To achieve system load peak shaving and valley filling, as well as real-time balancing of power differences between load peaks and valleys, a multi-energy collaborative optimization method for energy supply is proposed, integrating new and conventional energy sources. Based on the different generation characteristics of wind power, photovoltaic power, hydropower, and thermal power, an optimization model is constructed for a new energy alliance. The energy block method for peak and valley loads is first applied to determine the electricity demand to be undertaken by the alliance. Through internal resource optimization allocation and multi-agent generation combination-based multi-scale cost balancing optimization, an operational decision-making method for multi-energy alliance optimization based on multi-scale dynamic cost Shapley values is proposed. This method quantifies the dynamic allocation laws of cost allocation and output proportion of each entity, and a comparison is made between the kernel method and the Shapley value method. Furthermore, a comprehensive evaluation model based on the grey relational degree-entropy weight TOPSIS method is used to select optimal power generation combinations, aiming to enhance system optimization. Case analysis demonstrates that the proposed multi-energy alliance operation mechanism effectively smooths fluctuations in new energy output, increases the internal new energy consumption ratio, and enhances economic efficiency and system stability. The research findings provide theoretical support for generation planning, cost-sharing strategies, and combination optimization within the alliance and offer practical guidance for the efficient and coordinated operation of multi-energy systems.