偏最小二乘回归神经网络模型在燃煤锅炉结渣预测中的应用

偏最小二乘回归方法能较好地解决自变量之间的严重相关性问题，笔者将偏最小二乘回归与神经网络耦合，建立了克拉玛依市油田公司某燃煤供热锅炉结渣预测模型。利用偏最小二乘法对影响锅炉结渣的诸多因素进行分析，提取对因变量影响强的成分，从而克服了变量之间的多重相关性问题，降低了神经网络的输入维数。同时，利用神经网络建模可以较好地解决非线性问题。结果表明，预测值与实际值很接近，耦合模型的拟合和预报精度均优于独立使用偏最小二乘回归或神经网络建模的精度。模型对于提高燃煤锅炉的安全运行具有重要的指导意义。

Application of neural network model with partial least-squares regression on predicting slagging of coal-fired boiler

Partial least-squares regression method can resolve serious muhicollinearity problems among variables. A model for predicting slagging of coal-fired boiler based on the combination of neural network and partial least square method is proposed in oil field of Kelamayi. The factors affecting the slagging of coal-fired boiler are analyzed by means of partial least square method to extract the most important components so that not only the problem of multicorrelation among variables can be solves but also the amount of input dimensions of the neural network can be reduced. Besides, the application of neural network helps to solve the problem of non-linearity of the model. The resuits indicate that the prediction values approximate to the real values, the proposed model has higher precision than those models based on neural network method or partial least square method only. The model plays an important guiding role in improving operating safety of coal-fired boiler.