14-08-2012, 09:50 AM
Performance of SMES unit on Artificial Neural Network based Multi-area AGC scheme
[attachment=31356]
Abstract
This work investigates the performance of Superconducting Magnetic Energy Storage (SMES) unit on Artificial Neural Network (ANN) based multi-area AGC scheme. SMES units have been used to the power systems to inject or absorb active power. A three layer feed forward neural network (NN) is proposed for controller design and trained with Back propagation algorithm (BPA). The poolco based transaction can be implemented by optimizing the bids (price & capacity) submitted by the generating companies (Gencos) and distribution companies (Discos). The functioning of the proposed ANN based controller has been demonstrated on a two-area System.
INTRODUCTION
The parallel operation of interconnected systems is the today’s requirement with the increase of size of electric power system, controlling the frequency of interconnected power system has becoming the challenge for control engineer. The deviation of the frequencies and tie-line power arise because of unpredictable load variations, which occur due to a mismatch between the generated and the demanded power. The main objective of providing an Automatic Generation Control (AGC) has been to maintain the system frequency at nominal value and the power interchange between different areas at their scheduled values.
CONTROLLER DESIGN USING NEURAL NETWORK
The conventional PID controller is replaced by artificial neural network (ANN) trained controller to improve the dynamic response during the step load change of Discos for multi area AGC scheme.
Each area is equipped by a neural controller as shown in figure 5. The best value for controller parameters is obtained by training the ANN off line at different load parameters through Back propagation
CONCLUSION
A general-purpose ANN controller for multi area AGC, suitable for deregulated electricity market, has been developed with SMES and without SMES unit. The investigation shows that for the mixed transactions, the response is faster and less undershoots with SMES unit compared to without SMES unit. Effort has been made in this paper to reduce the cost incurred by earlier proposed systems by having SMES unit located only in one area to regulate multi-area frequency. The proposed ANN controller has been successfully tested on a 39-bus New England power system for all types of load following contracts. It has been shown that the system frequency and tie-line power oscillations can be effectively damped out with the use of a small capacity SMES unit in either of the areas following a step load disturbance. It has also been observed that the use of ACE for the control of SMES unit substantially reduces the peak deviations of frequencies and tie-power responses.
[attachment=31356]
Abstract
This work investigates the performance of Superconducting Magnetic Energy Storage (SMES) unit on Artificial Neural Network (ANN) based multi-area AGC scheme. SMES units have been used to the power systems to inject or absorb active power. A three layer feed forward neural network (NN) is proposed for controller design and trained with Back propagation algorithm (BPA). The poolco based transaction can be implemented by optimizing the bids (price & capacity) submitted by the generating companies (Gencos) and distribution companies (Discos). The functioning of the proposed ANN based controller has been demonstrated on a two-area System.
INTRODUCTION
The parallel operation of interconnected systems is the today’s requirement with the increase of size of electric power system, controlling the frequency of interconnected power system has becoming the challenge for control engineer. The deviation of the frequencies and tie-line power arise because of unpredictable load variations, which occur due to a mismatch between the generated and the demanded power. The main objective of providing an Automatic Generation Control (AGC) has been to maintain the system frequency at nominal value and the power interchange between different areas at their scheduled values.
CONTROLLER DESIGN USING NEURAL NETWORK
The conventional PID controller is replaced by artificial neural network (ANN) trained controller to improve the dynamic response during the step load change of Discos for multi area AGC scheme.
Each area is equipped by a neural controller as shown in figure 5. The best value for controller parameters is obtained by training the ANN off line at different load parameters through Back propagation
CONCLUSION
A general-purpose ANN controller for multi area AGC, suitable for deregulated electricity market, has been developed with SMES and without SMES unit. The investigation shows that for the mixed transactions, the response is faster and less undershoots with SMES unit compared to without SMES unit. Effort has been made in this paper to reduce the cost incurred by earlier proposed systems by having SMES unit located only in one area to regulate multi-area frequency. The proposed ANN controller has been successfully tested on a 39-bus New England power system for all types of load following contracts. It has been shown that the system frequency and tie-line power oscillations can be effectively damped out with the use of a small capacity SMES unit in either of the areas following a step load disturbance. It has also been observed that the use of ACE for the control of SMES unit substantially reduces the peak deviations of frequencies and tie-power responses.