基于LSTM神经网络的超超临界机组过热器温度预测

Prediction of Superheater Temperature in Ultra-Supercritical Units Based on LSTM Neural Network

提出一种对超超临界火电机组的过热器温度的温度进行预测的方法。首先介绍了火电机组过热器温度的重要性和预测方法的研究现状,然后详细描述了长短期记忆网络(LSTM)的原理和应用。将国内某1000 MW的超超临界火电机组的运行数据作为数据样本,将数据的80%作为LSTM模型的训练组、其余20%作为预测组,在计算中应用了交叉验证的方法。经验证,预测的最大误差为0.8%,平均误差在0.5%。最后,探讨了LSTM模型的不足与可能的改进方法。

This paper proposes a method for predicting the temperature of the superheater in ultra supercritical thermal power units.Firstly,the importance of superheater temperature in thermal power units and the current research status of prediction methods are introduced.Then,the principle and application of Long Short-term Memory Networks(LSTM)are described in detail.The operation data of a 1000 MW ultra supercritical thermal power unit in China is used as the data sample,80%of the data is used as the training group for the LSTM model,and 20%of the data is used as the prediction group.Cross validation method is applied in the calculation.The maximum error predicted after verification is 0.8%,with an average error of 0.5%.Finally,the shortcomings and possible improvement methods of the LSTM model are discussed.