基于主成分分析和长短期记忆神经网络的锅炉燃烧系统模型

Boiler Combustion System Model Based on Principal Component Analysis and Long-Short-Term Memory Neural Network

摘要: 燃烧系统建模是锅炉燃烧优化控制的基础。由于锅炉燃烧系统具有强耦合、多输入输出、时序性输入输出等特点，很难建立符合实际情况的机理模型。本文提出一种基于主成分分析法(PCA)数据降维和长短期记忆神经网络(LSTM)的锅炉燃烧系统模型，其中，PCA方法可以降低输入参数维度以减少数据复杂度，LSTM神经网络能够处理具有时间序列特性的大量数据。与传统的循环神经网络模型相比，模型运算速度和泛化能力都有所提高；与未进行PCA降维处理的LSTM模型相比，运算速度有显著提高。

Abstract: Combustion system model is the basis of boiler combustion optimization control. The boiler combustion system is strongly coupled, with multiple and sequential input/output signals, which is difficult to construct a mechanism model that is close to the actual operation. This paper proposes a boiler combustion system model based on principal component analysis (PCA) data dimensionality reduction and long-short term memory neural network (LSTM). PCA method can reduce the dimension of input parameters to reduce data complexity, and LSTM neural network can process a large number of time series data. Compared with the traditional recurrent neural network model, the operation speed and generalization ability of the model are improved; Compared with the LSTM model without PCA dimensionality reduction, the operation speed is significantly improved.