数据驱动的最优运行状态鲁棒评价方法及应用

在现代复杂工业生产过程中, 细致而稳健的运行状态评价及非优因素识别对指导工业生产具有十分重要的实际意义.考虑到复杂工业过程难以建立准确的数学模型和实际工业过程数据噪声及离群点污染比较严重的问题, 本文提出一种全潜鲁棒偏M估计的复杂工业过程最优状态的鲁棒评价方法.在建立离线评价模型时, 通过对过程数据主元和残差子空间的进一步分解, 提取出能够反映与原材料、生产消耗和产品质量等因素相关的经济指标的变化信息, 同时采用样本数据加权的方法消除离群点对评价模型的不利影响, 提高算法的鲁棒性; 在线评价时, 针对生产过程中存在不确定性因素, 引入在线数据窗口及相似度分析进行在线评价, 并给出在线评价的准则和流程, 提高评价结果的可靠性, 当评价结果非优时, 通过计算相应变量的贡献率识别非优因素.最后, 通过重介质选煤过程验证了所提方法的有效性.

Data-driven Robust Evaluation Method for Optimal Operating Status and Its Application

In the process of modern complex industrial production, a detailed and robust evaluation method of operation state is of great significance for guiding the production. Considering the difficulty to establish an accurate principle model and the process data which are easily polluted by noise and outliers, this paper proposes a robust optimal evaluation method for complex industrial processes based on total partial robust M-regression. In the off-line modeling stage, by further decomposing the principal and residual subspaces of the process data, the process variation information related to the economic indexes reflecting the factors such as raw materials, production consumption and product quality is extracted, and the adverse effects of the outliers are eliminated by sample data weighting to improve the robustness of the algorithm. In the stage of online evaluation, the online data window and similarity analysis are introduced for the uncertain factors of the production process, and the framework and procedure of online evaluation are given to improve the reliability of the evaluation results. If the evaluation results are not optimal, then the non-optimal factors are identified by calculating the contribution rates of the corresponding variables. Finally, the effectiveness of the proposed method is illustrated by a process of dense medium coal preparation.