The present invention provides a block chain multi-dimensional random number consensus recommendation method, including: all specific nodes in the alliance randomly generate a set of vector random numbers Xi(xi,1,xi,2,...,xi,j,..., xi, m); all specific nodes synchronize random numbers Xi(xi, 1, xi, 2, ..., xi, j, ..., xi, m); each specific node first performs component average calculation, and then performs Calculate the comprehensive average value, and finally calculate the average value of the random number; in each specific node, by comparing the distance between the average value of the random number and the comprehensive average value, select the node Pi with the smallest or largest distance holding the average value of the random number as a candidate Bookkeeper: All specific nodes synchronize candidate bookkeepers with each other, and use the same candidate bookkeeper recommended by each node as the bookkeeper.