Managing Supply Uncertainty in the Operational Production Planning in a Whitefish Value Chain - A Stochastic Programming Approach
MetadataShow full item record
The fishing industry is a significant contributor to Norwegian exports. However, the whitefish industry is experiencing a declining profitability. This research contributes in two ongoing projects, Qualifish and iProcess, directed by SINTEF. The combined objective of the projects is to develop and make use of new technology, to improve the performance of the whitefish industry. Lerøy Norway Seafoods are involved as industrial partner in both projects. The problem addressed in this study, is the uncertainty of daily available raw material, and the impact this uncertainty has on the operational production planning at Lerøy Norway Seafoods Båtsfjord. Today, the operational production planning is conducted by key personnel, based on experience. In this study, a stochastic programming approach is applied to optimize the allocation of available raw material of Cod to end-products. The objective is to maximize the profit. This study investigates how stochastic programming can support decisions in operational production planning with uncertainties in supply, and how the number of discretizations, i.e. scenarios, affects the performance of a stochastic program. The stochastic solution is compared to the deterministic solution, where all parameters are assumed to be known and constant. A two-stage stochastic linear program is developed for a three-day planning period. The available quantity of raw material is certain for the first two days, but considered uncertain for the third day. The first two days correspond to the first stage, and the third day corresponds to the second stage. The supplied quantity of raw material is continuously distributed, but are discretized into scenarios by applying moment matching as the scenario-generation method. The scenarios are assumed to occur with equal probability. The stochastic linear program optimizes the allocation for the first two days, given all possible scenarios of supplied quantity for the third day. The results show that the stochastic solution outperforms the deterministic solution for all discretizations. The stochastic linear program hedges for undesired scenarios on the third day, by saving raw material from the first two days. By taking uncertainty into account, the effect of hedging results in increased expected profit by avoiding unnecessary costs. The stochastic program is solved for a number of three, five, seven, nine, and eleven discretizations. For a number of three discretizations, the stochastic solution provides an expected profit 71.3% higher than the deterministic solution. For an increment in the number of discretizations, the expected profit increases with decreasing improvements. Eleven is the maximum number of discretizations considered in this study, and hence, the most accurate approximation to the continuous distribution. The stochastic solution for a number of eleven discretizations provides an expected profit 119.2% higher than the deterministic solution.