【摘要】 为支撑跟网/构网混合型新能源场站中机组控制方式快速切换，实现可适应电网强度变化的新能源场站安全稳定运行，提出基于时间卷积神经网络（temporal convolutional network,TCN）的跟网/构网混合型新能源场站并网系统小干扰稳定性快速评估方法。首先，构建跟网/构网混合型新能源场站聚合阻抗模型，通过特征值计算得到并网系统小干扰稳定裕度。然后，以并网系统短路比和新能源场站跟网/构网控制方式信息作为输入特征，以并网系统小干扰稳定裕度和阻尼比作为输出特征，训练TCN得到混合型新能源场站并网系统小干扰稳定性快速评估模型。经过训练的模型可根据短路比和跟网/构网混合型新能源场站中各机组的控制方式快速输出对应的小干扰稳定裕度和阻尼比。最后，以一个含10台风电机组的新能源场站为对象进行算例分析。结果表明：所提TCN方法相比于长短期记忆神经网络方法，在小干扰稳定裕度和阻尼比预测上的平均绝对百分比误差分别降低16.76%、14.75%；所提方法的计算耗时相对于特征值计算方法降低98.54%，从而验证所提小干扰稳定性快速评估方法的准确性与时效性。更多还原

【Abstract】 To support the rapid switching of unit control modes in grid-following/grid-forming hybrid new energy stations and achieve safe and stable operation of these stations that can adapt to changes in grid strength, a rapid assessment method for small-signal stability of grid-connected systems in grid-following/grid-forming hybrid new energy stations based on temporal convolutional network(TCN) is proposed. Specifically, an aggregated impedance model for grid-following/grid-forming hybrid new energy stations is constructed, and the small-signal stability margin of the grid-connected system is obtained through eigenvalue calculations. Furthermore, using the short-circuit ratio of the grid-connected system and the information on the grid-following/grid-forming control mode of the new energy station as input features, and the small-signal stability margin and damping ratio of the grid-connected system as output features, a TCN is trained to obtain a rapid assessment model for small-signal stability of grid-connected systems in hybrid new energy stations. The trained model can quickly output the corresponding small-signal stability margin and damping ratio based on the short-circuit ratio and the control mode of each unit in the grid-following/grid-forming hybrid new energy station. A case study is conducted using a new energy station with 10 wind turbines, and the results show that compared to the long short-term memory neural network method, the proposed method reduces the mean absolute percentage error of small-signal stability margin prediction and damping ratio prediction by 16.76% and 14.75% respectively.Additionally, the computation time of the proposed method is reduced by 98.54% compared to the eigenvalue calculation method, verifying the accuracy and efficiency of the proposed rapid assessment method for small-signal stability.更多还原