Abstract:
Aiming to address the instability caused by the high proportion of uncertain wind turbines connected to the grid, as well as the weakened frequency modulation ability due to the inertia reduction of the power system, a wind power prediction and primary frequency modulation strategy is proposed for integrated wind turbines and energy storage systems based on long and short term memory networks. Firstly, based on the spatial clustering algorithm and correlation analysis method, the equivalent modeling for a wind farm is conduced by employing historical data. With this model, the ultra-short term power prediction of wind power output is carried out based on long and short term memory networks. Secondly, the coordination control strategy of variable speed or variable pitch frequency modulation is selected according to the wind speed range. Finally, the auxiliary frequency modulation depth of power type flywheel storage and energy type battery storage is adjusted according to the predicted output deviation, frequency deviation and change rate of the wind turbines. Simulation experiments are carried out on a primary frequency modulation response model for integrated wind turbines and energy storage systems, which is built on Matlab/Simulink. The results demonstrate that the average prediction accuracy of wind power LSTM model reaches 97.62%, while the maximum frequency deviation of the proposed strategy is reduced by 11.3% and 28.6% respectively, compared with the fixed proportion strategy and the single wind power frequency modulation. In addition, the stabilization time is reduced by 27.1% and 35.2% respectively, thereby significantly improving both the frequency modulation performance of wind power and frequency characteristics of the power system.