تجزیه و تحلیل عملکرد مقایسه ای مصنوعی الگوریتم کلونی زنبور عسل برای تنظیم کننده اتوماتیک ولتاژ (AVR)

In this study, Artificial Bee Colony (ABC) algorithm is applied to the Automatic Voltage Regulator (AVR) system for obtaining optimal control. The tuning performance of this algorithm and its contribution to the robustness of the control system are also extensively and comparatively investigated. In the performance analysis, Particle Swarm Optimization (PSO) algorithm and Differential Evolution (DE) algorithm are used for the purpose of comparison. These analyses are realized by benefiting from different analysis methods such as transient response analysis, root locus analysis, bode analysis and statistically Receiver Operating Characteristic (ROC) analysis. Afterwards, the robustness analysis is applied to the AVR system, which is tuned by ABC algorithm in order to determine its response to changes in the system parameters. At the end of the study, it is shown that the ABC algorithm is successfully applied to the AVR system for improving the performance of the controller and shows a better tuning capability than the other similar population based optimization algorithms for this control application.

Providing constancy and stability of the nominal voltage level in an electric power network is also one of the main control problems for an electric power system, because all equipments that are connected with this power network have been designed for a certain voltage level called rated or nameplate voltage. If the nominal voltage level deviates from that value, the performance of these equipments decreases and their life expectancy drops. In addition to this, the other important reason for this control is that the real line losses depend on real and reactive power flow. In fact, the reactive power flow depends greatly on terminal voltages in the power system. However it is possible to minimize the real line losses by controlling the nominal voltage level. To solve these control problems, which are explained above, an Automatic Voltage Regulator (AVR) system is applied to power generation units [1]. The AVR system is a closed loop control system that provides terminal voltage at the desired value. The configuration of this control system will be investigated in the next section. In the related literature, to realize the AVR system with better dynamic response, a number of different control strategies such as optimal, adaptive, robust control, etc. have been reported by researchers so far. But the self-tuning adaptive control technique is distinguished from the other control techniques because it makes the process, which is under control, less sensitive to changes in process parameters, and in particular it is also simpler to implement than the other modern control techniques. For this purpose, this type of control is applied to the AVR system in this study. Previous works related to the AVR system, which uses the self-tuning methods, initiated in the years of the 1990s. For example, Swidenbank et al. [2] applied the classical self-tuning control techniques to the AVR system in 1999. After this study, Fitch et al. [3] used a generalized predictive control technique as a self-tuning control algorithm in the same year. Since the conventional self-tuning control techniques containing more mathematical computing may also be unsuitable in some operating conditions due to the complexity of the power system such as nonlinear load characteristics and variable operating points, the usage of artificial intelligence based self-tuning control and optimization techniques was preferred by researchers from the beginning of 2000. For example, Panda and Padhy [4] proposed PSO based optimal design method for STATCOM-based controller with multiple PSS, and they tested the stability of their design in two area power system. After three years, they used the improved genetic algorithm method in order to solve the optimal design problem of flexible AC transmission system (FACTS)-based controller for the power systems [5]. However, self-tuning PID controllers tuned by these optimization methods have also been initiated to be applied to the AVR system frequently in these years. Gaing [6] suggested a PSO based self-tuning PID controller for the AVR system, and he compared the result of his method with that of the genetic algorithm based method in 2004. Two years after, Kim and Cho [7] developed the hybrid method, which contained genetic algorithm and bacterial foraging optimization technique, in order to improve the performance of the self-tuning PID controller in the AVR system. In 2007, Mukherjee and Ghoshal [8] reported the Sugeno fuzzy logic self-tuning algorithm based on crazy-PSO for PID controller, and proposed a novel cost function in this optimization method. They also compared their results with the genetic algorithm based controller. Later on, Zhu et al. [9] suggested a chaotic ant swarm algorithm in order to optimize the gains of PID controller in the AVR system in the year of 2009. In the same year, Zamani et al. [10] designed the particle swarm optimization based fractional order PID controller for the AVR system. They investigated the basic performance and robustness of their controller and compared with that of the classical PID controller. Coelho [11] proposed the chaotic optimization approach for tuning of the PID gains in 2009. Chatterjee et al. [12] also made a comparison between the optimization performance of velocity relaxed and craziness based particle swarm optimization methods on AVR system in 2009. In this study, it is evaluated that the ABC algorithm may be used as an alternative tuning method due to its superior local and global search capability provided by separate artificial bee colonies such as employees, onlookers and scouts [13] and [14]. This algorithm was successfully applied first to the different optimization and control processes so far. Some of them are the neural networks training [15], the design of optimum IIR filter [16], the quadratic knapsack problem [17], the parameter extraction of MESFETs [18], the economic dispatch problem for power system [19], clustering [20], the prediction of protein tertiary structure [21] and the load-frequency control for interconnected power system [22]. The aim of this study, which is different from the above literature, is that the ABC algorithm is applied to the AVR system in order to optimize the control parameters of the PID controller, and its tuning performance for the application of optimal AVR system, which is determined comparatively using PSO and DE algorithms. In this way, the optimal voltage control of the power system is provided. All analyses are realized by benefiting from different analysis methods such as the transient response analysis, the root locus analysis, the Bode analysis and the ROC analysis for investigating the results from another point of view. In addition to these aims, the robustness analysis is also applied to the AVR system, which is tuned by the ABC algorithm, in order to determine the contribution of this algorithm to the robustness of the control system.

This study presents both the usage of ABC algorithm as the new artificial intelligence based optimization technique in order to optimize the control problem of AVR system and the comparative tuning performance analysis of this algorithm. To reveal the tuning performance of ABC algorithm on AVR system, artificial intelligence based PSO and DE algorithms are applied to the control system for the purpose of comparison, additionally. The transient response analysis, the root locus analysis and the Bode analysis are used in order to obtain the results. In another point of view for analyzing them, statistical ROC analysis method is also applied to the results. Hereby it is seen that the tuning superiority of ABC algorithm is proved by all analysis techniques. It is evaluated that this result is caused by the triple search capability of the ABC algorithm. As explained in the related sections, the local search is realized by the employed and the onlooker bee phases and global search is realized by the scout bee phase in ABC algorithm consecutively and separately. In contrast, the local and global searches are realized by the same update formula in PSO and DE algorithms. On the other hand, the robustness of this algorithm is also shown by the applied method. Accordingly, the AVR system that was optimized by ABC algorithm is affected in adequately small amount by changes of the process parameters in the range of ±50%. Consequently it is seen from this study that ABC algorithm can be applied to the AVR system successfully, and it allows to control this system optimally and robustly. In addition to this, it can be evaluated that this control technique obtained using ABC algorithm is also applied to different control applications.