Given the disadvantages of poor expansibility and strong experience dependence of existing binary consensus algorithms, we propose an Nstate distributed binary consensus algorithm. Firstly, based on the idea of average consensus of the Gossip algorithm and the roulette idea, the degree of deviation between the state average of the wireless sensor network and the current state average is updated, and the initial probabilistic distribution of all possible update states is calculated. Secondly, the genetic algorithm is applied to optimize the initial probability distribution and obtain the optimal probability distribution with better accuracy. The results show that the proposed algorithm has higher accuracy and shorter convergence time under the same number of states.