(288g) A Hybrid Model for Production Planning Under Demand Uncertainty in Refineries | AIChE

(288g) A Hybrid Model for Production Planning Under Demand Uncertainty in Refineries

Authors 

Li, C. - Presenter, Tsinghua University
He, X. - Presenter, Tsinghua University
Chen, B. - Presenter, Tsinghua University


The deterministic production planning model (DPPM) has been employed in planning activities by most of current plants. However, because of volatile raw material costs, fluctuating product demands, and other unpredictable market conditions, many parameters in a production planning model should be considered as uncertainties. Among them, the product demand is one of the most important ones. It has been demonstrated that failure to incorporate a stochastic description of the product demand could lead to either unsatisfied customer demands or unnecessary inventory holding costs. Thus, production planning under demand uncertainties is challenged as one of the major problems in supply chain management, which still needs further in-depth studies.

This paper introduces a hybrid model for production planning under uncertain demands in refineries, which incorporates the deterministic planning and stochastic planning models by weight factors. Normal and uniform distribution assumptions for uncertain demand are used to test the robust of the normal stochastic model and the proposed hybrid model. To solve these models, piecewise linear approximate functions are employed, which bears enough accuracy and fast solution speed. It shows that the optimal results obtained by only deterministic planning model are more aggressive, while the results by only stochastic model are more conservative. Using the hybrid model with appropriate weight factors, however, will result in better economic performance than both deterministic and stochastic models. Meanwhile, different probability distribution assumptions for uncertain demands have significant impacts on the optimal results from stochastic model, but are insensitive to the hybrid model. Therefore, the developed hybrid model provides a more applicable approach for better production planning under demand uncertainties.