基于粒子群优化双支持向量机的SCR烟气脱硝效率预测模型

Model for predicting SCR flue gas denitration efficiency based on particle swarm optimization and twin support vector machine

为了更好地分析燃煤电厂脱硝效率与相关影响因素之间的非线性关系,引入双支持向量机(TWSVM)作为分类预测框架,利用粒子群算法(PSO)对TWSVM的惩罚因子和核参数进行寻优,进而构建了针对选择性催化还原(SCR)烟气脱硝效率预测的PSO-TWSVM模型。通过提取某电厂工况监控系统近期数据,结合三倍标准差检验法与归一化法对数据进行预处理,并选取训练集与测试集。仿真结果表明:SCR烟气脱硝效率预测PSO-TWSVM模型最大相对误差小于±0.6%,平均相对误差保持在±0.3%以下,证明该模型的准确性高;对比支持向量机(SVM)模型发现,PSO-TWSVM模型既提高了预测精度,也节省了计算时间。

In order to better analyze the nonlinear relationship between the denitration efficiency of coal-fired power plants and related factors, the twin support vector machine （TWSVM） was introduced as a classification framework. The particle swarm algorithm （PSO） was used to optimize the penalty factor and kernel function parameter of the TWSVM, and then the PSO-TWSVM model for predicting the SCR flue gas denitration efficiency was built up. Through extracting recent data of the power plant monitoring system and combining with three times standard deviation test and normalization method for data preprocessing, the training set and test set can be obtained. The simulation results show that, the maximum relative prediction error of the PSO-TWSVM model is no more than ±0.6%, and the average relative error is about ±0.3%, which proves the high efficiency of the model. In contrast with the SVM model, the PSO-TWSVM model not only improves the prediction accuracy, but also saves time.