基于一致性数据融合的磨机负荷自适应软测量研究

项目名称： 基于一致性数据融合的磨机负荷自适应软测量研究

项目编号： No.61304118

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 司刚全

作者单位： 西安交通大学

项目金额： 24万元

中文摘要： 性能退化、模型失效和负担过重是软测量技术实际应用过程中面临的最大挑战，严重制约着其应用的长期稳定性和可靠性。本项目研究基于一致性数据融合的自适应软测量技术，旨在解决因运行工况和操作条件时变影响产生的模型性能退化和失效问题，同时改善因模型复杂度高引起的负担过重问题。针对实际工业样本集具有大量相似样本、各工况样本数量不平衡和异方差特征，研究降低模型复杂度、增强鲁棒性的软测量建模方法；针对软测量模型性能下降及失效问题，研究基于一致性数据融合的自适应策略，采用扰动解耦和一致性鲁棒测度的方法实现软传感器性能评估、按需更新以及失效软传感器的准确剔除，充分利用各软传感器信息的冗余及互补特性，提高软测量估计的准确性和可靠性；最终将研究成果应用于火电厂球磨机负荷检测，为磨机负荷优化控制提供准确、可信的负荷信息。本项目可为复杂过程工业参数检测提供一种新颖的自适应软测量方法，为其长期可靠在线应用打下坚实的基础。

中文关键词： 磨机负荷；软测量；数据融合；最小二乘支持向量机；

英文摘要： Performance degradation, model invalidity and overburden are the major challenges in practical application of soft-sensing method which affect its durable stability and reliability in application. In this project, we will study an adaptive soft-sensing method based on consensus data fusion, focusing on how to avoid the model deterioration or invalidity while affected by the variation of working conditions, in the meantime, solve the problem of overburden which caused by model complexity. Considering the abundance of similar samples, the unbalance of the sample numbers and heteroscedasticity of the training data set under various working conditions, we will study the model construction method for soft sensor to reduce complexity and enhance robustness. Considering the performance of soft-sensing model deteriorates or invalid, we will study adaption mechanism based on consensus data fusion technique, to complete performance evaluation, update according to the need and remove the invalid soft sensor by using disturbance decoupling and robust consistency measuring, achieving a more accurate and reliable measuring result by making full use of the characteristics of the redundancy and complementary information of multi soft sensor. The ultimate goal of the research is to use in mill load measurement for providing accu

英文关键词： mill load；softsensing；data fusion；LSSVM；