Abstract
Wind energy is fast developing energy resource as it is renewable, pollution free and abundant. The nonlinear and fluctuation of wind are large demand for enhancing the reliability and accuracy of the power system that combines the wind speed. With an exact wind speed data, power system operators predict the power output for system planning and scheduling. Wind speed prediction is an essential factor in forecasting to attain the efficient wind power utilization. But the prediction time and accuracy performance were not improved using existing techniques. In order to address this problem, Deep Multilayer Perceptron Neural Network based Clonal Selective Optimization (DMLPNN-CSO) technique is introduced. The main objective of DMLPNN-CSO technique is to improve the performance of power generation prediction through wind speed estimation with higher accuracy based on mean air temperature, relative humidity and vapor pressure data. DMLPNN-CSO technique comprises two processes, namely deep multilayer perceptron neural network and clonal data selection algorithm. Initially in DMLPNN-CSO technique, wind turbine is considered as the input and given to the input layer in deep multilayer perceptron neural network. Wind turbine information is extracted and given to the hidden layer 1. In this layer, data preprocessing is performed through the median filter for removing unnecessary data. After that, the preprocessed data is given to the hidden layer 2 to estimate the wind speed. The estimated values are collected and get matched with the pre-stored values in the output layer by using softsign activation function. After estimation of wind speed, the optimal data gets selected by using clonal data selection algorithm for predicting the power generation. Clonal data selection algorithm performs the sorting, cloning and guassian mutation for finding the optimal value in power generation prediction. This in turn helps to improve prediction accuracy and reduce the time consumption. Experimental evaluation of proposed DMLPNN-CSO technique is carried out with respect to number of wind turbine data and data samples. The results showed that DMLPNN-CSO technique produces higher prediction accuracy and wind speed estimation rate when compared to state-of-the-art works.