- - z ETEP zy Load Frequency Control for Power System with Reheat Steam Turbine and Governor Deadband Non-linearity by Using Neural Network Controller H.L. Zeynelgil, A. Demiroren, zyxwv N.S. Sengor Abstract zyxwvutsrqp In this puper, a neural network zyxwvutsr (NN) controller is presented for load-frequency control of power system. The NN controller uses back zyxwvut propagation-through-time algorithm. In the power system, the reheat effect of the steam tur- bine and the effect of governor deadband non-linearity are considered by describing function approach in the state space model. By comparing the results of simulations, the peflormance of the NN controller is better than conventional controller. NN controller gives a shorter settling time and eliminates the necessity of parameter estimation time required in conventional adaptive control techniques. 1 Introduction zyxwvuts Changes in load are accompanied by changes in system frequency and generation. The system fre- quency must be maintained within given limits. Many control strategies have been proposed in liter- ature [ 1-31 to obtain better performance. Fixed con- troller, which is optimal under one operating condi- tion, may no longer be suitable in another one. In view of this, the variable structure control has to be applied to make the controller sensitive to the plant parameter changes [4-61. On the other hand, various adaptive control techniques have been proposed deal- ing with large parameter variations [7, 81. However, this method needs the information of the system states, which is not completely known generally. Be- cause of the inherent non-linearity of power systems, neural network techniques can be considered to build non-linear, adaptive controllers with improved per- formances. In this paper, a simple, isolated generator unit connected to a power line or electric bus that operates for different consumers is considered. zyxwvu As load varies, the frequency of the generator unit varies. To return to the steady-state value of frequency after a given load variation, a control system is designed that acts on the setting of the steam admission valve of the unit tur- bine. Frequency transients must be eliminated as rapidly as possible. It is known that most load-frequency con- trol systems include an integral controller. The integra- tor gain is set to a level that compromise between fast transient recovery and low overshoot in dynamic re- sponse of the system [9]. Unfortunately, this type of controller is slow and impossible in account to generat- ing unit non-linearities. The NN load-frequency controller is proposed in this study. It is shown that the power system controlled by using neural network back propagation-through- time algorithm can give good dynamic response. The model of non-linear system to be controlled is given by a set of differential equations, i.e., the system is a continuous dynamical system modelled by state space equations. So the control rule to be imposed must cope with this dynamic. Since NN will be used to con- trol the system, back propagation-through-time algo- rithm is preferred to cope with the continuous time dynamics. 2 Plant model and conventional control Mechanical power is supplied by a turbine and it is given to a synchronous generator for different consum- ers at a single area system. Essentially, in practice, “area” refers to a system including many parallel work- ing generators [9]. The waveforms of electrical quanti- ties at the output of the generator are mainly deter- mined by the turbine steam flow. It is also affected by changes in user power demands [ 101. When the electri- cal load suddenly increases, the generator shaft slows down, and the frequency of the generator decreases. The control system must detect the load variation and command the steam admission valve to open more so that the turbine increases the mechanical power pro- duction, counteracts the load increase, and brings the shaft speed and hence the generator frequency to their nominal values. Fig. 1 shows schematically a real-power control mechanism [9]. By controlling the position of the governor controlled valves, measured by the coordi- nate xE, control over the flow of steam through the tur- ETEP Vol. 12, No. 3, MayIJune 2002 179