Research On Confidence Power's Real-time Evaluation Of Distributed Generators And Its Optimal Scheduling

The production and life of human beings can not be separated from energy supplying,electricity is the most important form of energy supplying at present.The traditional centralized power supplying has poor reliability,high dependence on fossil fuels,severe damage to the environment,and long-distance transmission over long lines,resulting in high investment costs.Distributed energy has the advantages of high power generation efficiency,less environmental pollution,more flexible and convenient facility construction,and has become one of the power reform directions.However,with the continuous increase of distributed power capacity,the problems in operation are becoming increasingly prominent.It is of great significance to study the confidence power's real-time evaluation of distributed generators and its optimal scheduling to the stable operation and healthy and sustainable development of power grid.The principle of power generation for a variety of distributed power sources is discussed systematically,and the power output characteristics are also studied,thus a method for distributed generators' real-time assessment of confidence power and its optimal scheduling are proposed in microgrid.Firstly,a real-time compensated adaptive Kalman filter algorithm is proposed and applied to the real-time evaluation of the clean distributed generators' confident power availability.With the short-term power forecasted value and historical output value,we can get the evaluated value of the clean distributed generators'power in the real-time dispatching cycle,and then use the Gaussian distribution to fit the error probability distribution,and use the Hoeffding inequality to determine the minimum required energy storage device capacity,then based on the evaluation value of clean distributed power and the state of charge of the energy storage system,the power's evaluation value of the hybrid system of photovoltaic or wind turbines with energy storage devices can be obtained,so that the output power can be stabilized under high confidence.Then,considering the constraint conditions such as the charge-discharge climbing and storage capacity limits of energy storage system,a double-mechanism Q-learning algorithm,based on rewards and punishments is proposed for the first time as a strategy algorithm for charging and discharging energy storage system,and guided by the global optimal penalty term converges quickly,the trained Q value table can be obtained by learning in stages;in the real-time scheduling cycle,the actual state quantity may be different from the predicted value,the energy storage system only needs to correct its state quantity in real time based on the trained Q value table can not only achieve the optimal effect of clipping and valley filling,but also meet the real-time requirements.Finally,a bidding game model is proposed to deal with the possible artificially high-price bidding problem of generators under single-price rule,and the existence of the game model Nash equilibrium is proved.By formulating the rules of reverse auctions,each agent of power generation companies negotiates bidding to obtain their respective purchase price and power output,so that the distributed energy resources with lower unit generation cost is gived priority to send power and the purchase price is more reasonable.