Adaptive backtracking search optimization algorithm with a dual-learning strategy for dynamic economic dispatch with valve-point effects

Dynamic economic dispatch with valve-point effect (DED_vpe) is a dynamic nonlinear high-dimensional optimization problem with non-smooth and non-convex characteristics. Meta-heuristic methods have become the mainstream for solving the DED_vpe problem. However, most of these methods only focus on minimizing the generation costs and ignore the algorithmic robustness. In this paper, an adaptive backtracking search optimization algorithm with a dual-learning strategy (DABSA) is proposed for solving the DED_vpe problem. In DABSA, a dual-learning strategy (DL) based on the current and historical optimal individuals is developed to update each individual. This updating strategy helps DABSA improve solution accuracy and overcome premature convergence. In addition, an adaptive parameter control mechanism (APC), which can automatically adjust parameter ‘mixrate’ value according to the current iteration number, is presented. To handle the system constraints, a ‘repair + penalty’ constraints-handling approach is employed to lead non-feasible solutions towards the feasible region quickly. The performance of DABSA is assessed by testing on four DED problems containing 5, 10 and 30 units. The experimental results show that DABSA is very competitive compared with reported representative methods in yielding low fuel costs along with high robustness. • A new method DABSA is proposed for solving DED problem with valve-point effects. • The core innovation of DABSA lies in the design of a dual learning strategy (DL). • BSA is the first attempt in solving DED problem with valve-point effects. • Results on 5, 10 and 30 units show that DABSA is competitive in terms of solution accuracy and robustness.