多模型自适应控制应用于大型光伏电站阻尼区间振荡

Application of Multiple-model Adaptive Control for Damping Inter-area Oscillations by Large-scale Photovoltaic Plant

研究表明对大型光伏电站引入附加阻尼控制,可以有效抑制互联电力系统的区间低频振荡。为解决传统阻尼控制方法在应对大扰动方面的不足,引入多模型自适应控制策略进行阻尼控制。采用K-medoids算法以低频振荡模式为多维特征点对运行状态对应的线性模型进行聚类,并基于各类的中心模型分别设计阻尼控制器。通过比较实际系统与模型的输出动态响应差异,用贝叶斯方法实时计算每一个模型匹配当前实际系统的概率,据此更新各阻尼控制器输出权重,各控制器输出的加权平均作为最终控制信号附加于光伏无功控制的参考指令中。在DIgSILENT/PowerFactory中搭建光伏并网四机两区域测试系统,非线性仿真结果表明,在无需系统受扰后状态先验知识的情况下,该策略可以有效抑制非预期扰动引发的区间振荡,并在反馈信号时延和光照随机波动下均保持良好的鲁棒控制性能。

Studies suggested that supplementary damping control of large-scale photovoltaic （PV） plants could be used to effectively damp inter-area low frequency oscillations in interconnected power system. Multiple-model adaptive control （MMAC） was introduced for damping to overcome the incapability of conventional damping control methods in dealing with big disturbance. The K-medoids algorithm was used to cluster models associated with operating conditions, low frequency oscillation modes being the features, and dampers were designed based on individual centroid models of the clusters respectively. Based on the deviation between the output dynamic responses of the actual system and models, the probability of each model representing the actual system in real time was calculated through Bayesian approach. Weights of damper＇s outputs were updated, and the weighted average of the individual damper＇s outputs was attached to the reactive power control of the PV plant as the resultant damping control signal. A four-machine two-area test system connected with a PV plant was built using DIgSILENT/PowerFactory, and the nonlinear simulation results show that the proposed control strategy is efficient for damping inter-area oscillations caused by unexpected disturbances without any prior knowledge about postdisturbance states of the system. Robustness for time delays and random fluctuation of irradiation is also verified.