• • The POA-GWO-CSO hybrid algorithm integrates GWO's alpha/beta/delta leadership hierarchy and CSO's horizontal-vertical crossover into the POA framework, directly addressing the local-optimum entrapment and parameter sensitivity that degrade standalone particle swarm, genetic, and basic POA solvers in source-load-storage dispatch; this matters industrially because dispatch decisions must remain robust across volatile wind-solar profiles without repeated parameter retuning.
• • Building envelope thermal inertia is formalized through a resistance-capacitance network (Rwall, Cwall, Rwin, Croom) that converts air-conditioning loads into virtual energy storage, enabling a peak-shaving/valley-filling strategy of pre-heating or pre-cooling before peak periods and power reduction during peaks; this reduces reliance on physical battery storage and lowers building cluster energy costs without violating indoor temperature comfort bands.
• • The source-side model couples wind and photovoltaic physical output prediction with explicit forecast-error quantification, while the storage-side model enforces battery full life-cycle constraints; this dual treatment prevents the conservative over-sizing that arises when storage requirements are assessed from only the generation or grid side, and it dynamically matches source-load temporal mismatch.
• • Validation on the actual Minning Town dataset in Ningxia confirms that the proposed multi-side coordinated strategy improves active distribution network economic operation and renewable consumption rate compared with alternative dispatch strategies, providing a replicable template for regions experiencing high wind and solar curtailment under the dual-carbon framework.