基于改进AlexNet的电力系统暂态功角失稳紧急控制策略

Emergency Control Strategy for Transient Angle Instability of Power System Based on Improved AlexNet

随着新能源渗透率的提升,电网环境日益复杂,电力系统安全稳定运行也面临着新的挑战。为了满足电力系统暂态功角失稳后的实时紧急控制决策,采用深度学习与紧急控制相结合的方法,提出一种基于改进AlexNet网络的电力系统暂态功角失稳紧急控制策略。首先基于改进AlexNet对失稳发电机功角轨迹进行预测,识别临界机群;然后定义紧急控制动作灵敏度指标,建立改进AlexNet灵敏度预测模型,拟合发电机功角特征与紧急控制动作灵敏度的映射关系,从而确定紧急控制的动作母线;最后以切除发电机和负荷容量最小为目标,建立紧急控制优化模型并求解最优策略,并在新英格兰10机39节点系统进行算例验证。结果表明,针对电力系统暂态功角失稳问题提出的基于深度学习的功角轨迹预测模型和紧急控制灵敏度预测模型,均有较高的预测精度。在此基础上制定的紧急控制策略能使失稳系统快速恢复稳定运行,加强电网安全稳定防御体系。

With the increase of new energy penetration,the power grid environment is increasingly complex,and the safe and stable operation of power system is also facing new challenges.In order to satisfy the real-time emergency control after power system transient angle instability,an emergency control strategy based on improved AlexNet is proposed by combining deep learning with emergency control.Firstly,the power angle trajectories of unstable generators are predicted based on the improved AlexNet to identify critical generators.Then,the sensitivity of emergency control action is defined,and the improved AlexNet sensitivity prediction model is established to fit the mapping relationship between power angle characteristics and the sensitivity,so as to determine the action bus of emergency control.Finally,the emergency control optimization model is established and the strategy is solved with the goal of minimum capacity of the generator tripping and load shedding,and an example is given to verify the model in New England 10 machine 39 bus system.The results show that both the power angle trajectory prediction model and the emergency control sensitivity prediction model based on deep learning have high prediction accuracy.The emergency control strategy formulated on this basis can enable the unstable system to quickly return to stable operation and to strengthen the security and stability defense system of power grid.