Transient Stability Analysis Using MATLAB Simulink and ... - IRD India

More documents

Recommendations

Info

I n p u t S i g n a l sO u t p u t S i g n a l sInternational Journal on Advanced Computer Theory and Engineering (IJACTE)adaptive system. By adaptive, it means that eachparameter is changed during its operation and it isdeployed for solving the problem in matter. This iscalled the training phase.A artificial neural network is developed with asystematic step-by-step procedure which optimizes acriterion commonly known as the learning rule. Theinput/output training data is fundamental for thesenetworks as it conveys the information which isnecessary to discover the optimal operating point. Inaddition, a non- linear nature make neural networkprocessing elements a very flexible system.Input LayerMiddle LayerOutput LayerFig 7 Structure of Artificial neural networkThe neuron computes the weighted sum of the inputsignals and compares the result with a threshold value,. If the net input is less than the threshold, the neuronoutput is –1. But if the net input is greater than or equalto the threshold, the neuron becomes activated and itsoutput attains a value +1.X ix i w i1 1,if X Y 1,if X nThe neuron uses the following transfer or activationfunction. This type of activation function is called a signfunction.Fig 8 Possible active functions used in Neural NetworkV. TRAINING OF NEURAL NETWORKTraining to neural network small adjustments in theweights to reduce the difference between the actual anddesired outputs of the perceptron. The initial weightsare randomly assigned, usually in the range [0.5, 0.5],and then updated to obtain the output consistent with thetraining examples. It can be done by going to matlabnntoolbox.NN toolbox can be open by enteringcommand >>nntool. It will open NN Network/ DataManager screen.Let P denote the input and T denote thetarget/output.In Matlab as per the guidelines of implementationthese are to be expressed in the form of matrices:T = [0 1 1 0]To use a network first design it, then train it beforestart simulation. We follow the steps in order to do theabove:Step-1: First we have to enter P and T to the NNNetwork Manager. This is done by clicking New Dataonce.Step-2: Type P as the Name, and corresponding matrixas the Value, select Inputs under Data Type, thenconfirm by clicking on Create.Step-3:Similarly, type in T as the Name, andcorresponding matrix as the Value, select Targets, underData Type , then confirmStep-4: Now we try to create a Network. For this clickon New Network.Make Sure the parameters are as follows:Network Type = Feed forward – Back propTrain Function = TRAINLMAdaption Learning Function = LEARNGDMPerformance Function = MSENumbers of Layers = as per the requirementStep-5: Select Layer 1, type in 2 for the number ofneurons, & select TANSIG as Transfer Function.Select Layer 2, type in 1 for the number of neurons, &select TANSIG as Transfer Function.Step-6: Then, confirm by hitting the Create button,which concludes the network implementation phase.Step-7: Now, highlight Network with DOUBLE click,then click on Train button.Step-8: On Training Info, select P as Inputs, T asTargets.136ISSN (PRINT) : 2320 – 8945, Volume -1, Issue -6, 2013
International Journal on Advanced Computer Theory and Engineering (IJACTE)On Training Parameters, specify: epochs = 1000Goal = 0.000000000000001Max fail = 50After, confirming all the parameters have been specifiedas indented, hit Train Network.Fig 9 Training process of neural networkNow we can get following plots Performance plot, Itshould get a decaying plot (since you are trying tominimize the error).Training State PlotRegression PlotFig 11 Neural network implemented in smib system.Now for training this neural network according tothe output of our smib system, the output shown in table1 will be taken as the input to this neural network. Forthe better performance of neural our reference datashould be accurate. Output result with fault clearing0.1sec is the most accurate result available, so thisoutput result will be used for training the neuralnetwork. Neural network will obtain its final output bychanging its weight according to the requirement.Neural network is hit and trail method, it will improveits result iteration by iteration and when the final goal isreached it is finally trained. Untrained neural networkcan-not be used for the simulation purpose, training ofthe network is compulsory for its use.Fig 9 shows the training of neural network inmatlab, all the parameters should be within precise limitfor the better operation of the system, after the trainingof neural network is done, various plots for theperformance of neural network is shown, by seeingthese performance plot we are sure that our neuralnetwork is trained as per our requirement, shown in fig10, after verifying all the performance plot now we haveto implement this neural network in our smib system, bytaking power as input characteristic which control theoutput torque angle of the system, we can control theoutput performance of the smib by providing nueralnetwork into the system. The output response of thesystem with neural network and with time domainsolution is shown in fig 12 and fig 13.Fig10 Performance plot of neural networkFig 12 output response of smib system with neural network137ISSN (PRINT) : 2320 – 8945, Volume -1, Issue -6, 2013
Page 1 and 2: Transient Stability Analysis Using
Page 3: International Journal on Advanced C

Transient Stability Analysis Using MATLAB Simulink and ... - IRD India

Create successful ePaper yourself

Delete template?

Save as template?