基于数据与先验知识融合生化反应过程辨识建模及其参数不确定优化研究

In view of the characteristcs of parameter uncertainty, nonlinearity, the incompleteness of information and the time delay of the key parameters' measurement in real bioprocess, this project intends to carry out the research on identification model and optimization with parameter uncertainty for the bioprocess driven by the experimental data and priori knowledge. Firstly, to depict the change trend of important variables in bioprocess, the unstructured parameterized model is established based on the mechanism of the bioprocess, and the kinetic parameters are identified by data. And then, the hybrid model is builded up bu using the artifical neural network to compensate the accuracy of mechanism model, which is used to accurate estimates of the state variables in the bioprocess. Thirdly, the multi-objective optimization problem of the bioprocess is presented and solved by a hybrid algorithm by integrating an improved particle swarm optimization with successive quadratic programming, which results in the approximate optimum operating condition and operation scheme of the bioprocess. In addition, in view of the uncertainty factors in the bioprocess, the identification and sensitivity analysis are made to identify the key parameters, and then the optimization model of bioprocess with uncertain parameters is builded up and solved by stochastic optimization algorithm. The last step is to design efficient algorithm, set up control platform to realize the bioprocess control according to the different goals to achieve in the bioprocess.

鉴于生化反应过程具有参数不确定性、过程非线性、信息不完全性和过程关键参数测量时滞性等特点，本项目拟开展基于数据和先验知识融合生化反应过程辨识模型与不确定优化问题研究。首先，为了刻画生化反应过程重要变量的变化趋势，建立了基于先验机理知识的非构造式参数化模型，并对模型参数进行了辨识。其次，利用人工神经网络模型对机理模型精度进行补偿，建立了基于数据和过程机理的生化反应过程融合模型，用于对反应过程状态变量的精确估计。第三，基于融合模型构建多目标优化问题，并利用改进粒子群优化算法结合序列二次规划算法对其进行求解，确定一组过程的最佳运行条件和操作方案。此外,针对反应过程中存在的不确定性因素进行识别和敏感性分析，选定反应过程主要关键不确定性因素，构建参数不确定条件下的反应过程优化模型，并利用随机优化算法进行求解；第四，根据反应过程所要实现的不同目标，设计切实有效的控制算法，搭建控制平台实现对其控制。

本项目围绕基于数据和先验知识融合生化反应过程建模与不确定优化展开研究。首先，为了预测反应过程重要变量的变化趋势，建立了基于先验机理知识和数据的反应过程非结构化模型，借助于连续与半连续动力系统理论与方法分析了模型的动力学性态，并通过构建优化模型确定了反应过程的一组最佳运行条件与操作方案。其次，针对反应过程参数及输入不确定性问题，研究了基于模糊集理论的反应过程参数不确定性度量与性能评价方法，为分析和处理参数不确定性及其影响提供了理论基础。第三，针对反应过程输入不确定性及外部随机扰动，开展了自适应控制方法的研究，主要用于实现对反应过程的稳定控制。本课题的研究成果为解决生化反应过程建模和控制积累了一定的研究经验和技术基础。