Abstract: Under deep peak regulation state, the performance of primary frequency regulation of thermal power units has undergone significant changes. In order to figure out the changes of generator primary frequency regulation output and provide reference for the power grid to assess the performance of primary frequency regulation of deep peak regulation unit, the model structure is determined based on the prior knowledge of the primary frequency regulation model. Then the unknown parameter values in the model are derived by using an improved group optimization algorithm with better global search capability. Steady state values are incorporated into the identification parameters in order to eliminate modeling error which was introduced from manually selected steady state values. In this way, both the initial state and the end state converge to the steady state as required in conventional model identification process. The model parameter calculation results show that under deep peak regulation state, the capability margin of primary frequency regulation is higher than that under the normal load conditions. Therefore, relevant parameters need to be adjusted to better utilize the primary frequency regulation capability of the deep peak regulation unit.

Key words: deep peak regulation, primary frequency regulation, particle swarm optimization algorithm, parameter identification