An Identification Method of Production Behavior of Enterprises Based on Neural Networks

The production behavior model of non-electricity polluting enterprises is complex and difficult to identify accurately. Aiming at the time series, multiplicity, and mutation position independence of temperature and vibration data in the production process of enterprises, an LSTM-CNNSE2 model is designed. The model extracts the temporal features on continuous data through a long short-term memory network (LSTM), mines the spatial features of the data with a convolutional neural network (CNN), and further calibrates the spatial features using a squeeze-excitation block (SE-block). The temporal and spatial features are fused, and the fusion features identify the production behavior. The actual monitoring data are used for experimental verification, and the results show that the identification accuracy of this method is higher than 94% under different training rounds, which is superior to other methods. It has a certain reference value for non-electricity polluting enterprises’ production behavior identification and environmental protection process management.