The invention discloses a boiler combustion oxygen content modeling method. The method comprises the following steps of: acquiring history big data of a total blast capacity, a total coal quantity anda hearth oxygen amount of a boiler combustion system to serve as original data; screening a sample data set representing system characteristics by utilizing an optimal matching empowerment method; carrying out improvement by utilizing a Bayesian algorithm on the basis of a classical neural network model; taking a weight value as probability distribution of the whole weight space; obtaining an approximate globally optimal solution under a probability meaning by utilizing a calculation method of random simulation; converting a weight value optimization problem into a regular error function minimum value solving problem; and finally obtaining a boiler combustion oxygen content modeling result. The method is capable of simplifying the data set and providing reference for dynamic system identification carried out through combining big data and an intelligent identification method, is relatively high in practicability, and has important practical significance for improving the combustion efficiency of heat-engine plants.