The invention relates to a self-tuning method of a main regulator. The method comprises the following steps: initializing parameters and a particle swarm, enabling each particle position to contain three variables, carrying out variation operation on the particles under the condition that the particles meet variation conditions, updating the positions of the particles, and carrying out updating iteration on weights, speeds, positions and adaptive values of the particles under the condition that the particles do not meet variation conditions; under the circumstance of meeting the variation conditions, using one particle as a current Pi under the condition that the adaptive value of the particle is better than the adaptive value of an optimal position Pi passed by the particle for each particle; for each particle, comparing the adaptive value with an adaptive value of an optimal position Pg which the whole particle swam passes and using the particle as a current Pg under the condition that the adaptive value of the particle is better than that of the optimal position Pg which the whole particle swam passes; and with the optimal position Pg of the current particle swarm as an initialsearching point, carrying out local searching by invoking a Rosenbrock algorithm, updating the Pi and Pg, and outputting three variables of the particle. Therefore, the system response speed can be increased.