پرديس سديفي

Real-time optimization faces challenges due to the need for accurate dynamic models, long computation times, an‎d the necessity of reaching steady-state conditions. In this study, two real-time optimization methods are proposed based on asynchronous steady-state parameter adaptation using the Hammerstein model. These methods are eva‎luated in a laboratory-scale oil production network with gas injection an‎d compared with existing approaches in terms of performance an‎d economic considerations. In both methods, the process model is initially divided offline into steady-state an‎d dynamic submodels. For the dynamic part, the corresponding Hammerstein model is identified. In the first method (asHROPA1), the online phase includes steady-state identification an‎d estimation in the steady-state submodel, an‎d dynamic estimation using the Hammerstein model in the dynamic submodel. The optimal process input is then obtained through steady-state optimization, an‎d this cycle is repeated periodically. In the second method (asHROPA2), the dynamic submodel follows the same procedure as the first method. However, in the steady-state submodel, the identification step is omitted, an‎d dynamic estimation using the Hammerstein model with a zero dynamic matrix is applied. The optimal input is again determined through steady-state optimization. Due to their independence from precise dynamic modeling an‎d short computation times, both methods have demonstrated significant economic benefits. These approaches offer promising solutions for real-time optimization in industrial processes.

بهينه سازي بي درنگ , كنترل فرآيند پيشرفته , مدل همرشتاين

Real-time optimization , Advanced Process Control , Hammerstein model

Pardis Sodeifi

Dr. Seyed Majid Esmaeilzadeh