A new differential evolution algorithm for dynamic scheduling problems with variant job weights

In the real world, many practical applications are non-stationary optimization problems. This requires the optimization techniques not only find the global optimal solution but also track the trajectory of the changing global best solution. Dynamic scheduling problems pose great challenges to traditional differential evolutionary algorithms due to the diversity loss and low optimization efficiency. This paper introduces a new multi-population strategy for differential evolution (DE) algorithm to address the dynamic scheduling problems with variant job weighs. DE has always been applied for optimization problems in continuous solution space, while this new algorithm uses random key coding scheme to convey the continuous position vector to the sequential vector for each individual, and introduces a self-organized multi-population strategy to partition the population into parent population and child populations. The parent population is assigned to continuously search for new peaks, and child subpopulations are assigned for further exploitation in some promising areas. In addition, population sizes are adjusted according to their qualities for accelerating the optimization speed. It has been applied to the dynamic scheduling problems with variant job weights, the satisfactory results have been achieved.