Reinforcement learning algorithm-based self-correction control method for double-fed induction wind power generator

Abstract

The invention discloses a reinforcement learning algorithm-based self-correction control method for a double-fed induction wind power generator. According to the method, an RL controller is added to a PI controller of a vector control system based on PI control to dynamically correct the output of the PI controller; the RL controller comprises an RL-P controller and an RL-Q controller; and the RL-P controller and the RL-Q controller are used for correcting active and reactive power control signals respectively. The Q learning algorithm is introduced to the method to be used as the reinforcement learning core algorithm; the reinforcement learning control algorithm is insensitive to a mathematical model and an operating state of a controlled object while the learning capability has relatively high adaptivity and robustness on parameter changes or external interference, so that output of the PI controller can be optimized rapidly and automatically online; and by virtue of the reinforcement learning algorithm-based self-correction control method, high dynamic performance is achieved, and the robustness and the adaptivity of the control system are obviously reinforced.

Description

technical field [0001] The invention relates to a self-calibration control method of a doubly-fed induction wind power generator, in particular to a self-calibration control method of a doubly-fed induction wind power generator based on a reinforcement learning (Reinforcement Learning, RL) algorithm. Background technique [0002] Variable-speed constant-frequency double-fed power generation is a power generation method commonly used in wind power generation at present, and its generator uses a double-fed induction generator (DFIG). When the unit works below the rated wind speed, by adjusting the rotor speed of the generator, the best blade tip speed ratio is maintained to achieve the maximum capture of wind energy. Its control system often adopts vector control based on stator field orientation to realize decoupling control of generator active and reactive power. [0003] Due to the strong randomness and time-varying nature of wind energy, and the system contains dynamic pa...

AI Technical Summary

Problems solved by technology

[0003] Due to the strong randomness and time-varying nature of wind energy, and the system contains dynamic parts that are not modeled or cannot be modeled accurately, the doubly-fed power generation system becomes a multivariable, nonlinear, and strongly coupled system, so it is difficult to use only traditional vector control Meet the requirements of the control system for high adaptability and high robustness

At present, various control schemes are used, but the control effect is not very ideal. For example, the neural network control scheme is adopted. This control scheme improves the control performance, but the steady-state error is relatively large.

However, the fuzzy sliding mode control strategy is used to combine fuzzy control and sliding mode control. Although a good control effect has been achieved, the implementation is more complicated.

Images