基于门控循环单元神经网络的浆液pH的预测建模

Predictive Modeling of Slurry pH Based on Gated Recirculating Unit Neural Network

吸收塔内浆液的pH是影响燃煤电厂湿法脱硫系统效率的重要参数。燃煤电厂的湿法脱硫系统具有大滞后、非线性、强耦合等特征,因而其吸收塔浆液的pH很难实现精准控制。利用门控循环单元(gated recurrent unit,GRU)神经网络在处理时间序列数据的优越性,对吸收塔内的浆液pH进行预测建模,通过将燃煤电厂采集的影响浆液pH的变量数据作为模型的输入,对模型进行训练处理,获得吸收塔内浆液pH的预测模型。将预测模型应用于辽宁省华能营口电厂600 MW机组湿法脱硫智能控制系统中吸收塔内浆液pH的预测。结果表明相比于反向传播(back propagation,BP)神经网络模型、径向基函数(radial basis function,RBF)神经网络、循环神经网络(recurrent neural network,RNN)和长短期记忆(long and short term memory,LSTM)神经网络,该模型精确度更高,实用性更强。

The pH value of slurry in absorber is an important parameter affecting the efficiency of wet desulfurization system in coal-fired power plant.The wet desulphurization system of coal-fired power plant has the characteristics of large hysteresis,nonlinear and strong coupling,so it is difficult to accurately control the pH value of the absorber slurry.The advantages of gated recurrent unit(GRU)neural network in processing time series data were utilized to predict and model the pH value of the grout in the absorber.The variable data that affect the pH value of the grout collected from coal-fired power plants were used as the input of the model to train the model.The prediction model of slurry pH value in absorber was obtained.The prediction model was applied to the prediction of pH value of slurry in absorber in the intelligent control system of wet desulfurization for 600 MW unit of Huaneng Yingkou Power Plant in Liaoning Province.The results show that compared with the back propagation(BP)neural network model,radial basis function(RBF)neural network,recurrent neural network,RNN and long and short term memory(LSTM)neural networks.This model is more accurate and practical.