Integrated condition-based planning of production and utility systems under uncertainty

Abstract

A general rolling horizon optimization framework for the integrated condition-based operational and maintenance planning of production and utility systems in process industries is presented. In brief, the proposed optimization framework considers for the production and utility units: (i) improved unit performance degradation and recovery models that depend on both the cumulative time of operation and the unit operating levels deviation of units; (ii) modified operating capacities under online cleaning periods; (iii) different types of cleaning tasks (flexible time-window and online or offline condition-based); (iv) alternative options for offline cleaning tasks; (v) limited availability of cleaning resources; (vi) the initial state of the overall system at the beginning of each planning horizon; and (vii) terminal constraints for the rolling horizon problem. Total cost constitutes the objective function of the resulting problem and includes unit operating costs, cleaning costs, energy consumption costs and resource purchases costs. The case studies solved show that when compared to solutions obtained by sequential approaches the proposed integrated approach provides significantly better solutions in terms of total costs (reduction from 5%-32%), and especially in cost terms related to utility units operation, energy consumption, cleaning and startup/shutdown operations. Unnecessary cleanings and purchases of resources can be avoided by the proposed integrated approach.

A rolling horizon stochastic programming approach for the integrated planning of production and utility systems

“…It is assumed that some information about the behavior of the uncertainty parameters is known (i.e., number of scenarios with associated probability of occurrence, and given parameter values for each scenario). In particular, this study is a major extension of our previous work (Zulkafli and Kopanos, 2017) by: (i) providing a two-stage scenario-based stochastic programming version of a modification of the previously deterministic model, (ii) introducing an improved cumulative operating level deviation model for condition-based cleaning policies, (iii) defining improved terminal constraints for the maximum allowable unit performance degradation level (i.e., minimum performance level) at the end of the planning horizon, (iv) incorporating the resulting two-stage scenario-based stochastic programming model into a rolling horizon framework to readily deal with various types of uncertainties. The proposed approach follows a plant-wide condition-based approach for the cleaning actions that explicitly consider the condition of the units as a result of the optimized operational planning of the production and utility systems.…”

“…In this study, the main parameters that describe the initial state of the overall system are: (i) the inventory levels for utilities and products; (ii) the cumulative time of operation for each unit; (iii) the cumulative operating level deviation for each unit; (iv) the current operating status of each unit; (v) the startup and shutdown history of each unit; (vi) the online and offline cleaning history of each unit; (vii) the cleaning resources history of units; and (viii) the demands for products per scenario considered. A more detailed description and discussion on the reactive planning via a rolling horizon framework can be found in Zulkafli and Kopanos (2017).…”

“…The stochastic programming model presented is an enhanced modified version of the deterministic model of our previous work (Zulkafli and Kopanos, 2017). For this reason, constraints that remain unchanged from its deterministic version, proper references will be given to the constraints of our previous work to avoid unnecessary repetitions.…”

“…There are two major types of planning approaches to deal with uncertainties, namely reactive and proactive approaches. In general, reactive approaches involve the repetitive solution of the deterministic planning problem within a rolling horizon framework, and are especially suitable for highly dynamic environments with limited information about the behavior of the uncertainty (Zulkafli and Kopanos, 2017). In simple words, these approaches basically rely on a wait and react approach with respect to unexpected events.…”

“…A brief literature reviews on sequential approaches in process industries and a discussion on the need for integrated plant-wide planning approaches can be found in Zulkafli and Kopanos (2016). Importantly, Zulkafli and Kopanos (2017) showed that significant total cost reductions (from 5% to 32%) can be achieved if an integrated planning approach is used instead of the sequential alternative. Therefore, there is an important need for development of integrated planning approaches that also account efficiently for uncertainty to deal with the dynamic nature of the process industries.…”