基于类脑计算的污水处理过程BOD软测量模型研究

基于神经网络的软测量模型是解决污水处理过程的重要评价指标及控制参数-生化需氧量（Biochemical Oxygen Demand, BOD）-在线测量困难问题的关键，然而目前缺少能够模拟人脑各功能脑区交互与协同作用的软测量模型，从而制约着出水BOD的预测精度与实时性，严重影响了污水处理过程的闭环控制。针对该问题，本项目分析污水处理过程特性，提取与出水BOD相关的特征参量；从脑与认知科学的角度出发，建立基于类脑计算的污水处理软测量模型模拟人脑处理复杂信息时各功能脑区的交互与协同作用；针对污水处理过程的时变、不确定特性，研究该软测量模型的结构演化机制，使其结构能够进行自适应动态调整以模拟人脑的结构可塑性，提高污水处理过程出水BOD的预测精度与实时性，为污水处理厂优化运行提供技术保障。该研究工作及成果具有非常广阔的应用前景，对我国人工智能理论发展起着重要的支撑作用。

Soft-sensing model based on neural network plays a key role in solving the problem of online measurement for the important evaluation index and control parameter - Biochemical Oxygen Demand (BOD) - in wastewater treatment process. However, at present, few soft-sensing models can simulate the functional brain regions to interact and cooperate with each other, which limit the prediction accuracy and instantaneity for BOD and have an impact on the closed-loop control of wastewater treatment process. To solve this problem, the project first analyzes the property of wastewater treatment process and extracts characteristic parameters related to BOD. Then, from the view of brain and cognitive sciences, a soft-sensing model based on brain-like computing is constructed to simulate the interaction and cooperation between functional brain regions. Besides, aiming at the time-varying and indeterminate characteristics of wastewater treatment process, we will study the evolution mechanism of the soft-sensing model structure to make it self-adapted dynamically, simulating the structural plasticity of brains. This may improve the prediction accuracy and instantaneity for effluent BOD in wastewater treatment process, and will provide technical support to optimal operation of wastewater treatment plants. The research outcome from this project brings a very broad application prospect, and will play an important supporting role on the theory of the artificial intelligence.