Wind Speed Forecasting Based on Neural Network and Least Mean Square Method
Keywords:
Forecasting, Neural Network, Least mean square methods, Short term wind forecastingAbstract
Prediction of future events and conditions are forecasts, and the act of making such predictions is called
forecasting. Forecasting is an integral part of the decision making activities of power sector. The proposed neural
network model is multilayer perceptron type with one hidden layer and one output layer Based on the mathematical
explanations and the Adaptive filter theory we have proposed least mean square (LMS) algorithm for short term wind
speed forecasting. We can make changes in the algorithm by changing the values of parameters disjointedly or
combine. After obtaining the required simulation results the values are finalized. The value of step size chosen to be
best suited for the original data to be generated. After training Neural Network algorithm we got the minimum error
for daily data MAPE 9.59% and monthly data 32.26%.After training Least Mean Square algorithm we got the
minimum error for daily data MAPE 16.68% and monthly data MAPE 54.86%. The results suggest that NN model
with developed structure can perform good prediction with least error.
