Multiple and nonuniform time grids in discrete-time MIP models for chemical production scheduling

The modeling of time plays a key role in the formulation of mixed-integer programming (MIP) models for scheduling, production planning, and operational supply chain planning problems. It affects the size of the model, the computational requirements, and the quality of the solution. While the development of smaller continuous-time scheduling models, based on multiple time grids, has received considerable attention, no truly different modeling methods are available for discrete-time models. In this paper, we challenge the long-standing belief that employing a discrete modeling of time requires a common uniform grid. First, we show that multiple grids can actually be employed in discrete-time models. Second, we show that not only unit-specific but also task-specific and material-specific grids can be generated. Third, we present methods to systematically formulate discrete-time multi-grid models that allow different tasks, units, or materials to have their own time grid. We present two different algorithms to find the grid. The first algorithm determines the largest grid spacing that will not eliminate the optimal solution. The second algorithm allows the user to adjust the level of approximation; more approximate grids may have worse solutions, but many fewer binary variables. Importantly, we show that the proposed models have exactly the same types of constraints as models relying on a single uniform grid, which means that the proposed models are tight and that known solution methods can be employed. The proposed methods lead to substantial reductions in the size of the formulations and thus the computational requirements. In addition, they can yield better solutions than formulations that use approximations. We show how to select the different time grids, state the formulation, and present computational results.     

An MILP continuous-time approach to short-term scheduling of resource-constrained multistage flowshop batch facilities

This work presents a new MILP mathematical formulation for the resource-constrained short-term scheduling of flowshop batch facilities with a known topology and limited supplies of discrete resources. The processing structure is composed of multiple stages arranged in series and several units working in parallel at each one. All production orders consist of a single batch and follow the same processing sequence throughout the plant. The proposed MILP approach is based on a continuous time domain representation that relies on the notion of order predecessor and accounts for sequence-dependent setup times. Assignment and sequencing decisions are independently handled through separate sets of binary variables. A proper formulation of the sequencing constraints provides a substantial saving in sequencing variables and constraints. By postulating a pair of conditions for the simultaneous execution of processing tasks, rather simple resource constraints requiring a few extra binary variables are derived. The proposed MILP scheduling approach shows a remarkable computational efficiency when applied to real-world problems.     

A unified graphical formulation for synthesis of basic and subcolumn distillation sequences

In contrast to the different approaches currently adopted for generating basic and side-split subcolumn distillation sequence for separating zeotropic multicomponent feed mixture, we present a unified graphical method applicable towards both basic and side-split subcolumn distillation sequence. For a given number of components in the feed mixture, we enforce constraints on a base graph to eliminate violations of conservation principles and to preclude distillation sequences that demand higher heat duty in all appraised practical scenarios. A compact set of algebraic constraints is transfixed using the graph counterpart for generating basic-only distillation configurations. These algebraic constraints utilize binary variables to quantify existence of submixture streams and this considerably reduces the number of variables in generating distillation sequences. We also suggest extension of the formulation to enable the exploration of thermally coupled configurations.     

A simpler better slot-based continuous-time formulation for short-term scheduling in multipurpose batch plants

Short-term scheduling of multipurpose batch plants is a challenging problem for which several formulations exist in the literature. In this paper, we present a new, simpler, more efficient, and potentially tighter, mixed integer linear programming (MILP) formulation using a continuous-time representation with synchronous slots and a novel idea of several balances (time, mass, resource, etc.). The model uses no big-M constraints, and is equally effective for both maximizing profit and minimizing makespan. Using extensive, rigorous numerical evaluations on a variety of test problems, we show that in contrast to the best model in the literature, our model does not decouple tasks and units, but still has fewer binary variables, constraints, and nonzeros, and is faster.     

MILP scheduling model for multipurpose batch processes considering various intermediate storage policies

MILP (Mixed Integer Linear Programming) scheduling models for non-sequential multipurpose batch processes are presented. Operation sequences of products have to be made in each unit differently by considering production route of each product under a given intermediate storage policy to reduce idle time of units and to raise the efficiency of the process. We represent the starting and finishing time of a task in each unit with two coordinates for a given storage policy. One is based on products, and the other is based on operation sequences. Then, using binary variables and logical constraints, we match the variables used in the two coordinates into one. We suggest MILP models considering sequence dependent setup times to guarantee the optimality of the solutions. Two examples are presented to show the effectiveness of the suggested models.     

An MILP-MINLP decomposition method for the global optimization of a source based model of the multiperiod blending problem

The multiperiod blending problem involves binary variables and bilinear terms, yielding a nonconvex MINLP. In this work we present two major contributions for the global solution of the problem. The first one is an alternative formulation of the problem. This formulation makes use of redundant constraints that improve the MILP relaxation of the MINLP. The second contribution is an algorithm that decomposes the MINLP model into two levels. The first level, or master problem, is an MILP relaxation of the original MINLP. The second level, or subproblem, is a smaller MINLP in which some of the binary variables of the original problem are fixed. The results show that the new formulation can be solved faster than alternative models, and that the decomposition method can solve the problems faster than state of the art general purpose solvers.     

Quantifying the uncertainty introduced by discretization and time‐averaging in two‐fluid model predictions

The two‐fluid model (TFM) has become a tool for the design and troubleshooting of industrial fluidized bed reactors. To use TFM for scale up with confidence, the uncertainty in its predictions must be quantified. Here, we study two sources of uncertainty: discretization and time‐averaging. First, we show that successive grid refinement may not yield grid‐independent transient quantities, including cross‐section–averaged quantities. Successive grid refinement would yield grid‐independent time‐averaged quantities on sufficiently fine grids. Then a Richardson extrapolation can be used to estimate the discretization error, and the grid convergence index gives an estimate of the uncertainty. Richardson extrapolation may not work for industrial‐scale simulations that use coarse grids. We present an alternative method for coarse grids and assess its ability to estimate the discretization error. Second, we assess two methods (autocorrelation and binning) and find that the autocorrelation method is more reliable for estimating the uncertainty introduced by time‐averaging TFM data. © 2017 American Institute of Chemical Engineers AIChE J, 63: 5343–5360, 2017     

Integrated process design,scheduling, and model predictive control of batch processes with closed-loop implementation

Simultaneous evaluation of multiple time scale decisions has been regarded as a promising avenue to increase the process efficiency and profitability through leveraging their synergistic interactions. Feasibility of such an integral approach is essential to establish a guarantee for operability of the derived decisions. In this study, we present a modeling methodology to integrate process design, scheduling, and advanced control decisions with a single mixed-integer dynamic optimization (MIDO) formulation while providing certificates of operability for the closed-loop implementation. We use multi-parametric programming to derive explicit expressions for the model predictive control strategy, which is embedded into the MIDO using the base-2 numeral system that enhances the computational tractability of the integrated problem by exponentially reducing the required number of binary variables. Moreover, we apply the State Equipment Network representation within the MIDO to systematically evaluate the scheduling decisions. The proposed framework is illustrated with two batch processes with different complexities.     

Piecewise linear relaxation of bilinear programs using bivariate partitioning

Several operational and synthesis problems of practical interest involve bilinear terms. Commercial global solvers such as BARON appear ineffective at solving some of these problems. Although recent literature has shown the potential of piecewise linear relaxation via ab initio partitioning of variables for such problems, several issues such as how many and which variables to partition, which partitioning scheme(s) and relaxation model(s) to use, placement of grid points, etc., need detailed investigation. To this end, we present a detailed numerical comparison of univariate and bivariate partitioning schemes. We compare several models for the two schemes based on different formulations such as incremental cost (IC), convex combination (CC), and special ordered sets (SOS). Our evaluation using four process synthesis problems shows a formulation using SOS1 variables to perform the best for both partitioning schemes. It also points to the potential usefulness of a 2‐segment bivariate partitioning scheme for the global optimization of bilinear programs. We also prove some simple results on the number and selection of partitioned variables and the advantage of uniform placement of grid points (identical segment lengths for partitioning). © 2009 American Institute of Chemical Engineers AIChE J, 2010     

Global optimization of multiphase flow networks using spline surrogate models

A general modelling framework for optimization of multiphase flow networks with discrete decision variables is presented. The framework is expressed with the graph and special attention is given to the convexity properties of the mathematical programming formulation that follows. Nonlinear pressure and temperature relations are modelled using multivariate splines, resulting in a mixed-integer nonlinear programming (MINLP) formulation with spline constraints. A global solution method is devised by combining the framework with a spline-compatible MINLP solver, recently presented in the literature. The solver is able to globally solve the nonconvex optimization problems. The new solution method is benchmarked with several local optimization methods on a set of three realistic subsea production optimization cases provided by the oil company BP.     

Hybrid time slots sequencing model for a class of scheduling problems

This article presents a new model for the short‐term scheduling of multistage batch plants with a single unit per stage, mixed storage policies, and multiple shared resources for moving orders between stages. Automated wet‐etching stations for wafer fabrication in semiconductor plants provide the industrial context. The uncommon feature of the continuous‐time model is that it relies on time grids, as well as on global precedence sequencing variables, to find the optimal solution to the problem. Through the solution of a few test cases taken from the literature, we show that new model performs significantly better than a pure sequencing formulation and better than a closely related hybrid model with slightly different sequencing variables. We also propose a new efficient heuristic procedure for extending the range of problems that can effectively be solved, which essentially solves relaxed and constrained versions of the full‐space model. © 2011 American Institute of Chemical Engineers AIChE J, 2012     

A generalized cutting-set approach for nonlinear robust optimization in process systems engineering

We propose a novel computational framework for the robust optimization of highly nonlinear, non-convex models that possess uncertainty in their parameter data. The proposed method is a generalization of the robust cutting-set algorithm that can handle models containing irremovable equality constraints, as is often the case with models in the process systems engineering domain. Additionally, we accommodate general forms of decision rules to facilitate recourse in second-stage (control) variables. In particular, we compare and contrast the use of various types of decision rules, including quadratic ones, which we show in certain examples to be able to decrease the overall price of robustness. Our proposed approach is demonstrated on three process flow sheet models, including a relatively complex model for amine-based CO₂ capture. We thus verify that the generalization of the robust cutting-set algorithm allows for the facile identification of robust feasible designs for process systems of practical relevance.     

A novel MILP formulation for short-term scheduling of multi-stage multi-product batch plants with sequence-dependent constraints

This paper presents a continuous-time mixed-integer linear programming (MILP) model for short-term scheduling of multi-stage multi-product batch plants. The model determines the optimal sequencing and the allocation of customer orders to non-identical processing units by minimizing the earliness and tardiness of order completion. This is a highly combinatorial problem, especially when sequence-dependent relations are considered such as the setup time between consecutive orders. A common approach to this scheduling problem relies on the application of tetra-index binary variables, i.e. (order, order, stage, unit) to represent all the combinations of order sequences and assignments to units in the various stages. This generates a huge number of binary variables and, as a consequence, much time is required for solutions. This paper proposes a novel formulation that replaces the tetra-index binary variables by one set of tri-index binary variables (order, order, stage) without losing the model's generality. By the elimination of the unit index, the new formulation requires considerably fewer binary variables, thus significantly shortening the solution time.     

Scheduling multistage, multiproduct batch plants with nonidentical parallel units and unlimited intermediate storage

Scheduling production optimally in multistage multiproduct plants with nonidentical parallel units is a very difficult but routine problem that has received limited attention. In this paper, we construct, analyze, and rigorously compare a variety of novel mixed-integer linear programming formulations using unit-slots, stage-slots, process-slots, a variety of slot arrangements and sequence-modeling techniques, 4-index and 3-index binary variables, etc. While two of our 4-index models are an order of magnitude faster than existing models on 22 test problems of varying sizes, we find that no single model performs consistently the best for all problems. Our work suggests that the best strategy for solving difficult scheduling problems may be to use a set of competitive models in parallel and terminate them all, when one of them achieves the desired solution. We also develop several heuristic models based on our formulations and find that even a heuristic based on an inferior model can surpass others based on superior models. Thus, it may not always be wise to just aim for a single best model for a given scheduling problem, but a host of novel and competitive models, as we have done in this paper.     

混合免疫算法及在相平衡计算中的应用

带或不带化学反应的相平衡计算为化学、化工领域的重要课题;可将其转换为带有约束的Gibbs自由能最小化问题。常用的序贯二次规划（SQP）收敛速度快;但依赖初始值;易陷入局部极小。人工免疫算法（AIS）具有全局寻优功能;但局部搜优性能差;收敛速度很慢;甚难找到痕量解。为此;在AIS算法中引入SQP操作;汲取两者的优点;构建混合免疫算法（HAIS）。还将相平衡的物质的量变量转换为摩尔分数;并采用适当策略处理约束;以基本可行解为基础;快速生成满足约束的抗体;以提高HAIS的操作速率。多个相平衡实例应用表明HAIS性能良好;优于其他方法（SQP;AIS）。     

A novel precedence-based and heuristic approach for short-term scheduling of multipurpose batch plants

Many continuous-time formulations have been proposed during the last decades for short-term scheduling of multipurpose batch plants. Although these models establish advantages over discrete-time representations, they are still inefficient in solving moderate-size problems, such as maximization of profit in long horizon, and minimization of makespan. Unlike existing literature, this paper presents a new precedence-based mixed integer linear programming (MILP) formulation for short-term scheduling of multipurpose batch plants. In the new model, multipurpose batch plants are described with a modified state-task network (STN) approach, and binary variables express the assignments and sequences of batch processing and storing. To eliminate the drawback of precedence-based formulations which commonly include large numbers of batches, an iterative procedure is developed to determine the appropriate number of batch that leads to global optimal solution. Moreover, four heuristic rules are proposed to selectively prefix some binary variables to 0 or 1, thereby reducing the overall number of binary variables significantly. To evaluate model performance, our model and the best models reported in the literature (S&K model and I&F model) are utilized to solve several benchmark examples. The result comparison shows that our model is more effective to find better solution for complex problems when using heuristic rules. Note that our approach not only can handle unlimited intermediate storage efficiently as well as the I&F model, but also can solve scheduling problems in limited intermediate storage more quickly than the S&K model.     

基于GPU加速求解MINLP问题的SQP并行算法

针对确定性算法求解大型复杂混合整数非线性规划的时间不可接受问题,通过对序贯二次规划算法(SQP)和图形处理器(GPU)的架构特点分析,提出了基于GPU加速策略的并行化SQP算法。算法的主要思想是通过枚举法确定二元变量的取值,在保证取值完整的基础上,使用CPU+GPU的并行策略,同时运用大量线程进行非线性规划子问题的求解。算例的数值实验结果表明:本文所提出的算法较之传统串行计算具有较好的加速效果,特别适合求解二元变量较多,约束条件相对少的MINLP问题。     

New approach for scheduling crude oil operations

Scheduling of crude oil operations is crucial to petroleum refining, which includes determining the times and sequences of crude oil unloading, blending, and CDU feeding. In the last decades, many approaches have been proposed for solving this problem, but they either suffered from composition discrepancy [Lee et al. 1996. Mixed-integer linear programming model for refinery short-term scheduling of crude oil unloading with inventory management. Industrial and Engineering Chemistry Research 35, 1630-1641; Jia et al., 2003. Refinery short-term scheduling using continuous time formulation: crude-oil operations. Industrial and Engineering Chemistry Research 42, 3085-3097; Jia and Ierapetritou, 2004. Efficient short-term scheduling of refinery operations based on a continuous time formulation. Computer and Chemical Engineering 28, 1001-1019] or led to infeasible solutions for some cases [Reddy et al., 2004a. Novel solution approach for optimizing crude oil operations. A.I.Ch.E. Journal 50(6), 1177-1197; 2004b. A new continuous-time formulation for scheduling crude oil operations. Chemical Engineering Science 59, 1325-1341]. In this paper, coastal and marine-access refineries with simplified workflow are considered. Unlike existing approaches, the new approach can avoid composition discrepancy without using iterative algorithm and find better solution effectively. In this approach, a new mixed integer non-linear programming (MINLP) formulation is set up for crude oil scheduling firstly, and then some heuristic rules collected from expert experience are proposed to linearize bilinear terms and prefix some binary variables in the MINLP model. Thus, crude oil scheduling can be expressed as a complete mixed integer linear programming (MILP) model with fewer binary variables. To illustrate the advantage of the new approach, four typical examples are solved with three models. The new model is compared with the most effective models (RKS(a) and RKS(b) models) presented by Reddy et al. [2004a. Novel solution approach for optimizing crude oil operations. A.I.Ch.E. Journal 50(6), 1177-1197; 2004b. A new continuous-time formulation for scheduling crude oil operations. Chemical Engineering Science 59, 1325-1341], which proves that the new approach is valid and feasible in most small-size and medium-size problems.     

Deterministic optimization of the thermal Unit Commitment problem: A Branch and Cut search

This paper proposes a novel deterministic optimization approach for the Unit Commitment (UC) problem, involving thermal generating units. A mathematical programming model is first presented, which includes all the basic constraints and a set of binary variables for the on/off status of each generator at each time period, leading to a convex mixed-integer quadratic programming (MIQP) formulation. Then, an effective solution methodology based on valid integer cutting planes is proposed, and implemented through a Branch and Cut search for finding the global optimal solution. The application of the proposed approach is illustrated with several examples of different dimensions. Comparisons with other mathematical formulations are also presented.     

Constrained adaptive sampling for domain reduction in surrogate model generation: Applications to hydrogen production

We propose a new approach for sampling domain reduction for efficient surrogate model generation. Currently, the standard procedure is to use box constraints for the independent variables when sampling the exact simulator. However, by including additional inequality constraints to account for interdependencies between these variables, we can drastically reduce the sampling domain and ensure consistency of unit operations. Moreover, we present a methodology for constructing surrogate models based on penalized regression and error-maximization sampling. All these algorithms have been implemented as a free and open-source software package. Through a case study on the water–gas shift reaction for hydrogen production, we show that sampling domain reduction reduces the required number of sampling points significantly and improves the accuracy of the surrogate model.