生物氧化预处理过程中不确定性数据处理

摘要

氧化槽进气量是生物氧化预处理过程中重要的操作调控参数,由于氧化槽内成分复杂,预处理过程具有强非线性、耦合性和大时滞性的特点,采集到的数据往往包含一定的不确定性.针对这一问题,提出一种基于机会约束的鲁棒支持向量回归方法.通过将支持向量回归模型中常规约束转换成机会约束,得到相应的鲁棒回归模型,将鲁棒回归模型转化为求解二阶锥规划问题.仿真结果表明,所提算法能够有效对数据不确定性进行处理,为生物氧化预处理过程提供新的方法.%Air input of oxidation tank is an important control parameter in the biological oxidation pretreatment process.Toget-her with many other complex factors in oxidation tank,the pretreatment process possesses the characteristics of strong nonlinea-rity,coupling and large time delay.And collected data often contain a certain amount of uncertainty.A robust support vector regression (RSVR) method based on chance constraint was proposed for the problem.The conventional constraints in support vector regression (SVR) model were converted to a chance constraint to get the robust regression model.Both linear and nonli-near formulations were converted to second-order cone programming (SOCP) problems.Simulation demonstrated that the proposed method can deal with uncertain data effectively and provide a new way for biological oxidation pretreatment process.