An ANN Based Approach For Optimal Placement Of Custom Power Devices
Power electronics based controllers used in distribution systems are called custom power devices. These devices have been proved to be quite effective in power quality enhancement. They may be of series, shunt and series-shunt or series-series type – depending upon their connection in the circuit...
- D K Tanti, MK Verma,
Brijesh Singh, O N Mehrotra
Harmonics have existed in power systems from the time of the very first generators. However, the harmonic components were so small that their effects on systems were negligible. Since last five decades, there has been proliferation in the use of non-linear loads such as electronic ballasts, computer power supplies, fax machines, arc furnaces and Variable Frequency Drives (VFDs). It has resulted in injection of large harmonics contents in the system. The presence of large harmonics contents in the system may lead to problems such as inductive interference with communication lines, degradation of conductors & insulating material in motors and transformers. Therefore it is important to gauge the total effect of harmonics in the system and recommend stringent control measures for reduction of harmonics to a significantly low level.
With the advancement in power electronics, new controllers known as Flexible AC Transmission System (FACTS) have been developed. These controllers have been proved to be quite effective in power flow control, reactive power compensation and enhancement of stability margin in AC networks. Power electronics based controllers used in distribution systems are called custom power devices. Custom power devices have been proved to be quite effective in power quality enhancement. The custom power devices may be series, shunt, and series-shunt or series-series type depending upon their connection in the circuit. Most prominent custom power devices include Distribution Static Compensator (DSTATCOM), Dynamic Voltage Restorer (DVR) and Unified Power Quality Conditioner (UPQC). Custom power devices may be quite helpful in reduction of harmonics. Research on placement of custom power devices has mainly considered harmonics reduction in radial distribution systems. Custom power devices are to be optimally placed in interconnected distribution systems due to their high cost. No effort seems to be made in optimal placement of custom power devices in interconnected distribution system for reduction of harmonics.
In this article, an Artificial Neural Network (ANN) based approach has been proposed for optimal placement of DSTATCOM, DVR and UPQC to reduce harmonics in an interconnected distribution system. The ANN has been trained with Levenberg Marquardth back propagation algorithm (trainlm). Case studies have been performed on IEEE 14-bus system considering it as an interconnected distribution system.
Custom power devices model
In the present work three types of custom power devices have been considered. The proposed models of these devices are presented below.
In the present work, the DSTATCOM has been represented as three independently controllable single phase current sources injecting reactive current in the three phases at the point of coupling (the load bus at which DSTATCOM is placed). The DSTATCOM model has been shown in Fig. 1.
The control scheme consists of three control switches, which can be set on/off as per compensation requirement. The amount of reactive compensation provided by DSTATCOM can be adjusted to reduce THD level at different buses. The three switches remain open during off condition and are closed when compensation is required. This permits injection of independently controllable reactive currents, to the three phases of DSTATCOM bus, which causes reduction of Total Harmonic Distortion (THD) at different buses.
In the present work, the DVR has been represented as three independently controllable single phase voltage sources injecting complex voltages in series with the line in the three phases. The magnitude and angle of injected voltages may be controlled to reduce THD level at different buses. The proposed DVR model has been shown in Fig. 2.
The control scheme consists of six control switches that can be set on/off as per compensation requirement. During off condition, the three control switches connected in series with the controllable single phase voltage sources are open and the other three control switches in parallel with controllable voltage sources, are closed. When compensation is required, the three switches connected in series with independently controllable voltage sources are closed, and the remaining three switches are made open. This permits injection of controllable complex voltages in the three phases of the line, which causes reduction of THD level at different buses.
In the present work, UPQC has been considered as combination of DSTATCOM and DVR models suggested in sections 2.1 and 2.2 respectively.
The simulation model of the power system network under study is developed using MATLAB/SIMULINK software. Harmonics are injected to different bus voltages using three phase harmonics voltage injector shown in Fig. 3, which is a three phase programmable voltage source capable of injecting a maximum of two harmonics component (say 3rd order & 5th order) to bus voltage. The magnitude, phase angle and phase sequence of harmonics can be adjusted in the programmable voltage source shown. The Total Harmonics Distortion (THD) at different buses were calculated using;
THDi = Total Harmonics Distortion (THD) of voltage at bus-i
Vij = jth harmonic component of voltage at bus-i
Vi1= Fundamental frequency component of voltage at bus-i
n = Maximum order harmonics present at bus-i
In the present work, only third harmonic has been injected to bus voltages considering it a major harmonic component. The total harmonic distortions so obtained for different buses have been used to train a feed forward neural network with back-propagation algorithm. The training process is carried out with large number of input data and output target data. The normal acceptable voltage THD value (taken as 5% in this work, considering it a significantly low value) at different buses have been considered as output target data. The Mean Square Error (average squared deviation of bus THD value from the defined target of 5%) is calculated for all the buses. The bus having highest Mean Square Error (MSE) is considered as the optimal bus for the placement of DSTATCOM. Total harmonics distortions were calculated by putting DVR in lines connected to the optimal bus considering one line at a time for DVR placement. The line resulting in maximum reduction of THD after DVR placement was considered as the optimal location for the placement of DVR. The UPQC placement was considered in the optimal line towards optimal bus.
The simulation model of IEEE 14-bus system was developed using MATLAB/SIMULINK software. The IEEE 14-bus system consists of 14 buses including 5 generator buses and 9 load buses, and 20 lines. The total real and reactive power demand of the system are 259 MW and 81.3MVAR, respectively. The simulation block diagram of the system has been shown in Fig. 4.
This plant model has been used to find THD at different buses by switching of harmonics source shown in Fig. 3.
For introducing harmonics at different buses the order of harmonics (n), amplitude (p.u.), phase (degrees) and sequence (0, 1, 2) are to be specified. Here, in present work these are taken as (3, 0.4, 30, 1) to inject 3rd order harmonics with 40% amplitude, 30 degree phase shift and positive phase sequence. The THD database of different buses was prepared to train the artificial neural network. The normal acceptable THD values of different buses (taken as 5% in this work) were considered as output target data. Some data were used to test the network and Mean Square Errors (M.S.E.) was calculated for all the buses. The ANN training performance has been shown in Fig. 5
It is observed from Fig. 5 that bus-5 has the highest value of Mean Square Error. Hence, bus-5 was considered as the optimal location for the placement of DSTATCOM. DVR placement was considered in all the lines connected to bus-5 viz. Line 5-1, Line 5-2, Line 5-4 and Line 5-6, respectively. The placement of DVR in Line 5-4 is found more effective in THD reduction as compared to DVR placement in Line 5-1, Line 5-2 and Line 5-6.
Therefore, Line 5-4 was selected as the optimal line for the placement of DVR controller. UPQC placement was considered in optimal line 5-4 towards optimal bus 5. In order to study impact of DSTATCOM, DVR and UPQC in reduction of THD, the DSTATCOM model presented in Fig.1, DVR model presented in Fig. 2 and UPQC model were considered and their SIMULINK models were developed. Total Harmonic Distortion (THD) for voltage at different buses were obtained without any controller, with placement of DSTATCOM at bus-5, with placement of DVR in line 5-4 and with placement of UPQC in line 5-4 towards bus-5, using the software package MATLAB/SIMULINK.
Total Harmonic Distortion for voltages at few selected buses without controller, with DSTATCOM placed at bus-5, with DVR placed in Line 5-4 and with UPQC placed in Line 5-4 towards bus-5, have been shown in Fig. 6.
It is observed from Fig. 6 that placement of DSTATCOM, DVR and UPQC results in significant reduction of voltage THD at each and every bus of the considered system. However, UPQC is more effective in THD reduction compared to other two controllers.
In this article, an ANN based approach has been considered for optimal placement of DSTATCOM, DVR and UPQC to reduce THD level in an interconnected distribution system. Case studies have been performed on IEEE 14-bus system with the help of MATLAB/SIMULINK software. The time domain simulations for voltage THD at different buses have been obtained without any controller, with DSTATCOM controller, with DVR controller and with UPQC controller placed optimally in the system.
The simulation results obtained on the test system establish effectiveness of placement of DSTATCOM, DVR and UPQC in reduction of THD level at different buses. The placement of UPQC seems to be more effective than the placement of DSTATCOM and DVR in reduction of Total Harmonic Distortion (THD) in meshed interconnected power system.
D K Tanti, MK Verma, Brijesh Singh, O N Mehrotra are from Bihar Institute of Technology, Sindri and Indian Institute of Technology, Varanasi.
What are your views on ANN based approach considered for optimal placement of custom power devices? Comment your answer now!