The invention discloses a power plant boiler NOx prediction model optimization method based on an improved quantum particle swarm algorithm, and the method comprises the following steps: 1, carrying out the mechanism analysis of a boiler combustion system of a coal-fired unit, and determining the input variable of a NOx emission concentration prediction model; 2, combining a cosine decreasing function with a quantum particle swarm optimization algorithm, and providing an improved quantum particle swarm optimization algorithm; and 3, optimizing initial parameters of the extreme learning machineby utilizing an improved quantum particle swarm optimization algorithm. Establishing an accurate NOx emission model by taking the error absolute value sum minimization of a training data prediction value and an actual value as a target; and 4, through simulation verification, the precision of the model optimized by the improved quantum particle swarm algorithm is higher than that of the model optimized by other methods. The method has the advantages that the optimal initial parameters of the extreme learning machine can be efficiently and rapidly calculated through the improved quantum particle swarm optimization algorithm, then the accurate thermal power plant boiler NOx emission model is obtained, and the method is of great significance for reducing pollutant emission of a coal-fired unit.