Load balance just reasonably allocate request from users to each edge cloud storage servers. It can avoid overloading or congestion to ensure data transmission efficiency. Referring to idea of ant colony algorithm, the paper brought out Content Delivery Load Balance Algorithm Based on Probability Matching (CDLBA-PM) algorithm. The method calculated access probability of each edge cloud storage peer and selected edge cloud storage peer with probability matching algorithm comprehensively considering about history access costs and real-time access cost of edge cloud storage peers. It fully takes into account current network bandwidth, edge cloud storage peer performance as well as history access cost. Random probability matching was also introduced for cache copy selection in the algorithm. Simulation experiment results with CloudSim show that CDLBA-PM can not only reduce response time of user request, but also effectively balance load of edge cloud storage peer.