Event Causality Extraction (ECE) aims to extract structured cause-effect event pairs from text. Existing classification methods overly rely on argument information for Cause-Effect Type Identification (CET), making them error-prone in cases of sparse arguments or complex scenarios (e.g., multiple causes and effects, one cause with multiple effects, or multiple causes with one effect). Generative methods face limitations in extracting multi-causal events and precisely associating arguments with event types, particularly in complex scenarios. To address these issues, this paper proposes Label-Aware Multi-Prompt Generation Network (LA-MPGN) for two key ECE subtasks: 1) CET and 2) Event Argument Extraction (EAE). In CET, a Masked Language Model (MLM) predicts \(\left[{\text{MASK}}\right]\) in continuous prompts, while a Graph Convolutional Network (GCN) captures label correlations to identify cause-effect types, avoiding dependence on arguments. For EAE, a multi-prompt learning strategy uses sub-prompts to locate arguments and main prompts to assign them to corresponding slots. By integrating causal pair information into multi-prompt templates, this method effectively reduces argument classification errors. Experiments on the benchmark dataset ECE-CCKS demonstrate that LA-MPGN outperforms previous state-of-the-art (SOTA) model and performs well in ECE for complex multi-causal scenarios.