Research On Consensus Of The Multi-agent Systems And Its Applications On Distributed Control Of Smart Grids

In August 2009,the State Grid Corporation of China released the"Comprehensive Research Report on the Construction of a Uniform and Strong S-mart Grid".In 2015,the electric power equipment was included in one of the ten key development areas of "Made in China 2025".In July 2016,the "Thirteenth Five-Year National Science and Technology Innovation Plan" released 9 major projects,including smart grids,big data,intelligent manufacturing and robotics etc.,and the smart grid is one of them.It can be seen that the smart grid industry has gradually become a national key development area.By 2020,the new energy will be intercon-nected orderly.The grid will incorporate more than 150 million kilowatts of wind energy,more than 32 million kilowatts of hydropower,and more than 20 million kilowatts of photovoltaic power.Smart grids must have a high level of intelligence and extensive distribution,and be able to adapt to the flexible access and offset adjustment of all types of clean energy to meet the diverse needs of users.Distributed artificial intelligence will be the core of the smart grid construction,Coordinated control of multi-agont systems is a hot issue in distributed artificial intelligence research.One of the most basic issues in this field is consensus problem.With developing of the research on the cooperative control of multi-agent systems,the consensus problem has be viewed as a hot cutting-edge issues in this field.For smart grids,due to the large number of intermittent energy sources,dis-tributed energy sources,and flexible loads,smart grid modeling becomes more and more difficult,even impossible to model or the model is not available.In addition,the smart grid has plug-and-play features,so the grid will inevitably have some inherent defects,which will affect the stable operation of the grid,and even cause the collapse of the grid.Based on these difficulties,the main research work of this discussion is as follows:(1).An internal model compensator is designed for the uncertain part of smart grids,and the distributed compensator is designed for the heterogeneous smart grids.Then a distributed state feedback control law is give based on internal model compensator and distributed compensator.Thus this distributed control law could achieve the control objective,and also reject the external disturbance of smart grids.(2).A reduced-order observer is designed for the unmeasurable part of smart grids.Then based on the observer,a distributed dynamic output feedback control law is designed.Applying this algorithm can make the observer error of the unmea-surable part of the smart grid quickly converge to zero,which also greatly reduces the synchronization time of the entire smart grid system.(3).A model-free approximate dynamic programming(ADP)algorithm based on policy iteration is proposed.The main advantage of the model-free distributed design method is that it can achieve the specified convergence,speed without knowing the precise mathematical model of the system,and the speed and accuracy of the smart grid systems is also improved.Then it proves that the model-free distributed control algorithm can make the tracking error of the smart grid system strictly converge to zero.(4).A finite-step stable approximate dynamic programming algorithm is pro-posed.It is proved for the first time that for a value iterative algorithm,there is a finite iterative step that makes the iterative controller stable.The performance of the optimal problem is a quadratic performance index with a discount factor.It is proved that the optimal control of the smart grid system based on this algorithm is stable,and at the same time,the superiority of introducing the discount factor is proved.