电工技术学报2026，Vol.41Issue(2)：575-590,648,17.DOI:10.19595/j.cnki.1000-6753.tces.250156
Koopman响应驱动预测控制的光伏并网系统次同步振荡鲁棒抑制策略研究
Research on Subsynchronous Oscillation Suppression Strategy of PV Grid-Connected Systemvia Koopman-Based Response-Driven Predictive Control
摘要
Abstract
Under the background of the development of the power system driven by the goals of"carbon peaking"and"carbon neutrality",renewable energy sources represented by wind and photovoltaic(PV)power in China have witnessed leapfrog growth,which are gradually becoming the main primary power sources of the power system.Due to the inherent weak anti-interference ability and low damping characteristics of PVs,their adverse dynamic interactions with the power grid have significantly exacerbated the risk of subsynchronous oscillation(SSO)instability in the power system.The PV grid-connected system exhibits dynamic characteristics of high-dimensional complexity,strong nonlinearity,and significant uncertainty,rendering traditional model-based SSO suppression strategies more prone to failure.Therefore,to address this issue,this paper builds a bridge between response information and system behavior,and proposes a SSO robust suppression strategy for PV grid-connected systems via Koopman-based response-driven predictive control. Firstly,the system is reconstructed using finite time-domain data based on behavioral system theory,and then the dynamic behavior of nonlinear systems is accurately identified in the Koopman linearized observable space.Secondly,the identification,prediction,and control of system behavior are integrated into an optimal control strategy design,and regularization and relaxation techniques are introduced to enhance control robustness.Subsequently,online SSO suppression is accomplished through rolling optimization.Thirdly,theoretical analysis reveals the robustness enhancement mechanism of regularization and relaxation techniques in the suppression strategy,and a fast solution method for this strategy is developed based on the analytical expression of the optimal control law.Finally,through time-domain simulations and hardware-in-the-loop(HIL)experiments,the effectiveness,robustness,and practicability of the SSO suppression strategy are verified in uncertain conditions such as noise interference,parameter variations,source/load fluctuations,and power grid topology changes. The main conclusions of this paper are summarized as follows: (1)A behavioral system theory combined with the Koopman operator is proposed to realize finite-time domain data reconstruction of nonlinear systems,establishing a bridge between response information and the dynamic behavior of complex systems.Simulation results show that the proposed response-driven prediction model accurately identifies the SSO behavior of PV grid-connected systems with significant nonlinear characteristics. (2)A Koopman response-driven predictive control framework for SSO suppression is constructed,organically integrating system behavior identification,prediction,and control into a linear quadratic optimization problem.This integration enables online multi-step receding horizon optimization,which effectively addresses bounded disturbances and suppresses SSOs. (3)Based on Matlab time-domain simulations and HIL experiments,the effectiveness,robustness,and practicality of the proposed suppression strategy are verified in different uncertain conditions.Case studies indicate that compared to linear control strategies or those based on deterministic equality constraints,the proposed response-driven suppression strategy yields more accurate predictions,providing technical control support to enhance the dynamic stability of PV grid-connected systems.关键词
次同步振荡/响应驱动预测控制/非线性/Koopman算符/鲁棒性Key words
Subsynchronous oscillation(SSO)/response-driven predictive control/nonlinearity/Koopman operator/robustness分类
信息技术与安全科学引用本文复制引用
王子涵,郑乐,曹竣淇,伍珀苇,李佳晏,李庚银..Koopman响应驱动预测控制的光伏并网系统次同步振荡鲁棒抑制策略研究[J].电工技术学报,2026,41(2):575-590,648,17.基金项目
国家重点研发计划项目"响应驱动的大电网稳定性智能增强分析与控制技术"(2021YFB2400800)、金风科技项目"构网型与随网型控制的交互影响机理"(10012000176124082701)资助. （2021YFB2400800）
