基于机器学习可解释代理模型的风电次同步振荡 在线预测及优化控制方法

随着新能源发电的快速发展,大规模风电场与电网相互作用引起的次同步振荡问题不断凸显,对次同步振荡进行预测并采取预防性控制措施具有重要意义。为此,提出了一种基于机器学习可解释代理模型的风电并网系统次同步振荡的在线预测和优化控制方法。采用Prony算法分析电网小扰动过程以辨识系统阻尼水平。建立了基于梯度提升树模型的系统阻尼评估系统。提出了基于可解释代理模型的优化控制辅助决策方法。在Matlab Simulink中搭建了多个直驱风电场并网系统的仿真模型,验证了所提方法能够有效进行次同步振荡在线预测和优化控制,从而抑制次同步振荡,提升系统的稳定性。相较于传统的抑制方法,所提方法不依赖风电系统的详细模型,能够对风电场进行有针对性的控制,且控制措施的效果是可估测的。

Online prediction and optimal control method for subsynchronous oscillation of wind power based on an interpretable surrogate model for machine learning

With the rapid development of new energy power generation, the problem of Subsynchronous Oscillation (SSO) caused by the interaction of large-scale windfarms with the power system has become increasingly prominent. It is of great significance to predict SSO and adopt preventive control measures. In this paper, an online prediction and optimal control method for SSO of wind power grid-connected systems based on an interpretable surrogate model for machine learning is proposed. The Prony algorithm is used to analyze the small disturbance process of a power system to identify the damping level. An evaluation method for system damping based on a gradient boosting decision tree model is established. An optimal control auxiliary decision-making method based on the interpretable surrogate model is proposed. Simulation experiments are conducted on a wind power grid-connected system with multiple Direct-Drive Permanent Synchronous Generators (D-PMSGs) built in Matlab Simulink to verify the effectiveness of the proposed method for online prediction and optimal control of SSO. This could suppress the SSO and improve the stability of the system. Compared with the traditional suppression method, the proposed method does not rely on a detailed model of the wind power grid-connected system while regulating the wind farm pertinently, and the effect of the control measures can be estimated. This work is supported by the National Key R & D Program of China (No. 2016YFB0900600) and the Science and Technology Project of State Grid Corporation of China (No. 52094017000W).