Font Size: a A A

Smart Power Generation Control For Microgrids Islanded Operation Based On Reinforcement Learning

Posted on:2013-11-04Degree:MasterType:Thesis
Country:ChinaCandidate:H H LiangFull Text:PDF
GTID:2232330374975829Subject:Power system and its automation
Abstract/Summary:PDF Full Text Request
The smart grid researches rise all over the world. As the intermittent wind power andphotovoltaic energy have make increasingly strong randomness to the system environment,the micro-grid requires the generation dispatch control system transfer from the traditional"automatic control" to "intelligent control" rapidly. A new Smart Generation Control is able tomaintain good frequency performance and quality in a dynamic random environment by thesecondary frequency control.This paper studies the characteristics of a variety of distributed generations, andestablishes a typical Microgrid model with a wind turbine, a photovoltaic generation, flywheelenergy storage, a smallhydro unit, a micro gas turbine and the load model. In the microgirdLFC model, the hydro unit and the micro gas turbine are adopted as AGC units. The AGCcontrol principle for the large interconnected power systems is introduced into the Microgrid,and the multi-step R(λ) learning is applied to propose a novel reinforcement learning basedAGC controller so as to achieve the smart generation control for Microgrids islandedoperation. Simulation analysis and comparison with PI control, Q-learning and Q(λ) learningshow that the R(λ) learning controller has rapid convergence rate and good dynamicperformance as well as strong model adaptability.By using different algorithms to AGC controller simulation, analysis from the simulationresults show that, R(λ) of the controller has a faster convergence characteristic and gooddynamic performance and a strong model adaptation comparing with the PI control,Q-learning and Q(λ) learning.The distributed AGC system is also a feasible micro-network control strategy, which isdifferent from the traditional centralized AGC system. The frequency control units are able tocomplete frequency modulation by establishment of the coordinate mechanism and thecontrol objectives. Simulation analysis show that, the appropriate size of weight coefficientensures normal operation of the microgrid and maintenances the coordination objectives.
Keywords/Search Tags:Reinforcement learning, Smart generation control, Microgrid, Distributedgeneration, Load frequency control
PDF Full Text Request
Related items