(558d) Process Cost Modeling and Production Planning for Petrochemical Industries under Uncertainties

Production planning provides industries various opportunities to improve economic profits, promote enterprise decision management, reduce process upset, and/or mitigate environmental burden. It becomes increasingly important for petrochemical industries under today's economic globalization. Although notable research progress has been made in recent years, however, two major difficulties in the petrochemical production planning area are still presenting big challenges for both academic and industrial people. One is the systematic uncertainty implicated in both internal and external production, which includes cost variation of water, electricity, steam, fuel, catalyst, raw material, product, and etc. The other is the lack of precise models for evaluating process economic performance, which are mainly due to the considerable complexity of petrochemical manufacturing process.

Facing these two difficulties, this paper introduced data mining techniques into modeling and planning procedures. Based on the available plant data, this paper first developed a data preprocess method to keep the data integrality, reduce data noise and decrease the number of input variables. Then a modeling methodology correlating variable costs and processing amount for various processing units is proposed. The integrated variable cost curve with respect to processing amount is verified to be consistent with the Economy of Scale Theory. Assisted by variable cost models, this paper further developed a graph-assisted production-planning modeling system for a general petrochemical plant. Since systematic uncertainties may present tangible characteristics in historical and new plant data set, data mining techniques are embedded into the system to identify and revise the possible dynamic relations among various uncertain variables. This will help the production-planning model more applicable in real industries.

To solve the nonlinear problem induced by the proposed production-planning model, an improved tempering simulated annealing (ITSA) algorithm together with tempering-rule based heuristic restarting-point techniques were introduced to advance the computational efficiency. The efficacy of the proposed modeling and planning methodologies and ITSA algorithm are demonstrated by successfully tackling of a real industrial problem, where seven main production units with 839 variables and 836 constraints are employed. The planning results show more reliable and precise economic benefits and also facilitate the decision-making process.