| 초록 |
Since the physical properties of distillates from the vacuum distillation unit (VDU) determine the performance of the lube base oil plant, it is imperative to control the kinematic viscosity at 100℃ (KV100) and distillation (D5%, D95%). However, in the commercial plant, product qualities have been manually controlled depending on the insights of plant operators, therefore, it is required to develop the automated model-based control system. In this study, we formulate and solve the real-time optimization problems to find the operating conditions that satisfy the setpoint of the product specification based on the inferential sensor. The inferential sensor is developed using the one-year time series data of VDU and it predicts the properties of distillates from the process variables such as properties of inlet materials and operating conditions. It predicts KV100, D5%, and D95% within a mean absolute error of 0.035cst and 2ºC, respectively. Optimization is performed to track the setpoint in response to changing inlet properties in real time. We verify the applicability of the model and evaluate the performance of optimization quantitatively, through the commercial plant test. |
