Optimal scheduling of batch processes for heat integration

While batch plants are generally of smaller scale than continuous plants, they nonetheless can incur significant heating and cooling needs. Thus the opportunity exists of obtaining energy savings through heat integration methods involving matching of hot and cold streams. However, because of the fact that streams or batches of material are only available over specific time intervals, the potential matches are constrained by timing considerations, that is by the schedule under which the batch process is operated. The resulting heat exchanger synthesis problem incorporates aspects of batch task rescheduling, in order to jointly consider the tradeoff between energy recovery and production rate reductions. In this work, this idea is applied to a single product batch production line containing both in-phase and out-of-phase units and operating with a cyclic schedule under the no intermediate storage (NIS) policy. Only a single match of a given batch at a given stage is allowed with another batch. Moreover, it is assumed that there are no heat losses and gains when a batch is held in a processing unit under the NIS policy. A maximum allowable holding time formulation is proposed which jointly determines the operating schedule and stream matches. The formulation is solved using conventional MILP solvers such as MPSX. Since the size of the model is dependent on the number of repeated batches which are considered, predictive relations are derived to allow estimation of the minimum number of repeated batches which is sufficient. Numerical examples are presented to illustrate the applicability of the formulation. The examples show that with modest schedule revision, it is possible to achieve energy recoveries equivalent to the maximum recovery obtainable if time effects are disregarded. The MILP formulation is then extended for the heat integration of two independent batch processes. An algorithm based on Bender\u27s decomposition method is developed to solve the resulting large scale integer programming problem. The solution using the decomposition method is shown to be obtained much faster than with conventional MILP solvers. (Abstract shortened with permission of author.