在“双碳”目标驱动下，新型电力系统用户侧因分布式光伏、电动汽车充电桩、储能设备等新型并网主体的大规模接入面临多维安全风险挑战。首先系统分析了新型并网主体大规模接入引发的设备安全、电网安全及社会安全等问题，并指出了当前技术与管理层面的核心瓶颈，如设备质量不均、跨模态数据融合困难、安全防护缺失等。针对多模态数据的异构性、语义差异及关联特性，提出了多模态大模型构建技术，涵盖数据增强生成、跨模态语义对齐及参数高效微调方法，以提升风险感知精度。此外，从网架优化、智能运维、韧性提升三方面提出一体化安全防控对策，为新型电力系统安全治理提供了理论支撑与实践路径，助力“双碳”目标下能源转型与安全协同发展。

Driven by the "dual carbon" goals, the user side of the new power system faces multidimensional security risks due to the large-scale integration of new grid-connected entities such as distributed photovoltaic systems, electric vehicle charging stations, and energy storage equipment.Firstly, the challenges to equipment safety, grid security, and societal safety posed by the massive integration of these new entities are systematically analyzed, core bottlenecks at the technical and managerial levels, such as uneven equipment quality, difficulties in cross-modal data fusion, and insufficient safety protections, are highlighted. To address the heterogeneity, semantic differences, and correlation characteristics of multimodal data, a multimodal large model construction technology is proposed, incorporating data augmentation generation, cross-modal semantic alignment, and parameter-efficient fine-tuning methods to enhance risk perception accuracy. Furthermore, integrated security prevention and control strategies are introduced from three perspectives: grid optimization, intelligent operation and maintenance, and resilience enhancement. These measures provide theoretical support and practical pathways for the security governance of the new power system, facilitating the coordinated development of energy transition and security under the "dual carbon" objectives.