<p>With the growing demand for intelligent electronic devices, the scale of analog integrated circuits is rapidly expanding. However, the automated design of such circuits faces major challenges, including a large number of design parameters and complex constraints. Existing algorithms often struggle to efficiently identify optimal parameter configurations. This paper proposes a Knowledge-Injected Constrained Multi-Objective Evolutionary Algorithm (KI-CMOEA). The algorithm integrates analog circuit domain knowledge into the evolutionary strategy, accelerating the generation of high-quality candidate solutions. A stagnation detection mechanism is introduced to identify suitable moments for knowledge injection. Based on the severity of stagnation, the degree of knowledge integration is dynamically adjusted. In addition, an optimization principle assisted by a Large Language Model (LLM) is proposed, enabling effective knowledge injection and performance enhancement through prompt engineering. Finally, the proposed algorithm is validated on several analog circuit design tasks, and experimental results demonstrate that it can efficiently identify optimal parameter combinations under complex constraints, significantly improving the overall efficiency of analog circuit design.</p>