An Intelligent QCNN-Based Hybrid AZOA–RPOA Framework for Optimal Load Shedding in Renewable and EV Integrated Islanded Distribution Networks

The increasing penetration of renewable generation and electric-vehicle charging complicates emergency load-shedding decisions in islanded distribution networks, particularly when critical loads must be protected while voltage and economic constraints are maintained. This paper presents a quantum convolutional neural network assisted hybrid American Zebra Optimization Algorithm–Red Panda Optimization Algorithm framework for priority-aware partial load shedding. A four-qubit quantum convolutional neural network assigns continuous priority scores to curtailable buses using normalized load magnitude, power-factor-based stress, renewableenergy proximity and customer criticality. The resulting scores and Grades A–D are incorporated into the optimization through priority penalties and grade-dependent shedding limits. The hybrid optimizer employs adaptive selection between exploratory and refinement operators, together with greedy replacement, elite preservation and stagnation recovery. The framework is evaluated on a modified IEEE 33-bus islanded distribution system under conditions without and with electricvehicle charging. Thirty independent runs are performed using a common budget of 30,000 objective-function evaluations. The proposed method achieves mean composite objective values of 294.23 and 325.69 for the two operating cases, with corresponding interruption costs of $213.72 and $220.77. Mean minimum voltages remain at 0.9655 and 0.9550 p.u., while Grade-A curtailment is practically negligible. The proposed method also achieves a 100% feasible-run rate in both cases. Comparison with a compact multilayer perceptron and conventional metaheuristics shows that the proposed framework provides a favorable balance among economic curtailment, priority protection and voltage feasibility without claiming quantum computational advantage.