To address the existing problems in Ethernet measurement and control networks, such as real-time and reliability caused by the collision of measurement and control data, a double-layer compression method based on Partial Least Square Regression-self Based Regression(PLSR-SBR) was induced to the IP Mode Measurement and Control System(IMMCS). In the first layer, the model of main parameters and all auxiliary parameters was built up based on PLSR.The principal score of main parameter was determined by the efficient compression index to finish the information compression of main parameter. In the second layer, the auxiliary parameters were compressed based on a modified SBR to set up a basic signal. To ensure the fitting error to be smaller than the setting value, every auxiliary parameter sequence in the auxiliary parameters or the decomposed sequences were mapped in turn to a definitive basic signal to complete the data compression of the auxiliary parameters. Innovative technologies for some key problems were addressed in the method, including the confirmation of maximum correlation degree of interpreting latent variables and reflecting latent variables, basic signal with least basic sequences, the minimum number of decomposed sequences from auxiliary parameters and the independent update rule of basic signal. Finally, experiments using IMMCS for alcohol concentration were performed. The results show that under the conditions that IMMCS possesses main parameters and auxiliary parameters, and simultaneously data variety of parameters is wide, the compression ratio based on the method reaches 68%,which is higher than that based on the Swinging Door. The data collision of Ethernet measurements and control networks are decreased effectively.