ADAPTIVE AND NON-LINEAR EXCITATION CONTROL OF SYNCHRONOUS GENERATOR‟S STABILITY THROUGH NEURAL NETWORK

ABSTRACT
Adaptive and non linear Excitation control of synchronous generator‟s stability through Neural Network Control equipment of synchronous generators such as automatic voltage regulators, speed governors and power system stabilizers have been developed to maintain stability and to improve damping of the power systems. When an operating condition changes greatly, however, such controllers may become less effective because of nonlinearity of the power system. And hence these drastic changes in the power system caused by faults and circuit switching may cause control performance to become unsatisfactory. In this project, a nonlinear adaptive generator control system using neural networks is proposed. In this controller we have integrated a voltage regulator and a power system stabilizer. The proposed neural network based controller generates appropriate control signals enhancing transient stability and damping of the power system. The proposed system is demonstrated by computer simulation in MATLAB by first modeling a power system with conventional controller (AVR & PSS) and then these controllers are used to TRAIN the neural network controller. After training the neural network based controller is used to control the same power system and it is proved through simulation that the neural network controller performs better than the conventional controller, improving the transient and dynamic stability.

EXECUTVE SUMMARY

HYPOTHESIS: Adaptive and non linear Excitation control of synchronous generator‟s stability through Neural Network BACKGROUND: Both the historical and the present-day civilization of mankind are closely interwoven with energy, and there is no reason to doubt but that in the future our existence will be more and more dependent upon the energy. Electrical energy occupies the top position in the energy hierarchy. Therefore, it is more favorable to make the generation and transmission of electrical energy more economical and reliable. Keeping in mind this economic condition 3φ synchronous generators (known as alternators) are used for large scale power generation. Here the armature winding is placed on the stator while the field winding is placed on the rotor. The field winding is responsible for excitation control of the generator which maintains generator voltage and controls the reactive power flow [01]. Most synchronous generators are connected to large interconnected power system and hence work on an infinite bus. The control of active and reactive power keeps the system in steady state. Changes in real power affect mainly the system frequency while the reactive power is mainly dependent on voltage magnitude. In synchronous machine this real power is controlled by governor action i.e. by controlling the input mechanical power. The reactive power and hence the terminal voltage is controlled by Voltage Regulator i.e. by controlling the excitation voltage. Hence the controller for voltage regulator holds an important position in determining the power system stability. Today‟s automatic control theory is all based on the concept of feedback. The essence of feedback theory consists of three components, measurement, comparison and correction. In order for the controller to perform its best under all operating conditions it must be capable of having good learning and adaptation capabilities to cope with changes and uncertainties in the system.
EXECUTVE SUMMARY
2
A basic approach to controller design for synchronous machines is an implementation of state feedback optimal control. It is typically designed for a linear model about a specific operating point, which does not necessarily guarantee sufficient robustness to handle changes in system power loads and variations due to system parameter uncertainties. Due to its limitations, this approach has lost its original popularity. Subsequently, adaptive control was developed over the past decade. Most algorithms are still based on a linear model. However, the synchronous machine is a nonlinear, fast-acting multivariable system and interconnected in a power system. The machine operates over a wide range of operating conditions, and is subject to different types of disturbance. The conditions change, but the outputs have to be coordinated so as to satisfy the requirements of power system operation. For this type of system it is recognized that classical control theory and mathematical model-based control algorithms can not be successfully employed. [02] To overcome the above problems, a new approach to controller design which uses new technologies such as artificial neural networks is needed. And therefore there has been some research on using the neural network approach for nonlinear systems control. This project presents an application of artificial neural network as a controller for a synchronous machine excitation system. A hierarchical architecture of an ANN is adopted for controller design, which is used for data mapping and control respectively, based on the Back Propagation Algorithm (BPA). An artificial neural network (ANN), usually called "neural network" (NN), are applied in this work because they are remarkable on several counts. First, they are adaptive: they can take data and learn from it. Thus they infer solution from the data presented to them, often carrying quiet subtle relationship. This ability differs radically from standard software techniques because it doesn‟t depend on the programmer prior knowledge of rules. Second, NN can generalize: they can correctly process data that only broadly resembles the data they were trained on originally. They can also solve problems that lack existing solutions. Third NN are non linear, in that they can capture complex interaction among an input variable in a system. These are some reasons why NN is used in this project.

PROPOSED SCHEME:
This thesis provides a means of determining the application of neural network in the excitation systems of synchronous generator. The responses of synchronous machine in a power system are observed by computerized simulation. In fig.1.1 the block diagram representation of the power system model is shown, which is a simple representation of a general power system model. It is highlighted here that in the proposed scheme the conventional based controller of AVR and PSS are replaced by neural network based controller.
Before the neural network based controller can be employed in the system they must be trained [chapter 04]. For the training of network we need the training data i.e. telling the neural network the inputs and the desired outputs so that it can adjust itself in such a way that the next time when it is provided with such input data it generates the desired output. Before training the network we have to decide two important things: firstly its architecture and then the training Algorithm [chapter 02].
[attachment=13991]