Optimization of Sequence-dependent Harvesting Time at Sugarcane Farms using Meta-heuristic Methods

Abstract

The framework of the present problem is based on predicting the recoverable sugar percentage, which has a fundamental role in sugarcane harvesting time. After forecasting this parameter, two objectives are pursued one tries to maximize the sugar production quantity according to the proper sugarcane age and variety, and the other tries to minimize the completion time of harvesting operations in a specified sequence of sugarcane farms. To estimate the recoverable sugar percentage parameter was used an Elman Neural Network (ENN). Then, the sequence-dependent harvesting time problem was formulated by a hybrid model called the Travelling Thief Problem (TTP). To solve this bi-objective problem has been used two meta-heuristic algorithms, called NSGA-II and SPEA2. Results indicate that the bi-objective optimization problem can be increased the sugar production quantity by 32.93% and be decreased the completion time of harvesting operations via finding optimal routes by 57.7% compared to the actual harvesting sequence. The statistical testing results show that the NSGA-II is superior to the SPEA2 in terms of achieving better convergence, generating more non-dominated solutions, improving the distribution of solutions, and shortening the running time.