基于生成机理的燃煤电站锅炉NO_x排放量神经网络模型

ARTIFICIAL NEURAL NETWORKS MODEL FOR NOX EMISSIONS IN A PULVERIZED-COAL POWER STATION BOILER ON GENERATION MECHANISM

到目前为止,对燃煤电站锅炉 NOx 生成规律的研究主要集中在试验和基于化学反应动力学的 CFD 模型研究上,而对基于 NOx 排放规律的人工神经网络模型研究得较少。为数不多的研究者也只是采用人工神经网络黑箱的特点,没有充分应用现已逐渐成熟的 NOx 生成机理。该文基于 NOx的生成机理,针对某燃煤电站锅炉,提出 NOx 排放量的神经网络模型。该神经网络模型具有可以预测各一次风粉单元 NOx 生成量、锅炉 NOx 排放量、网络隐节点数少、泛化能力强、鲁棒性好、学习速度快等优点。所提出的模型可以为大型电站锅炉通过燃烧系统自动调整或结构改造降低 NOx排放提供依据。

Until today, studies on the NOx generation mechanism in a boiler of pulverized-coal power station mainly concentrate on experiments and CFD model on chemical kinetics, but little on artificial neural networks. A few researchers only adopt characteristic of ANN black box without making good use of NOx generation mechanism. This paper presents an ANN model of NOx emissions on this mechanism. The model possesses many advantages such as predicting NOx emissions of each primary air and pulverized coal unite, NOx emissions of boiler, few hidden nodes, strong generalization, good robust and fast learning. The model presented can provide reference for huge power station boiler to reduce NOx emissions by auto-adjustment or reconstruction.