| Process industry has great contribution to the development of our nation and it helps to improve our lives. However, it faces enormous challenges inside and outside the industry. Confronting competitions in quality and quantity of products and environmental aspects, an enterprise must rely on advanced techniques of simulation and optimization and response promptly to market. For the critical demands, people began to study integrated enterprise-wide solutions.In recent years, information technologies, especially the rapid development of computers help simulation and optimization to become a widely used technique in industry. As the operating skills and computer technologies are improved, real-time optimization (RTO) develops rapidly and is put into practical use. There are two kinds of RTO, one is steady-state based and the other is dynamic. Steady-state based RTO is widely used in industry, as it is easier to carry out and maintain.Actually steady-state based RTO is an extension of steady-state simulation. It uses rigid non-linear models and equation-oriented solvers. A necessary condition is that there is no big change in the system's states when carrying out steady-state RTO. This character determines that optimization speed is a critical fact in steady-state RTO, which means that the optimization must be fast enough in large-scale systems otherwise the optimization results cannot be applied. That is to say, in order to applied steady-state RTO in more complicated systems, our solving skills must be further improved. In practice, operators will maintain a set of base cases. When a new field condition comes, the system will be configured using the matched base case if there is one matched. This simple method can improve the performance of RTO and it can improve stability as well. However, according to our research, it still has not utilized all the power of experience data. Experience data can be dynamically updated and used as "base cases", so that to improve the whole optimization process. It can be further used by selecting some of the data and approximate with it for better starting points. Hence, the concept of Mnemonic Enhancement Optimization (MEO) is proposed.In order to apply the MEO theory in practical applications, this article proposed an algorithm framework which is flexible to implement. MEO framework and its application under Aspen Plus are implemented. The performance of implemented MEO framework is verified through different numerical experiments in the two-column flowsheet and the ethylene simulation flowsheet (more than 30 thousand equations). Compared to the traditional method in industry, MEO is superior in solving time and stability. In the experiments of two-column flowsheet, MEO has the same convergence rate as the traditional method, but it solves the problems with about 80% time of that of the traditional method. In the experiments of the ethylene simulation flowsheet, MEO has 100% convergence rate, however, the traditional method has only 67%.The detailed work of this article can be divided into the following parts:Firstly, the background of RTO was introduced, the characters of RTO were studied, and how to improve optimization process and analyzed the possibilities to use former optimization experience in improving solving speed were illuminated. MEO's basic idea was explained and its algorithm framework was proposed.Secondly, Service Oriented Architecture (SOA) was studied and used in the implementation of MEO framework. Middleware techniques were analyzed and the MEO framework was built based on COM. The inner message framework of Windows was analyzed and the MEO's message dispatch mechanism was implemented.Thirdly, MEO framework under Aspen Plus was implemented and a useful OOMF script (used by Aspen Plus) generator was designed. Specific numerical examples were used to illustrate MEO application under Aspen Plus and the results were analyzed. Finally the prospect of MEO and future work were given. |
PDF Full Text Request