时刻清扫执行机构的调度方案，实现对物理实体的精准控制。为保证调度过程实时性，该方案中的参数优化采用改进鲸鱼优化算法（IWOA），以较高收敛速率获取全局最优解。研究以撒料量、关键部件磨损情况、清扫机构总功率和清扫力预测准确率作为DT系统性能的评价指标，在多工况下与多组主流优化算法对比验证该系统性能，并通过Sobol和Fast两种敏感性分析方法验证预测模型输入参数合理性。实验结果表明，基于DT的智能输送系统中煤炭撒料量低于100 g/min，刀片磨损率下降了8.99%，实际功率降低了8.61%，清扫力预测的平均绝对百分比误差（MAPE）和均方根误差（RMSE）分别为3.274 8%和0.017。该DT系统建立了虚拟与现实之间动态信息交互映射，实现了带式输送系统清扫力的实时精准获取并合理调度多级清扫执行机构，有效减少了设备维护工人的工作量和环境污染问题的发生，显著提升了输送设备的安全性和稳定性。

The performance inconsistency in belt cleaning post-discharge presents significant challenges in long-term usage of mine belt conveyors. Issues include material adhesion, dust dispersion, and material spills, posing a direct impact on transportation safety and the effective utilization of energy. Responding to this problem, we propose an effective and environmentally friendly intelligent conveyor system leveraging Digital Twin (DT) technology. In this conceptual model, an Online Sequential Extreme Learning Machine (OS-ELM) is applied within the virtual component to construct a near real-time model predicting sweeping force, using physical feedback data as input parameters. Within the physical counterpart, the sweep force in practice is computed by integrating mechanical models, intellectual and experiential knowledge from the engineering field, as well as operational data. Moreover, forecast values of sweep force at the current moment help formulate the dispatch plan for the cleaning mechanism in the next moment, thus ensuring precise control of the physical model. The Improved Whale Optimization Algorithm (IWOA) is employed to optimize parameters in this blueprint, facilitating real-time scheduling by rapidly converging to the global optimum. In evaluating system performance, we establish key performance indicators: the amount of spillage, wear of crucial components, total power used by the cleaning mechanism, and accuracy in predicting the sweeping force. We compare our system’s performance under various operational scenarios against an array of common optimization algorithms. Simultaneously, we use Sobol’s method and the Fast sensitivity analysis to verify our model's input reasonability. Results show that the DT-based intelligent conveyor system reduces coal spillage to less than 100 g/min, decreases blade wear rate by 8.99%, cuts actual power consumption by 8.61%, with Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE) in sweep force prediction being% and 0.017 respectively. The innovative solution establishes dynamic informational mapping between the virtual and real worlds. This process helps gain precise real-time readings of sweeping forces in conveyor systems and enables the reasonable dispatch of multiple levels of sweeping mechanisms. Consequently, it significantly lightens maintenance workloads, mitigates environmental pollution, and greatly enhances the safety and stability of conveyor equipment.