Hybrid Modeling For Soft-sensor Of Carbon Content Of Catalyst In FCC

The main technology of catalytic cracking in China is FCC(Fluid Catalytic Cracking).The real-time and accurate measurement of the carbon content of catalyst in FCC has great significance in ensuring a safe production environment,improving the production of FCC and prolonging the service life of catalyst..Based on the existing research in catalytic cracking technology,this paper develops a soft sensing modeling work for carbon content of catalyst in MIP-CGP reaction-regeneration device.To achieve accurate and real-time measurement requirements,and to solve problems of other industrial soft sensor modeling methods,this paper mainly presents a deep learning based soft-sensor modeling method for carbon content of catalyst in Fluid Catalytic Cracking.Firstly,a data-based modeling method for soft-sensor of carbon content of catalyst in FCC is studied in this paper.In order to build an accurate soft sensing model,by combining the production process and mechanism analysis,this paper selects primary variables and auxiliary variables of the soft sensing model.Then a continuous deep belief network is set up to process the combination of data features.After that,a LSSVR optimized by a heuristic algorithm is used as a regression layer model to output the final predicted value.This paper takes MAPE and MSE as the measurement accuracy verification indexes.Secondly,in order to improve the performance of carbon deposition softsensor in FCC industry,this paper continues to explore the problem of building the data-based model of soft sensing,and studies the series hybrid modeling method of soft-sensor.To optimize the accuracy and generalization of databased model,this paper analyzes the mechanism of FCC reaction-regeneration system,and then builds a mechanism model of reaction-regeneration system based on pseudo-component.Then a soft sensing hybrid model with series combination method is set up in paper.By comparing with data-based modeling method,the series hybrid modeling method has a remarkable improvement in measurement performance.Finally,to improve the digitization level of refinery,this paper develops a production monitoring platform for key process indicators of FCC reactionregeneration production system,and applies the soft sensing hybrid model to the monitoring platform.The monitoring platform is designed as browser/server architecture and developed based on.NET framework.This platform provides the monitoring function of the key indicators in production process,and designs the function realization of the hybrid soft sensor.This monitoring platform has the ability to provide data guidance for enterprise production.