Variable neighborhood search guided differential evolution for dynamic load dispatch with Non-Smooth cost function

This paper presents a new and novel approach for solving the dynamic load dispatch (DLD) problem of generating units. Operational constraints, such as valve point loading and ramp rate limits are taken into account in the formulation of DLD problem. The conventional fuel cost curve of the generating units is modeled as a highly non linear curve by inclusion of valve point effect as a rectified sinusoidal component. The paper investigates the hybridization of differential evolution algorithm by variable neighborhood search (VNS) for solving DLD problem considering non-smooth characteristics (NSDLD). The basic idea of VNS is to explore increasingly distant neighborhoods of the current solution. The hybrid method incorporates DE as the global optimizer and VNS to fine-tune the solution space discovered by DE run. Thus, the VNS guides DE for better performance in the complex solution space. To validate its efficiency and effectiveness, the VNS guided DE is applied to solve a non-smooth DLD problem of power systems with 5 and 10 units. It is shown that, the proposed technique for non-smooth DLD problem generates satisfactory and superior solutions when compared with previous approaches.