基于VMD-IASO-ELM的吸收塔出口SO_(2)浓度组合预测模型

Combined Prediction Model of SO_(2) Concentration at Outlet of Absorber Based on VMD-IASO-ELM

为提高火电厂SO_(2)污染物排放控制水平,提出一种基于变分模态分解(VMD)改进原子搜索算法(IASO)极限学习机(ELM)的吸收塔出口SO_(2)浓度组合预测模型。首先,利用机理和相关性分析确定吸收塔出口SO_(2)浓度的初始相关变量,并采用VMD算法对其分解,保留分解结果与输出互信息中大的低频分量;然后,采用结构简单、学习速度快的ELM建立预测模型,并利用基于混合策略改进的IASO优化网络参数,提高预测精度;最后,利用模糊规则推理出误差修正项以校正ELM模型预测结果。应用历史数据仿真建模,结果表明该模型具有较高的预测精度和学习能力,能够准确跟踪吸收塔出口SO_(2)浓度变化趋势。

In order to improve the emission control level of SO_(2) pollutants in thermal power plants,a combined prediction model of SO_(2) concentration at outlet of absorber was proposed based on VMD-IASO-ELM.Firstly,the mechanism and correlation analysis were used to determine the initial correlation variables of SO_(2) concentration at outlet of absorber,and variational mode decomposition(VMD)algorithm was used to decompose it.The low-frequency components with large mutual information between the decomposition results and the output were retained.Then,the simple structure was established based on fast-learning extreme learning machine(ELM),and the improved atomic search optimization(IASO)based on hybrid strategy was used to optimize network parameters and improve the prediction accuracy.Finally,the error correction term was deduced from fuzzy rules to correct the prediction results of ELM model.Using historical data for simulation modeling,the results show that the model has high prediction accuracy and learning ability and can accurately track the change trend of SO_(2) concentration at the outlet of absorber.