Data-Driven Optimization for Energy-Constrained Dietary Supplement Scheduling: A Bounded Cut MP-DQN Approach

Document Type

Article

Publication Date

2-1-2024

Abstract

Energy rationing exerts a substantial influence on the landscape of manufacturing operations. When mandatory energy rationing occurs, manufacturers find themselves compelled to adapt their production strategies to align with the prescribed rationing timetable. In more challenging scenarios when energy availability is uncertain, manufacturers often find it very difficult to arrange their production based on historical data. This research studies an energy-constrained multi-period production scheduling problem for dietary supplement manufacturers. The primary objective is to streamline the scheduling of production across multiple periods, factoring in the stochastic availability of electricity, to minimize the cumulative weighted tardiness of customer orders. To address this, we formulate a rigorous mathematical programming framework, denoted as ICMP. It is meticulously designed to unveil the optimal production schedule through meticulous system logic analysis, contingent upon knowledge of the energy rationing schedule. For the stochastic case where access to the rationing schedule is not readily attainable, it is demonstrated that the problem can be transformed into a parameterized action Markov decision process. An innovative data-driven reinforcement learning strategy named BC-MP-DQN is proposed to fine-tune the production schedule. Extensive computational experiments are performed to evaluate the effectiveness of the proposed algorithm. Its performance is also benchmarked with the optimal solution and the popular Model Predictive Control (MPC) algorithm. This study imparts valuable managerial insights to manufacturing enterprises grappling with the challenges of energy rationing. Moreover, it stimulates further research into the problem of stochastic intermittent production systems with intricate setups and hybrid decision spaces.

Share

COinS