AbstractIn this paper proposed an algorithm to control dynamic voltagerestorer (DVR). The proposed controller is using an artificialneural network with Levenberg Marquardt method ConstructiveNeural Network (CLMNN). The purpose of this control is to obtaina DVR with a quick and accurate response while improving thequality of the voltage due to the voltage flicker in low-voltagesystems. CLMNN use that is based on the configuration of hiddenlayer neural network used can be changed to obtain the optimalvalue. CLMNN used in the detection of the change in voltageconditions, either in the form of fluctuations in the voltageamplitude and phase changes in load. Once detected, CLMNNgenerate a PWM signal as a control. To determine the performanceCLMNN controller, then the simulation is used as a comparison ofconventional controllers. Based on the simulation results it isknown that overshoot voltage of sensitive load decreased by 20%when using CLMNN. In addition, the DVR is more stable with afaster settling time, ie less than 0.01 seconds, whereas when using aPI controller reaches 0.02 seconds.Keywordsâ Dynamic Voltage Restorer (DVR), voltage flicker,Levenberg Marquardt Constructive Neural NetworkÂ
Copyrights © 2015