A Memory-Assisted Self-Adaptive Bonobo Optimizer for Economic Dispatch with Valve-Point Effects Including Combined Heat and Power Units
Abstract
Combined Heat and Power Economic Dispatch (CHPED) is a nonlinear and nonconvex optimization problem involving the coordinated operation of power-only, heat-only, and combined heat-and-power units under electrical and thermal balance constraints, coupled CHP operating regions, valve-point loading effects, and transmission losses. This study proposes a Memory-Assisted Self-Adaptive Bonobo Optimizer (MSBO), which extends the original Bonobo Optimizer (BO) through old, worse, and bad memory populations, adaptive partner selection, and repulsion-based parameter learning. The retained evolutionary information is used to improve search diversity, mitigate stagnation, and regulate the exploration-exploitation balance according to search progress. MSBO is evaluated on 7-unit and 48-unit CHPED systems under both lossless and transmission-loss conditions. Repeated-run analysis shows that, for the 7-unit system, MSBO reduces the standard deviation of the final cost by 99.94% and 99.95% in the lossless and lossy cases, respectively, while the Mann–Whitney tests (p = 0.1485 and p = 0.4095) indicate that the principal benefit is improved repeatability rather than a significant shift in the central cost level. For the 48-unit system, MSBO achieves mean operating-cost savings of 931.09 $/h and 782.54 $/h, with highly significant Mann–Whitney results of p = 3.08 × 10−8 and p = 5.23 × 10−7, respectively. The estimated probabilities that an MSBO run outperforms a BO run are 91.67% and 96.50% for the two 48-unit scenarios. These results demonstrate that MSBO primarily strengthens solution reliability for the smaller CHPED system while improving both operating-cost quality and repeated-run performance for the larger-scale system.