A parallel Probability-preserving Random Walk Algorithm Considering Bi-level Optimization of Microgrids

To improve the computational efficiency and global optimization performance of the integrated capacity planning and operation dispatch for microgrids, a parallel probability-preserving random walk algorithm is proposed. Based on a bi-level optimization model for microgrid equipment capacity and operation, the upper level takes the capacity of each device as optimization variables, and the random walk algorithm with compulsive evolution (RWCE) adopts a parallel strategy to enable simultaneous computation of multiple populations. The lower level receives the equipment capacity from the upper level, takes the real-time output of each device as optimization variables, and introduces a probability-preserving mechanism into the RWCE algorithm to enhance local exploration by retaining the original solution with a small probability, then feeds the operational cost back to the upper level. Case study results show that the proposed algorithm improves optimization efficiency by approximately 35% and reduces the total annual cost by 1.7%, effectively addressing the issues of long computation time and insufficient local search accuracy. Moreover, it is applicable to the efficient co-optimization of capacity and operation for grid-connected microgrids.