A Self-learning Hyper-Heuristic Algorithm for Energy-Efficient Distributed Flexible Job Shop Scheduling

The distributed flexible job shop scheduling problem (DFJSP) has been studied intensively over the past decades, and the energy-efficient distributed flexible job shop scheduling problem (EDFJSP) has emerged as a hot and challenging research topic. A self-learning hyper-heuristic algorithm (SLHH) is proposed in this paper to solve EDFJSP. First, an EDFJSP model is established, and a coding scheme is designed based on the characteristics of the problem. Second, a Q-learning-based high-level strategy and an improved ϵ - greedy policy are designed to guide six knowledge-based low-level heuristics (LLHs) to explore the solution space and move acceptance to ensure the diversity of solutions. Third, an energy-saving strategy is proposed which can effectively reduce the energy consumption of the solution. The performance of SLHH is tested against three state-of-the-art algorithms in 20 benchmark instances. The experiment results demonstrate the effectiveness of SLHH in addressing the multifaceted challenges of EDFJSP.