Model gain scheduling and reporting for ethylene plant on-line optimizer

An ethylene plant is one of the largest chemical plants. As such, there are frequent changes in hydrocarbon feed mix, individual feed quality and demand for its olefin products. This makes it difficult to operate an ethylene plant in an optimal way. Showa Denko K.K. (SDK) has applied on-line optimization software packages of Honeywell Process Solutions to its ethylene plant. The software packages consist of large-scale model predictive control (MPC) controllers and optimization systems. As a function of the software, furnace yield model gains of MPC are scheduled by using a rigorous first principle model to cope with the strong non-linearity of the reactions in the furnace. In addition, non-linearity of consumption of fuel and steam is important, especially when user demand or storage tank room is limited and productions rates are kept constant. In that case, the direction of optimization of fuel and steam users could be wrong if non-linearity is ignored. The software, however, cannot cope with that non-linearity. Therefore, the achievable performance of the original software is limited. To cope with non-linearity, SDK developed a function of gain scheduling for utility models. As a result, both throughput and efficiency were improved more than expected. Since the on-line optimizer copes with many constraints and checks benefits, it is difficult for operators to intuitively understand the optimization results. Therefore, SDK also developed user interfaces and daily reporting systems for the on-line optimizer. The user interfaces indicate the operating condition and the bottlenecks in the ethylene plant every minute. Thus, operators can recognize the operation determined by the on-line optimizer. In daily reports, process trends of bottleneck points and manipulated variables of the on-line optimizer are reported. The daily reports are automatically generated and sent to members in associated sections by e-mail. With the report, the operation of the ethylene plant can be monitored and fine-tuning of the on-line optimizer can be performed.