AI-powered multistage grid planning optimization method for medium-voltage distribution networks based on AHA algorithm

Simply connecting some lines based on fixed rules can only meet short-term, static load demands and power access conditions, resulting in a low rationality in distribution network planning. The "domain search" and "migratory behavior" properties are utilized to enhance global search capabilities. An adaptive step size and non-uniform mutation strategy are employed to search for medium-voltage distribution network planning solutions. Combining the global optimization results of the AHA algorithm with dynamic weight adjustment and a spiral foraging strategy, an optimal AI-powered multi-stage grid planning solution for medium-voltage distribution networks is generated through phased iteration and evaluation, balancing economy, reliability, and renewable energy absorption capacity. Experiments show that the proposed method produces an orderly and efficient grid structure. The average per-unit voltage value of 33 nodes reaches 1.002, and the average voltage deviation is as low as 0.09%, with all node voltage deviations falling within the acceptable range. Compared with the comparison method, the proposed method ultimately stabilizes at approximately 90,000 yuan, a significantly lower objective function value, and achieves faster convergence and stronger search capabilities. Experiments demonstrate that the proposed method leverages the powerful global search and adaptive adjustment capabilities of the AHA algorithm to improve planning rationality.