Adaptive On-line Performance Optimization Of Gas Fractionation Plant

Along with the time of plant operation proceeds, many factors such as the property and price of the feed, specifications and requirement of products, production environment and the nature of units are varying and impossible to keep the same as specified in original design. In these cases, it is necessary to adjust the set points of operation variables and make sure it really works in an optimized state to cope with the changing condition. In this work, we will emphasize our study upon the problem of optimization technique, in particular, we will combine the process analysis with economical analysis as well as production decision, with least or without investment in equipments by this way. The better use of existing equipment will be achieved, the intrinsic power of the plant in order to increase productivity will be explored, and finally to raise the quality of products and decrease the operational cost of the plant. This is the technique called Optimization of Plant Operation (OPO). It has attracted the attention of scientists and engineers from many process industries.The research work of this thesis consists of the following major parts:1.According to the target problem, the current operation status of the separation section of a plant was analysed. The major controllable independent variables, dependent variables and the object function of optimization problem was formed.2.Regarding one of the core work in operation optimization, Both the practical data directly recorded from practical plant and the data obtained by executing the theoretical mathematical model of the process were need to construct different kinds of models to correlate the quantitative relationships among of the independent variables and dependent ones.3.The procedure and methods for variables sieving, data sieving, and model sieving were systematically investigated, and a sequential procedure for applying the multivariate regression analysis to our object plant were determined.4. Considering the requirements in real-time using the theoretical model, thevimapping of the inputs and outputs of theoretical model to a artificial neural network models were constructed.~.Considering that the non-linear statistical models based on multivariate polynomials were not well satisfactory, a set of artificial neural network models were also established directly from the data take from the practical plant.6.Properly selecting a suitable method of optimization is very important to operation optimization. Three optimization algorithms such as linear programming. complex algorithm and genetic algorithm, corresponding to three kinds of models were used to solve the optimization problem according. Moreover, improve complex algorithm. genetic alaorithm were worked out and used in our studies successfully.7.Expert System is introduced in our work, it is used to manipulate different kinds of model to properly solve the problem which occurring in process of optimization.8.A practical on-line adaptive optimization expert system monotype was established for liquefied hydrocarbon separation process.