Design of Decentralized Neuron Based LFC in a Deregulated Power System HEIDAR ALI SHAYANFAR HOSSEIN SHAYEGHI Electrical Engineering Department Iran University of Science and Technology Technical Engineering Department University of Mohaghegh Ardebili Tehran, IRAN Ardebil, IRAN Abstract: - This paper presents a new decentralized Artificial Neural Network (ANN) controller based on the mixed H 2 / H ∞ control technique for Load Frequency Control (LFC) in a deregulated power system. To achieve decentralization, the effects of possible contracted scenarios and interfaces between control areas are treated as a set of new input disturba- nce signals. In order to account modeling uncertainties, cover practical constraints on control action and minimize the effe- cts of area load disturbances, the idea of mixed H 2 / H ∞ control technique is being used for training ANN based controller. This newly developed design strategy combines advantage of the ANN and mixed H 2 /H ∞ control techniques for improve- ing robust performance and leads to a flexible controller with simple structure, which can be useful in real world complex power systems. The proposed method is tested on a two-area power system to demonstrate its robust performance with possible contracted scenarios under large load demands and modeling uncertainties. The results of proposed controller are compared with mixed H 2 / H ∞ and PI controllers in the presence of Generation Rate Constraints (GRC). Key-Words: LFC, Decentralized Control, Deregulated Power System, ANN, Mixed H 2 / H ∞ control, Robust Control. 1 Introduction In the restructured power system, Load Frequency Con- trol (LFC) will serve as ancillary service and acquires a principal role to enable power exchanges and to provide better condition for electricity trading [1]. In an open energy market a DISCO has the freedom to have a contr- act with any GENCO for power transaction in its area or other areas. Currently, all transactions have to be cleared through Independent System Operator (ISO) or other responsible organizations. In a real world deregulated power system, each control area contains different kinds of uncertainties and various disturbances due to increasing the complexity, system mo- deling errors and changing power system structure. As a result, a fixed controller based on classical theory is certa- inly not suitable for LFC problem. Thus, it is required that a flexible controller is developed. Recently, several optim- al and robust control strategies have been developed for LFC synthesis according to change of environment in po- wer system operation under deregulation [2-6]. The pro- posed methods show good dynamical response, but robus- tness in the presence of modeling uncertainties and system nonlinearities were not considered. Also, some authors suggest complex state feedback or high order dynamical controllers, which are not practical for industry practices. In this paper, a new decentralized Artificial Neural Net- work (ANN) controller is developed based on the mixed H 2 / H ∞ control technique for LFC problem in an open ene- rgy market. Following the idea presented in Ref. [7] a generalized model for LFC scheme is developed based on the possible contracted scenarios in deregulated environm- ents. To achieve decentralization, the effects of possible contracted scenarios and interface between areas is treated as a set of new input disturbance signals in each control area. LFC goals, i.e. frequency regulation and tracking the load demands, maintaining the tie-line power interchanges to specified values in the presence of model uncertainties and Generation Rate Constraints (GRC) determines the LFC synthesis as a multi-objective optimization problem. Thus, first the LFC problem is formulated as a multi obje- ctive optimization problem via a mixed H 2 / H ∞ control technique and solved by linear Matrix Inequalities (LMI) approach to obtain optimal controllers. Then, these contro- llers are reconstructed by using learning capability of neu- ral networks to obtain the desired level of robust perform- ance in different operating conditions. The main feature of ANN based controller is that it provides a non-model ba- sed control system and do not require the accurate model of the plant. The proposed strategy is tested on a two-area power system for two scenarios in the presence of model uncertainties and GRC under various load changes. The results show that the proposed method guarantees the rob- ust performance for various operating conditions and sup- erior to the mixed H 2 / H ∞ and conventional PI controllers. 2 Generalized LFC Scheme Model The deregulated power system consists of three company- ies, GENCOs, TRANCOs and DISCOs with an open access policy. In this environment, GENCOs may or may not participate in the LFC task and DISCOs have the lib- erty to contract with any available GENCOs in their own or other areas. This makes various combinations of possi- ble contracted scenarios between DISCOs and GENCOs.