Detection of Abnormal Energy Efficiency for HEN Based on Hybrid PCA with CNN-SVM

Energy efficiency of a heat exchanger network(HEN) not only reflects the operating conditions but also influences the economic benefits. However, the dynamic and complex operating conditions of HEN make it challenging to calculate the energy efficiency directly. This paper introduces an abnormal detection approach to enhance the energy efficiency. The principal component analysis(PCA) is applied to reduce the dimensionality of HEN logistics data. Then, the reduced dimensional data is passed through convolutional neural network(CNN) to extract digital features. Consequently the stable detection of abnormal energy efficiency state of HEN is realized by support vector machine(SVM). The proposed method is applied to a HEN of the crude distillation unit before desalination. The simulation results demonstrate that the proposed method can achieve effective detection of the energy efficiency status of the HEN by maintaining the accuracy above 95% for all test samples, with the mean square deviation less than 0.005.