间歇过程生产调度与用水网络同步优化

Simultaneous Optimization of Production Scheduling and Water-Using Network for Batch Processes

与分步优化相比,间歇过程生产调度和用水网络同步优化解空间更完备,能更好地权衡生产与用水费用。但两者时空间维度高度耦合导致问题规模组合爆炸式增长,选择恰当的时间表达方式成为控制模型复杂度的关键。首先,为了克服以往单一采用全局连续时间模型复杂度大和特定单元事件点模型解空间丢失的缺陷,提出了混合时间表达方式,即分别用特定单元事件点和非均匀离散时间描述调度模型与用水网络模型,再用辅助约束关联两个时间维度。然后,基于状态任务网络和用水网络超级结构构建混合整数非线性模型(MINLP)并采用分步法(MILP-NLP)和随机扰动相结合的初值策略进行求解。两个算例结果表明:与分步法相比,同步优化大大降低了用水量,且所提方法避免了解空间丢失,有效缩减了模型复杂度。

Compared to hierarchical methods, simultaneous optimization of batch production scheduling and water-using network can ensure the integrity of solution space and achieve better trade-offs between production and water-using costs. However, the high time-space coupling between two modules inevitably leads to a combinational explosion of alternative configurations, thereby the choice of proper time representation is critical to control model complexity. In order to overcome drawbacks of high complexity of the global continuous time model and solution space reduction of the unit-specific event point model, a hybrid time model was proposed which consisted of unit-specific event point for schedule model, non-uniform discrete time for water-using network model and auxiliary time correlation constraints. A mixed-integer nonlinear programming （MINLP） model was then formulated based on state task network （STN） and water-using network superstructure Moreover, a solution strategy was also developed, which integrated hierarchical methods （MILP-NLP） and stochastic disturbance to provide efficient initial values. The results of two illustrative examples demonstrate that synchronous optimization significantly decreases water consumption compared to hierarchical methods, and the proposed methodology can avoid loss of solution space as well as reduce computational complexity.