Research Initiation: A Bilevel Programming Methodology for Optimizing Control of Chemical Processes

工业过程控制和操作的目标 都是以层级模式设置的。 短期、随机、快速 波动由监管机构处理， 选择的经营方针，并在底部的控制 等级制度。 中期瞬变是一个 优化层位于层次结构的中间， 响应操作策略、过程变化的设定点 性能和饲料质量。 长期的变化是由一个 位于层次结构顶部的生产计划层， 分配饲料，分配生产率， 产品组合，以应对不断变化的市场条件。 因此，在本发明中， 该过程通过以下方式与较高层的故障绝缘： 存在较低的，例如，如果优化计算机 向下，操作仍将继续下的方向， 调节器 过去这些控制方案的设计 假设监管机构的行动比监管机构快得多， 优化器，因此优化器可以忽略 调节器 这项研究的主要目标是开发一种新的 优化控制框架，明确说明 优化器和调节器之间的相互作用。 继续 由于库存减少， 在“作为”的基础上， 需要”的基础。 如果一种商品可以从几个生产商那里买到， 被用作饲料，这引入了一个重要的 变化成一个过程。 这种更高频率的饲料变化 需要优化器更频繁地执行操作。 然而， 对工业过程进行更快速的优化， 动力学推动优化器和调节器的时间尺度 更接近，使一个优化器忽略的行动， 调节器不足。 拟议的核算 优化器和调节器之间的相互作用， 在这种情况下，使用称为 二层程序 PI建议制定一个有效的 算法来解决这个优化问题。

The objectives of control and operation of industrial processes are set up in a hierarchial mode. Short-term, random, rapid fluctuations are handled by regulators which implement some chosen operating policy and are at the bottom of the control hierarchy. Medium-term transients are the domain of an optimizing layer in the middle of the hierarchy, which produces setpoints to respond to changes in operating policy, process performance and feed quality. Long-term changes are managed by a production planning layer at the top of the hierarchy, which allocates feeds, assigns production rates and determines the best mix of products in response to changing market conditions. Thus, the process is insulated from a failure of a higher layer by the presence of the lower ones, e.g., if the optimizing computer goes down, operation will still continue under the direction of the regulator. These control schemes in the past were designed assuming that the regulator acts much more quickly than the optimizer, so the optimizer can ignore the dynamics of the regulator. The principal goal of this research is to develop a novel optimizing control frame work that explicitly accounts for the interplay between the optimizer and the regulator. Continuing pressures to operate with reduced inventories have caused feedstocks to be purchased in smaller quantities, on an "as needed" basis. If a commodity available from several producers is used as a feed, this introduces a source of significant variations into a process. This higher frequency of feed changes demands more frequent action by the optimizer. However, imposing a more rapid optimizing action on industrial processes with slow dynamics pushes the time scales of the optimizer and regulator closer, making an optimizer which ignores the actions of the regulator inadequate. The proposed approach to account for the interaction between the optimizer and the regulator in such circumstances uses an optimization formulation known as the bilevel program. The PI proposes to develop an efficient algorithm for solving this optimization problem.