State Feedback Design for Multi-machine Power System Using Genetic Algorithm Abolfazl Jalilvand, Reza Aghmasheh and Amin Safari Department of Electrical Engineering, Faculty of Engineering, Zanjan University, Zanjan, Iran ajalilvand@znu.ac.ir , r.aghmasheh.ir@gmail.com and asafari1650@yahoo.com Abstract-An optimal design of state feedback, as a power system stabilizer (PSSs) is presented based on genetic algorithm (GA) by using eigenvalue-based objective functions. The practical implementation of PSSs and The relative stability of low-frequency modes are all included in the constraints. The locally measured states are fed back at the AVR reference input of each machine after multiplication by suitable feedback gains. The proposed method is confirmed by eigenvalue analysis and simulation results for a multi-machine power system under different operating conditions. I. INTRODUCTION Stability of power systems is one of the most important aspects in electric system operation. This arises from the fact that the power system must maintain frequency and voltage levels, under any disturbance [1]. Since the development of interconnection of large electric power systems, there have been spontaneous system oscillations at very low frequencies in order of 0.2 to 3.0 Hz. Once started, they would continue for a long period of time. To enhance system damping, the generators are equipped with power system stabilizers (PSSs) that provide supplementary feedback stabilizing signals in the excitation systems. PSSs augment the power system stability limit and extend the power-transfer capability by enhancing the system damping of low-frequency oscillations associated with the electromechanical modes [2]. Several approaches based on modern control theory have been applied to the PSS design problem including adaptive [3], robust [4] variable structure control [5]. These stabilizers provide better dynamic performance but they suffer from the major drawback of requiring model parameter identification, state observation and feedback gain calculations. The eigenvalue sensitivity analysis has been used for PSS design under deterministic system operating conditions. It has been used for the parameter design of power system damping controllers [6]. A new approach for the optimal decentralized design of PSSs with output feedback is investigated in [7]. If complete state feedback control scheme are adopted, the requirements of estimators and centralized controls may be used for the unavailable states and control signals. These increase the hardware cost and reduce the reliability of the control system. Novel intelligent control design methods such as fuzzy logic controllers [8-9] and artificial neural network controllers [10] have being used as PSSs. H optimization techniques have been also applied; however the importance and difficulties in the selection of weighting functions of H optimization have been reported. The order of the based stabilizer is as same as the plant. This causes the complex stabilizers and reduces their applicability [11]. Recently, heuristic methods are widely used to solve global optimization problems. Techniques such as genetic algorithms [12], tabu search algorithm [13], simulated annealing [14], evolutionary programming [15] and swarm optimization techniques [16] have been applied earlier to the PSSs design. In this paper the proposed design process for PSSs with state feedback schemes is applied to a multi-machine power system. Local and available states ( δ Δ , ω Δ , q E Δ and fd E Δ ) are used as the inputs of each PSS. The design problem is converted to an optimization problem and GA is employed to solve it. Robustness is achieved by considering several operating conditions. The eigenvalue analysis and the time domain simulation results under various operation conditions are given, which show that the proposed objective functions damps the low frequency oscillation in an efficient manner. II. PROBLEM STATEMENT A. Power System Model A power system can be modeled by a set of nonlinear differential equations as: ( , ) X f XU = (1) In this study, the two-axis model is used for time–domain simulations [2]. When a power system operates at one operating point, the system is linearized. The model of this linearized system can be represented by [10]: () () () () [ () ( )] s d X t AX t BU t A X t BU t U t = + = + + (2) Where the input vector ) (t U is composed of the control vector ) (t U S and disturbance vector ) (t U d , both 1 × m vectors; ) (t X is an 1 × n state vector; A is an n n × plant matrix of the open-loop system and B is an m n × input matrix; n and m are the number of state variables and control signals, respectively. The following control input is defined as the state feedback: () () s U t KX t = (3) Where K is the feedback gain matrix with appropriate dimensions. Applying (3) to (2): () ( ) () () d Xt A BK X t BU t = + + (4) c A A BK = + (5) 978-1-4244-3388-9/09/$25.00 ©2009 IEEE 244