<p>Multi-intent Spoken Language Understanding (SLU) constitutes a fundamental component in dialogue systems. In recent years, prompt learning paradigms based on Large Language Model (LLM) have emerged as the predominant approach for SLU models. However, two critical issues persist in existing models. First, the label decision-making process lacks transparency, where semantic correlations between intents and slots are often implicitly modeled as black-box processes. Second, hierarchical semantic relationships in nested slots prove challenging to effectively capture through static graph structures. To address these challenges, we propose a Coupled Hierarchical Chain-of-Thought Prompt Learning Model for Multi-intent Spoken Language Understanding(CHCoT-MSLU). Specifically, we design a Coupled Hierarchical Chain-of-Thought (CHCoT) prompt template to guide the LLM Mixtral in generating explanatory rationales for each intent and slot label. These rationales serve as supervisory signals to train the model in replicating human-like reasoning processes. Furthermore, we introduce a Hierarchical Slot Label Dynamic Feedback Attention mechanism(HSL-DFA). It enables adaptive parameter adjustment at slot label hierarchies to enhance semantic comprehension, eliminating the need for graph construction. Experimental results on benchmark datasets MixATIS and MixSNIPS demonstrate that the CHCoT-MSLU model achieves average improvements of 1.25% in intent accuracy, 1.38% in slot F1 and 0.62% in overall accuracy, respectively, compared to the strong baselines ENSI-Mistral and ECLM models.</p>