With the continuous increase of accumulated data of power supply companies, the processing and management of massive data is becoming more difficult. In order to ensure high-quality data, provide more thoughtful marketing services, and improve the execution efficiency, a power supply company marketing data mining system based on Apriori algorithm is designed. In terms of hardware design, through the deployment of the physical architecture of the marketing data mining system, the system hardware environment is built to achieve the synchronization of database information. In terms of software design, design the power marketing data warehouse, collect and sort out the original business data, extract the power marketing data, adopt the Apriori algorithm, process frequent item sets by mapping pruning, calculate confidence, and mine association rules, so as to establish a multidimensional data mining model, select data mining space, calculate the contribution components of each dimension, and realize the data mining function of the system. Through experimental demonstration and analysis, the difference between the predicted results and the actual values of the system in the application of power load forecasting is small. Under the conditions of minimum support and transaction data volume, the execution time of data mining is less than 2s and 10s respectively, which has high execution efficiency, indicating that the system in this paper is feasible.