摘要
预测脱硫系统出口SO_(2)浓度对于计算脱硫效率、提升电厂运行经济性具有重要意义。针对出口SO_(2)浓度受较多因素影响且难以准确测量的问题,提出了基于KPCA和SSA-BP神经网络的脱硫出口SO_(2)浓度预测模型。首先,利用核主元成分分析法(KPCA)计算各影响因素的贡献率,并对采集数据进行降维,提取特征信息。其次,将降维后的数据作为麻雀算法改进的BP神经网络预测模型的输入变量,对脱硫出口SO_(2)浓度进行预测。针对BP网络参数初始化的问题,引入了麻雀算法对网络权值、阈值等进行寻优,以优化预测模型性能。最后,通过电厂实测数据验证模型性能。结果表明,所提SSA-BP模型的RMSE为0.197 4,MAPE为3.053 4%,预测精度高,具有工程应用价值。
Predicting the SO_(2) concentration at the outlet of the desulfurization system is of great significance for calculating desulfurization efficiency and improving the operational economy of power plants.This paper proposes a desulfurization outlet SO_(2) concentration prediction model based on KPCA and SSA-BP neural networks to address the issue of export SO_(2) concentration being affected by multiple factors and difficult to accurately measure.Firstly,the contribution rate of each influencing factor is calculated using Kernel Principal Component Analysis(KPCA),and the collected data is reduced in dimensionality to extract feature information.Secondly,the reduced dimensional data is used as the input variable for the improved BP neural network prediction model of the sparrow algorithm to predict the SO_(2) concentration at the desulfurization outlet.To address the issue of parameter initialization in BP networks,the sparrow algorithm is introduced to optimize network weights,thresholds,etc.,in order to improve the performance of the prediction model.Finally,the performance of the SSA-BP model was validated through actual testing data from the power plant.The results showed that the proposed model had an RMSE of 0.1974 and a MAPE of 3.0534%,indicating high prediction accuracy and engineering application value.
作者
张永强
郭锦涛
ZHANG Yongqiang;GUO Jintao(Guoneng Longyuan Environmental Protection Co.,Ltd.,Beijing 100089,China)
出处
《自动化应用》
2025年第15期291-294,共4页
Automation Application
