At present,there are few technologies for predicting the life cycle of residual oil hydrogenation catalyst.The research on the solution scheme which can conveniently detect the active state of catalyst will effectively assist the catalyst replacement management of residual oil hydrogenation unit.According to various factors affecting catalyst activity,direct variables such as the running time,raw materials and product properties of device operation data were selected in the process of machine learning modeling,and combined variables such as metal deposition amount and total processing load were added as the features through calculation to carry out algorithm screening and parameter adjustment,and fit the temperature rise law in device running cycle process.After training,mean absolute percentage error of the model for predicting the reaction temperature for each test cycle is 0.51％,and then the catalyst life expectancy can be obtained through the experience threshold of reaction temperature.Considering the possible distribution differences between production and experiment data,the software system designed and implemented according to the research results is flexible in operation,providing convenient and effective auxiliary research tools for professional engineers and technicians.
关键词
渣油加氢/催化剂寿命/反应温度预测/机器学习/系统实现
Key words
residual oil hydrogenation/catalyst life/reaction temperature prediction/machine learning/system realization
维普
万方数据