Recurrent Neural Network-based Performance Prediction Model for Compressor Units
The monitoring values of key parameters of compressor unit operation are the primary factors for determining whether or not the units are operating normally.However,at each individual grass-roots unit station,lots of data are only stored as backup files for post-incident investigation,and are not effectively used.A compressor unit performance prediction model based on recurrent neural networks is proposed.The monitoring data values are first discretized using data nodes to reduce data redundancy.Then,the correlation among different parameters at different time points are obtained according to correlation coefficient.By mining neighboring nodes,a set of temporal neighboring nodes for the key parameters are obtained,and it is used as the training set for the recurrent neural network to predict the values of the key parameters,and it is used to judge the operating status of the units.Experimental validation is carried out with SCADA dataset.The result indicates that the proposed model performs well in predicting the values of the outlet temperature.The evaluation metrics,MAE and RMSE,for a typical abnormal event in the outlet temperature are 0.88 and 0.92,respectively.These results further indicate that the proposed model exhibits strong generalization ability and predictive accuracy.
compressor unitrecurrent neural networkbig datadata collecting and monitoring systemprediction model
万方数据
维普
