Some decentralized optimization and control algorithms for the control of wind farms

Decentralized control;Network control;Wind farm control

Cheng, Albert Z. ; Langbort ; Cedric

关键词: Decentralized control;    Network control;    Wind farm control;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/24125/Cheng_Albert.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This is a preliminary study of decentralized algorithms that can be applied to windfarm controls. Traditionally, wind farm control is comprised of the wind farm levelcontrol and the wind turbine level control. The wind farm level control is a centralizedcontroller that takes the demands from the grid and generates operating points for eachwind turbine within the wind farm. The wind turbine level control then generatesthe optimal control for each turbine to match the operating point. Unfortunately, thistraditional control scheme does not constitute the optimal operation of a wind farm dueto it’s disregard at either level of control for the interactions between wind turbines inthe wind farm .Consequently, a different two level control scheme is proposed in this thesis. Thiscontrol scheme is shown to be a decentralized controller in that each wind turbine hasthe ability to both generate its own operating point and calculate its own optimal control.Through the communication of the wind turbines with each other, the interactionsbetween the wind turbines are incorporated into both levels of control. The generationof the operating point is posed as a stochastic resource allocation problem that takesinto account the stochastic wind and other wind farm characteristics. We develop astochastic algorithm based on network dynamic system theory to solve the resource allocationproblem. We show that the algorithm converges to the solution of the resourceallocation problem almost surely. The calculation of each wind turbine’s optimal controlis formulated as an Linear Quadratic Regulator (LQR) optimization problem witha equality constraint. We develop an algorithm that is based on the Tatonnement processin Economics to solve the LQR problem. We first consider the performance of thealgorithm in a dynamically decoupled system and show that the algorithm solves theLQR problem. We then consider the performance of the algorithm in a dynamicallycoupled system and discuss the difference between the two cases.