1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2823062, IEEE Transactions on Sustainable Energy 1 Abstract— Large scale power systems have distributed structure since power generation units are distributed in the network. In such systems, where the communications are sometimes imperfect, centralized control structures have concerns to be applied. For such inherently distributed systems, multi-agent control structures are more appropriate. On the other hand, power systems have nonlinear dynamics. When all generation units are synchronous generators, linear approximation of swing equation of generators is common choice. However, by penetration of renewable energy generation units, those have highly nonlinear dynamics, nonlinear control approaches should be applied. In this paper nonlinear multi-agent feedback linearization approach is adopted for optimal load frequency control of a power system with wind generation units. Multi-agent controllers with both distributed and decentralized structures are proposed. Centralized feedback controller is also applied for comparison purposes. The parameters of the controllers are optimized using GWO algorithm to improve time response of the system and reduce generation costs. IEEE 30-bus system with 6 synchronous generators and 4 wind turbines is considered as test system. Simulation results verify inefficiency of centralized controller and show that distributed control structure can produce a balanced performance in terms of time response and is better from the cost reduction view point than the decentralized controller. Index Terms— Distributed control, Decentralized control, Multi- agent systems, Optimal load frequency control, Feedback linearization, Economic load dispatch I. INTRODUCTION UTOMATIC generation control (AGC) is a wide area concept in power systems that consists of load frequency control (LFC), economic dispatch, interchange transaction scheduling, reverse monitoring and related data recording [1]. Among these function, LFC is the fundamental task and the most important issue. LFC prevents the frequency deviations to achieve zero steady-state errors; meanwhile optimal transient behavior is one of the most important objectives of LFC in the multi-regional integrated power systems [1][2]. Typically, the control loops at lower levels of a power system (locally in a generator) has smaller time constants in the order of a second or less. Whereas, the LFC loop operates Manuscrpit received 05 July, 2017; The authors are with the Electrical Engineering Department, Faculty of Technical Engineering, University of Mohaghegh Ardabili, Iran. (e-mails: akbarimajd@uma.ac.ir, mohsenolyai@gmail.com, b.sobhani@gmail.com, hshayeghi@gmail.com ). in a timescale of several seconds or minutes. Fig. 1 shows the schematic diagram of different timescales of power system dynamics and controls [3]. Accordingly, the excitation system time constant is much smaller and it does not affect the LFC dynamic, i.e. the LFC control loop can be considered decoupled from excitation system and is studied individually. In large scale power networks where power generation units are distributed in the network and the communications are limited or unreliable, employing of centralized control structures encounter some serious challenges [4]. Delays in communication channels, limited bandwidth and network cyber-attackers are some of these challenges. In addition, in recent years, rapid growth in utilization of renewable energy recourses, those are operated as distributed generation units, has created new challenges in power grids. Namely, uncertainties and nonlinearities have been increased and management, control and decision making have become more distributed. In large scale power grids, when distributed generation units participate in the grid, new control structures including topologies, algorithms and optimization methods should be developed. Among distributed generations, wind turbines are more uncertain and have more nonlinear dynamics. In the wind turbine generators, the output power is changed by wind shift, and the shift causes the frequency change in the power system [5]. In [6], the effect of wind power generation on the frequency responses of a power generation system has been investigated. In [7] small signal stability of power system in presence of wind power system with full-load converter interfaced wind turbines was studied. As it was mentioned, one of the challenges in modern large scale power systems is dealing with the high dimensionality of the system. To reduce the size and complexity of LFC problem, centralized control approach is repudiated and distributed control or decentralized control structures are applied [8][9]. These structures provide simple control applied Fig. 1. Schematic timescale schematic diagram of different timescales of some dynamics and controls loops of a power system. Nonlinear Multi-agent Optimal Load Frequency Control based on Feedback Linearization of Wind Turbines A. Akbarimajd, Member, IEEE, M. Olyaee, B. Sobhani, H. Shayeghi, Senior Member, IEEE A time miliseconds seconds minutes Fast protective devices Slow protective devices FACTS, AVR, PSS Generator/Excitation dynamics LFC Voltage stability Frequency stability