【作者】 李淋雨； 邱寅祺； 何星；

【Author】 Linyu Li;Yinqi Qiu;Xing He;Department of Automation, Shanghai Jiao Tong University, and Key Laboratory of System Control and Information Processing, Ministry of Education of China;Power Station Automation Department of Shanghai Minghua Power Technology Corporation;

【机构】 上海交通大学自动化系系统控制与信息处理教育部重点实验室； 上海明华电力科技有限公司电站自动化部；

【摘要】 NOx的排放测量具有时滞特性,受氧量、烟温、磨煤量等多方面因素影响,这给烟气出口处污染物的预测及控制带来了较大的挑战。为了有效预测出口硝化物浓度,本文引入基于长短期记忆神经网络(Long Short-Term Memory,LSTM)预测模型,采用相似度比较算法筛选出对出口硝化物浓度有影响的主要变量,并依次作为神经网络的输入变量。文中针对LSTM网络隐藏层进行讨论建立最优LSTM预测模型。以上海某火电厂实际数据进行预测分析,结果表明所提方法具有较高的模型预测精度,达到实际应用要求。更多还原

【Abstract】 The NOx emission measurement has time-delay characteristics and is affected by oxygen, smoke temperature, coal grinding capacity and other factors, which brings great challenges to the prediction and control of pollutants at flue gas outlet. In order to effectively predict the concentration of nitrification at the exit, this paper introduces the prediction model based on Long Short-term Memory(LSTM), and uses the similarity comparison algorithm to screen out the main variables that have an impact on the concentration of nitrification at the exit, and takes them as the input variables of the neural network in turn. In this paper, the LSTM network hiding layer is discussed and the optimal LSTM prediction model is established. Based on the actual data of a coal-fired power plant in Shanghai, the prediction results show that the proposed method has high model prediction accuracy and meets the requirements of practical application.更多还原