The Application Of Intelligent Control In AGC Control Of Power System

The power system is a complex structure, but at the same time strongly nonlinear, modeluncertainty and time-varying typical high-dimensional dynamic system. In recent years, asintelligent control algorithm is proposed and the smart grid concept, more and moreintelligent control method is applied to power system control. In the area of automaticgeneration control (AGC) in the power system, NERC officially launched the tie-line powersystem frequency deviation mode automatic generation control of interconnected powersystem control performance standards (CPS) in1997. Subsequently, the standards of the CPS,a large number of researchers have proposed various AGC control strategy, such as theclassical control strategy, adaptive control strategy, and modern intelligent control strategy.Variable universe fuzzy controller is the improvement of the fuzzy controller, it inherits thesimple structure of the fuzzy controller, robustness, and no need to control the accuratemathematical model, to overcome the fuzzy controller steady-state control accuracy is nothigh, subjective factor and other shortcomings. Reinforcement learning is an importantartificial intelligence control method, it has the advantages of self-learning and dynamicstochastic optimization, and adaptation of the power system the actual running state with goodresults, therefore, in recent years, reinforcement learning method also began to study appliedto power system Automatic Generation Control. This paper is focused on parametersoptimization variable universe fuzzy control and multi-agent reinforcement learning in theAGC control of the power system to commence; its main contents are as follows:First, this paper briefly introduces the AGC control structure based on the CPS standard.This paper analyzes the application in this model, the variable universe fuzzy control and itscorresponding fuzzy rules and membership functions are designed for both the AGCrelaxation control features. Since then the variable universe fuzzy controller based on thegeometric factor, while the introduction of self-optimization method for parameteroptimization, the formation parameters from the optimization of variable universe fuzzycontroller, and the final application in the AGC control of the power system effect.Secondly, in the fourth chapter describes the equilibrium theory and based on theequilibrium theory of multi-agent reinforcement learning. Describes the calculation of theequilibrium theory solution based on MATLAB S-function algorithm to achieve, given thesolution and calculation procedures for the Nash equilibrium when the two-agent game with the equilibrium solution.Finally, in Chapter IV, the main use of the CEQ, this is easier to implement multi-agentreinforcement learning method in the AGC control of the power system simulation run, theinformation on the interaction between the agents to obtain better results, but also forshortcomings on the outlook for the future research work.The thesis is supported by the National Natural Science Foundation of China (50807016,51177051), the Fundamental Research Funds for the Central Universities （2012ZZ0020），State Key Laboratory project of Tsinghua University （SKLD10KM01）..