Research On OIDQ Automatic Generation Control Strategy For Large-scale Wind Power Access

The effective way to solve the "carbon peak,carbon neutral" problem is to build a new power system with new energy as the main body.Driven by the technological innovation of new energy generation and the transformation of energy structure,the modern power grid has been gradually transformed into a new form of multi-regional energy interconnection under the deep information-physical integration.However,with the large-scale interconnection of new energy sources,traditional synchronous machines are gradually replaced,primary frequency regulation is weakened,and the generation methods represented by new energy sources such as wind and photovoltaic have uncontrollable factors such as resource intermittency and randomness,making it difficult to maintain a balance between the actual power supply of the power system and compliance with demand,and the traditional AGC strategy is no longer sufficient to meet the frequency control requirements of complex power grids.At the same time strong random disturbances cause a large impact on the interconnected grid,making the grid frequency control more and more difficult.The distributed multi-region large-scale new energy grid-connected mode needs to explore novel frequency control methods from the mechanism of frequency stability.Therefore,in this paper,an improved reinforcement learning algorithm is proposed from the perspective of AGC for the multi-regional interconnected grid system with large-scale wind power grid connection to realize the optimal control of multi-regional interconnected grid with different wind power penetration rates,improve the grid frequency dynamic control performance,and solve the frequency instability problem of power system brought by wind power grid connection.However,the power generation mode represented by wind energy,photovoltaic and other new energy sources has uncontrollable factors such as intermittency and randomness of resources,which makes it difficult for the power system to maintain a balance between actual power supply and demand.The traditional AGC strategy is no longer enough to meet the frequency control requirements of complex power grids.Therefore,this paper proposes an improved reinforcement learning algorithm from the perspective of AGC for the multi-regional interconnected power grid system with large-scale wind power grid connection,so as to achieve the optimal control of multi-regional interconnected power grid under different wind power permeability,improve the dynamic control performance of power grid frequency,and solve the problem of frequency instability of power system caused by wind power grid connection.Firstly,this paper summarizes the research status of AGC control strategy and wind power participation in frequency modulation at home and abroad,introduces the basic principle and system structure of multi-region interconnected power system AGC,and focuses on the operation mode and control strategy of AGC system.The application of reinforcement learning in AGC is described,and the mechanism of reinforcement learning and the control mode of Q learning in AGC are introduced.Secondly,a multi-region interconnected AGC system model is built.Each control part of the classical two-zone load frequency control model,doubly-fed fan model and frequency modulation strategy are introduced in detail.Considering the participation of electric vehicle PEVs in frequency modulation,an improved two-zone load frequency control model was established.For large-scale wind power grid-connection,a four-region interconnected power grid model with wind turbines as disturbance and participation system frequency modulation under different wind power permeability is established.The experimental foundation is laid for the simulation analysis of the control performance of the proposed intelligent algorithm.Then,an optimistic and exploratory double estimator reinforcement learning algorithm,OIDQ algorithm,is proposed,which adopts the optimistic initialization principle to expand the space of action exploration of agents,and integrates double estimators to solve the problem of overestimating the traditional reinforcement learning action value,so as to obtain the optimal control.Based on this algorithm,AGC system design and a series of simulation analysis under different interconnected power grid models are carried out.By comparing the simulation results of various reinforcement learning algorithms,it is proved that OIDQ algorithm has good performance in optimizing the performance indexes of the system such as frequency deviation,ACE and CPS,and has better control performance compared with other algorithms.Finally,the paper summarizes the work,reflects on the improvement of the research content,and explores the future research work of AGC system in which fans participate in frequency modulation.