Research On Low Frequency And Ultra-low Frequency Stability Analysis And Control Strategy For Power Systems With High Renewable Energy Penetration

In recent years,with the rapid development of renewable energy power generation technology,the proportion of renewable energy power generation in the power system increases rapidly.Among them,the hydropower and wind power have the highest proportion.Large scale wind power and hydropower integration into the grid has significantly changed the characteristics of the system.In particular,the hydropower output is seasonal,while the wind power output is stochastic.The fluctuation of wind power and hydropower output,the uncertain fluctuation of its load,and the frequent occurrence of natural disasters on the transmission channel makes the transmission power of the system,the operation mode of the power grid,and the security and stability characteristics of the system complex and changeable,showing high-dimensional,timevarying,nonlinear characteristics.Which makes the transient stability problem of the system is prominent,and the ability of the power grid to withstand power disturbances is greatly reduced.Moreover,the problem of low frequency and ultra-low frequency oscillation is particularly prominent,which seriously threatens the safe and stable operation of power system.In this context,it is urgent to analyze the system stability characteristic of the wind power and hydropower access system and study the stability control strategies for such systems.Under the support of the general project of Nation Science Foundation of China "Analysis of low-frequency and ultra-low frequency oscillation characteristics and research on unified control strategy of integrated physics data driven enriched hydropower system"(52277083),this dissertation focuses on the characteristics analysis and stability control of low-frequency and ultra-low frequency oscillation of power system under renewable energy access.The main research contents are as follows:1)In order to study the impact of hydropower and wind power access on the stability of the power grid,this dissertation constructs the Philips-Heffron model of the wind turbine connected to the infinite bus system.Then,the complex torque coefficient method is utilized to analyze this model and the results show that the wind turbine connected to the system will produce damping torque,which will become negative under some operating conditions.In this context,the wind turbine can provide negative damping to the system,causing low-frequency oscillation of the system.In addition,this dissertation applys Routh criterion to analyze the small disturbance stability of hydraulic turbine governor,and the results show that the reason for the ultra-low frequency oscillation of the system is that the water hammer effect of the hydro-turbine causes the system to generate a large negative damping torque in the ultra-low frequency band,which can be alleviated by optimizing the PID parameters of the governor.Therefore,optimiting PID parameter settings can be applied to suppress the ultra-low frequency oscillation.2)This dissertation proposes two low frequency oscillation suppression strategies.One is online sparse adaptive control of multi machine PSS.Specifically,both the reinforcement learning algorithm and sensitivity analysis theory is combined to train the agent to learn the sparse adaptive parameter adjustment strategy of multi machine PSSs.Then,the well-trained agent is utilized for collaborative sparse update of PSS parameter settings under different wind speed conditions,so as to acheive the best performance of the controllers and improve the low frequency stability of the system.The other is the robust adaptive control of the data-driven STATCOM additional damping controller.Specifically,an additional damping controller is added to the STATCOM device.Moreover,considering the time-varying characteristics of the system,the neural network is applied to realize the dynamic identification of the equivalent transfer function under different operating conditions.Based on this,the agent is trained via the small gain theory and reinforcement learning method to obtain the adaptive robust parameter adjustment strategy of the additional damping controller,which has the advantages of both adaptive control and robust control.3)This dissertation proposes two kinds of ultra-low frequency oscillation suppression strategies.One is that a novel PR-PSS controller is proposed to suppress the ultra-low frequency oscillation.Specifically,the structure of the traditional PSS is improved,and the PR controller is introduced to form a novel PR-PSS controller.This type of PSS makes up for the defect that the traditional PSS always provide the ahead of the phase in the ultra-low frequency band,and cannot provide sufficient damping for the ultra-low frequency oscillation mode,which can be used for ultra-low frequency oscillation suppression.The other is that a governor PID robust optimztion is proposed to suppress ultra-low frequency oscillation.During the optimization process,three principles is considered: adapting to extreme operating conditions,not deteriorating other oscillation modes,and retaining the dynamic regulation capability of the governor.To this end,the governor PID optimization problem is modeled as a Min-Max two-layer robust optimization model,and the reinforcement learning training-enabled agent is introduced to replace the inner model,The two-layer optimization problem is transformed into a single-layer nonlinear optimization problem,which is solved by combining the heuristic algorithm,thus improving the model solving efficiency.4)This dissertation proposes a multi-band oscillation suppression strategy to prevent low-frequency and ultra-low frequency oscillation.In order to effectively suppress the potential risk of low-frequency and ultra-low frequency coupled oscillation in the system,a multi-band damping controller is proposed to realize the frequency division decoupling control of low-frequency and ultra-low frequency oscillation.Considering that multiple multi-band damping controllers can be configured in the system,the parameter tuning problem of the multi controller is modeled as a Markov game,and the multi-agent reinforcement learning algorithm is introduced to train multiple agents to learn the coordinate parameter settings adjustment strategy of severeal multi-band damping controllers.After the training,each agent adjusts the parameters of the proposed controller according to local state.In this way,the distributed collaborative self optimizing control is realized to ensure the stability of the system under different working conditions.