| Batch process is a very important industrial production,which is widely used in high-added-value production,such as chemical,metallurgy,food,metalworking,biopharmaceutical and other fields.With the development of industry,operation optimization is furtherly popular with batch process.Batch process optimization can improve product quality and the economic benefits of enterprises.The data-driven model has been widely used in batch optimization with its convenience and rapid advantages.This method needs a large number of data.However,the new batch process data is insufficient,which makes it difficult to used data-driven model.There are many similar processes in the industrial process,which contains a lot of data.Traditional data method cares about datas which come from plant itself and ignore the existence of multiple plants.This paper mainly studies the optimization method of batch process operation based on process transfer method.This strategy improve product quality of the batch process and the comprehensive economic benefit of the industrial enterprises.The main research contents are summarized as follows:(1)This paper takes batch process as the research background.Firstly,it introduces the characteristics of batch process and the significance of batch process optimization.Then,it introduces multivariate statistical regression modeling methods,such as PCA,PLS,EPCA,JY-PLS.In order to solve the new batch process’ data insufficient,this paper uses joint-y partial quality least squares(JY-PLS)method in the construction of batch process model,the old process’ data is transfered to the new process to bulit new process model.In order to adapt to the characteristics of the batch process,the new available data is used for the online update.The process transfer model lays a theoretical foundation for the operation optimization.(2)Process transfer model can make better use the data of new process model,but there is a difference between old process and new process,which result in necessary conditions of optimality(NCO)mismatch.In order to solve this problem,a modified-adaptive optimization strategy based on process transfer is proposed in this paper,and a good operation optimization effect is obtained.Firstly,we get the predictive quality and the actual quality of the batch,then calculate the error between them,and use the optimization algorithm to calculate the optimal input variables of the next batch.In order to control the variable in the effective range,the T2 variable is used as the construction of the restricted variable.With the operation of batch process,the optimization process is divided into three stages to solve the differences of data from the old batch process(1)New data supplement stage,this stage is in the initial stage of the new batch process optimization,(2)The new process data satisfies the models’ requirement,the newly available data are used to replace old data one by one(3)The optimization model is transformed into a single process’ optimization model.The proposed algorithm is applied to the crystallization process of cobalt oxalate and achieve good results.(3)In order to solve real-time effect and deviate from the normal range in batch operation,a real-time correction adaptive operation optimization method based on process transfer is proposed in this paper.Based on the batch-to-batch optimization method,within-batch operation period uses mid-course correction(MCC)optimization method to furtherly divided into a number of decision points,at each decision point,process optimization is used to solve the mismatch.A method of compensation is used in batch correction.The proposed method is applied to the simulation of cobalt oxalate crystallization process,which verifie the effectiveness of the proposed operation optimization method. |
