LAUSR

Abstract Production capacity identification and analysis in industrial processes plays a more and more important role at home and abroad, which not only can improve the energy efficiency, but also reduce the carbon emission. However, the data of complex industrial processes exist multi-dimension and strong noise, which make traditional linear models difficult to identify and analyze the production capacity. Therefore, this paper proposes a novel production capacity identification and analysis method based on multivariate nonlinear regression (MNR) integrating the affinity propagation (AP) clustering algorithm (AP-MNR) for energy saving and resource optimization. The elements that mainly affect the production capacity are extracted by the AP algorithm. Then the extracted elements and the final yield are set as inputs and outputs to build the production capacity identification model by using multivariate nonlinear regression method. At last, the AP-MNR method has been applied for energy saving and resource optimization of actual ethylene and PTA industrial processes. The evaluation indexes with the goodness of fit in ethylene and PTA industrial processes are 0.984 and 0.993, which have proved the effectiveness of the proposed method. Furthermore, the reasonable resource allocation of complex industrial processes can be optimized to achieve energy saving and carbon dioxide emission reduction.