Abstract: Currently, the allocation and management of pipeline welding tasks in thermal power construction rely on the personal experience and intuition of management personnel. This article constructs a dynamic portrait by quantifying the skill characteristics and quality trends of welders, and combines the Apriori algorithm to mine the welder defect association rules. A knowledge graph is constructed to achieve intelligent matching and risk warning of welding tasks and welding skills. A pipeline welding task allocation system for thermal power construction is constructed, which includes five modules: data collection, welder dynamic portrait construction, knowledge graph construction and reasoning, task adaptability evaluation, and task allocation suggestions generation. The problems of subjectivity, lack of risk foresight, and insufficient data value mining in existing welding task allocation are solved, and the quality and efficiency of pipeline welding operations in thermal power construction are improved.