非平稳噪声条件下软测量系统量子随机滤波方法研究

项目名称： 非平稳噪声条件下软测量系统量子随机滤波方法研究

项目编号： No.61273069

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 金福江

作者单位： 华侨大学

项目金额： 80万元

中文摘要： 本项目针对实际软测量系统的非平稳噪声概率密度函数（PDF）难以准确估计，对软测量结果影响较大的关键问题，提出采用量子随机滤波和即时核学习方法进行研究。提出核学习方法建立势场函数模型，用绝热量子计算代替Schrodinger方程，应用递归贝叶斯得到状态变量后验PDF的即时核学习量子随机滤波方法。分析了势场函数与PDF模型之间的关系，论证了滤波算法的稳定性、收敛性和参数的灵敏度。最终，实现在非平稳随机噪声下软测量系统的最优滤波。通过实验验证微观建模和量子随机滤波算法的可行性和有效性。

中文关键词： 软测量；量子随机滤波；概率密度函数；即时核学习；薛定谔方程

英文摘要： It is difficult to accurately assess the probability density function (PDF) of non-stationary noises in actual soft sensor systems and consequently the prediction results of most soft sensor models generally degrade. The project proposes to adopt the quantum stochastic filtering and just-in-time kernel learning to research the mechanism of state function of stochastic signal. The model of potential field function is first established using the kernel learning approach. Then, the adiabatic quantum computation method is adopted to solve the complex Schrodinger equations. The recurrent Bayesian inference can be utilized to obtain the post probability of PDF of state variables with the recently updating models using just-in-time kernel learning. The relationship between the potential field function and PDF is analyzed. Also, the stability and convergence of the filtering algorithm, and the sensitivity of related parameters are investigated. Finally, the optimal filtering for state variables of soft sensor systems with non-stationary stochastic noises is achieved. Additionally, an efficient experimental scheme is designed, which can be used to verify the feasibility and validity of microcosmic modeling method and quantum stochastic filtering algorithm.

英文关键词： soft sensor；Quantum stochastic；Probability density function；Just-in-time kernel learning；Schrodinger equation