基于变分模态分解和SSA-LSTM的SCR脱硝系统入口NO_(x)浓度预测

Prediction of Inlet NO_(x)Concentration in SCR Denitration System Based on Variational Mode Decomposition and SSA-LSTM

燃煤机组的SCR入口NO_(x)浓度测量具有较大迟延、影响因素复杂和波动性大等特点,往往不能真实反映NO_(x)浓度的实时变化。针对上述问题,提出一种基于变分模态分解(VMD)和SSA-LSTM的SCR入口NO_(x)浓度的预测模型。首先采用变分模态分解法分解SCR入口NO_(x)浓度,互信息选择算法选择与目标变量强相关的辅助变量;然后利用SSA算法寻优LSTM神经网络参数构建SSA-LSTM预测模型;最后进行VMD-SSA-LSTM与LSTM、VMD-LSTM的仿真对比实验。结果表明,VMD-SSA-LSTM预测模型具有更高的预测精度、更小的误差和更强的泛化能力。

The measurement of NO_(x)concentration at the SCR inlet of coal-fired power units has the characteristics of significant delay,complex influencing factors,and high volatility,often unable to accurately reflect the real-time changes in NO_(x)concentration.A predictive model for SCR inlet NO_(x)concentration based on variational mode decomposition(VMD)and SSA-LSTM is proposed to address the aforementioned issues.Firstly,the variational mode decomposition method is used to decompose the NO_(x)concentration at the SCR inlet,and the mutual information selection algorithm selects auxiliary variables that are strongly correlated with the target variable.Then,the SSA algorithm is used to optimize the LSTM neural network parameters and construct an SSA-LSTM prediction model.Finally,conduct simulation comparative experiments between VMD-SSA-LSTM,LSTM,and VMD-LSTM.The results indicate that the VMD-SSA-LSTM prediction model has higher prediction accuracy,smaller errors,and stronger generalization ability.