通用大模型难适配新型电力系统，定制化AI成发展方向

通用大模型在适应新型电力系统的物理约束方面存在明显短板，专用的混合架构大模型成为必要选择1。我国能源结构正在加速调整，国家能源局数据显示，截至7月底，光伏装机容量首次超过煤电，占全国装机的31.5%1。光伏发电已深度融入社会用电，全国每8度电中就有超过1度来自光伏，占全社会用电量的13%1。

为适应这一转变，国家明确了非化石能源发展目标与人工智能应用规划。到2030年，非化石能源发电量占比要达到50%1。与此同时，国家密集出台政策推进人工智能在能源领域的深度应用，计划到2027年推动五个以上专业大模型在电网、发电、煤炭、油气等行业实现深度应用1。从更长远看，到2030年，能源领域人工智能专用技术与应用总体要达到世界领先水平1。

China's energy landscape has undergone a dramatic transformation as renewable energy continues to outpace traditional sources. As of July 2026, photovoltaic installations have surpassed coal-fired capacity for the first time, accounting for 31.5% of the nation's total generating capacity, according to data from the National Energy Administration 1. This shift reflects the accelerating transition toward clean energy, with solar generation now contributing over 13% of total electricity consumption nationwide—meaning more than one in every eight kilowatt-hours consumed comes from photovoltaic sources 1.

To support this transition, the government has set ambitious targets for the energy sector's decarbonization. Non-fossil fuel energy is projected to represent 50% of total power generation by 2030 1. However, integrating such high proportions of renewable energy into the grid presents significant technical challenges that general-purpose artificial intelligence models struggle to address. These broadly-designed systems cannot adequately accommodate the complex physical constraints inherent in the emerging power system, necessitating specialized hybrid-architecture AI models tailored specifically for energy applications 1.

Recognizing this gap, China has committed to advancing AI capabilities within the energy sector. The government aims to deploy at least five specialized large models across power grids, generation facilities, coal mining, and oil and gas operations by 2027, achieving deep integration into these industries 1. By 2030, China targets achieving world-leading levels in AI technologies and applications specifically designed for the energy domain 1.