CN-122026397-A - Low-frequency oscillation on-line inhibition method and device for wide-area additional damper of doubly-fed wind turbine based on reinforcement learning algorithm

CN122026397ACN 122026397 ACN122026397 ACN 122026397ACN-122026397-A

Abstract

The invention discloses a low-frequency oscillation on-line inhibition method of a doubly-fed wind turbine wide-area additional damper based on a reinforcement learning algorithm, which comprises the following steps of obtaining operation data of a wind power generation system, including wind speed, wind turbine output power, rotating speed and grid frequency; the method comprises the steps of preprocessing acquired data, carrying out trend removal and normalization processing by utilizing a band-pass filter, training a model by utilizing a depth deterministic strategy gradient algorithm to predict the running state of a fan and the oscillation condition of a system, generating an optimal damping control strategy based on the trained depth deterministic strategy gradient algorithm model, adjusting running parameters of the fan in real time, and applying the generated damping control strategy to the fan control system to improve the damping characteristic of the system and reduce low-frequency oscillation.

Inventors

LI ZEWEI
SUN ZHENGLONG
ZHANG LIN

Assignees

东北电力大学

Dates

Publication Date: 20260512
Application Date: 20251231

Claims (7)

1. The low-frequency oscillation on-line inhibition method of the doubly-fed wind turbine wide-area additional damper based on the reinforcement learning algorithm is characterized by comprising the following steps of: The method comprises the steps of obtaining operation data of a wind power generation system, including wind speed, fan output power, rotating speed and grid frequency, carrying out trend removal and normalization processing on the obtained operation data of the wind power generation system, including wind speed, fan output power, rotating speed and grid frequency data, utilizing a band-pass filter, training a model to predict the operation state of the fan and the oscillation condition of the system by utilizing DDPG namely depth deterministic strategy gradient algorithm, generating an optimal damping control strategy based on the trained DDPG model, adjusting the operation parameters of the fan in real time, and applying the generated damping control strategy to the fan control system to improve the damping characteristic of the system and reduce low-frequency oscillation.
2. The online low-frequency oscillation suppression method for the wide-area additional damper of the doubly-fed wind turbine based on the reinforcement learning algorithm according to claim 1, wherein the DDPG algorithm improves the learning stability of a model by creating a target Actor network and a target Critic network.
3. The online low-frequency oscillation suppression method for the wide-area additional damper of the doubly-fed wind turbine according to claim 1, wherein in the step of generating the optimal damping control strategy, the running parameters of the wind turbine are adjusted in real time by utilizing a strategy generated by a trained DDPG model so as to improve the damping characteristic of the system.
4. A reinforcement learning algorithm-based double-fed fan wide-area additional damper device that performs a reinforcement learning algorithm-based double-fed fan wide-area additional damper low-frequency oscillation on-line suppression method according to any one of claims 1 to 3, characterized by comprising: the acquisition module is used for collecting operation data of the fan and power grid frequency data; the processing module is used for preprocessing the acquired data and training DDPG models; and the control module is used for generating and applying a damping control strategy according to the trained DDPG model, so that the stability of the system is improved.
5. The reinforcement learning algorithm based doubly fed wind turbine wide area additional damper assembly of claim 4 wherein said control module is configured to collect wind speed, turbine output power, rotational speed and grid frequency data.
6. The reinforcement learning algorithm based doubly fed fan wide area additional damper apparatus according to claim 4, wherein said processing module uses a bandpass filter to trend and normalize the collected data.
7. The reinforcement learning algorithm-based doubly-fed wind turbine wide-area additional damper device of claim 4, wherein the control module utilizes DDPG models to generate an optimal damping control strategy to adjust operating parameters of the wind turbine in real time.

Description

Low-frequency oscillation on-line inhibition method and device for wide-area additional damper of doubly-fed wind turbine based on reinforcement learning algorithm Technical Field The invention relates to the technical field of power systems, in particular to a low-frequency oscillation on-line inhibition method and device of a wide-area additional damper of a doubly-fed fan based on a reinforcement learning algorithm. Background With the rapid development of wind power in the global scope, wind power generators, particularly doubly fed fans, have become the main stream model of the wind power industry. The doubly-fed fan is connected with the grid through power electronic equipment, the back-to-back converter is controlled to decouple the rotating speed of the fan from the frequency of the grid, and the rotating kinetic energy of the system is hidden, so that insufficient damping is caused. This characteristic, together with the randomness and volatility of the wind energy, leads to a significant increase in the risk of low frequency oscillations of the power system. Modern power systems inherently present a risk of low frequency oscillations due to the expansion of the scale, large capacity long distance transmission and the widespread use of fast exciting devices. The risk is aggravated by the wind power connection, and the traditional synchronous generator set can be directly connected with a power grid, and generates damping torque through a rotor and stator loop of the generator, so that system oscillation is effectively restrained. However, the fan is connected with the grid through the power electronic device, the internal resistance is small, the contribution to oscillation suppression is weak, and the stability of the system faces new challenges. To address these issues, a double-fed fan wide area damping controller based on a reinforcement learning algorithm has been developed. The reinforcement learning algorithm has the characteristics of strong self-adaptability and capability of self-learning optimization in a complex environment, and is particularly suitable for the application scene with high uncertainty of a wind power system. Through reinforcement learning algorithm, the controller can adjust the operation parameters of the fan in real time, improve the damping characteristic of the system, reduce low-frequency oscillation and ensure the stable operation of the wind power system. Content of the application The invention aims to provide an online low-frequency oscillation suppression method for a wide-area additional damper of a doubly-fed wind turbine based on a reinforcement learning algorithm, so as to improve the stability and reliability of a power system. The invention utilizes the reinforcement learning algorithm to adaptively adjust the operation parameters of the fan and enhance the damping characteristic of the system, thereby effectively inhibiting low-frequency oscillation. In order to solve the technical problems, the invention provides a low-frequency oscillation suppression method of a wide-area additional damper of a doubly-fed fan based on a reinforcement learning algorithm, which comprises the following steps: The method comprises the steps of obtaining operation data of a wind power generation system, wherein the operation data comprise wind speed, fan output power, rotating speed and grid frequency, conducting trend removal and normalization processing on the obtained operation data of the wind power generation system, comprising wind speed, fan output power, rotating speed and grid frequency data, training a model to predict the operation state of the fan and the oscillation condition of the system by utilizing DDPG namely depth deterministic strategy gradient algorithm, generating an optimal damping control strategy based on the trained DDPG model, adjusting the operation parameters of the fan in real time, applying the generated damping control strategy to the fan control system, improving the damping characteristic of the system, reducing low-frequency oscillation by 1. Data acquisition comprises the steps of obtaining the trend data of grid operation, and collecting the operation data of the fan, comprising wind speed, fan output power, rotating speed, grid frequency and the like in the fan operation process. Preferably, the DDPG algorithm improves the learning stability of the model by creating a target Actor network and a target Critic network; preferably, in the step of generating the optimal damping control strategy, the running parameters of the fan are adjusted in real time by using a strategy generated by a trained DDPG model so as to improve the damping characteristic of the system. In order to solve the technical problem, the invention provides a DDPG algorithm-based wide-area additional damper device of a doubly-fed fan, which comprises: and the acquisition module is used for collecting the operation data of the fan and the power grid frequency data. And