A Novel Short Term Wind Speed Forecasting based on Hybrid Neural Network: A Case Study on Smart City in India

Abstract

In the smart cities projects, short term wind speed forecasting is a challenging task in the computational science. This paper presents a novel approach for the optimization of neural network parameters on short term forecasting using Design of Experiments (DOE). Experiments are conducted by varying the neural network parameters related to architecture of the neural network. 18 years collected MERRA data (2000- 2018) at the Madurai in Tamil Nadu, India were used. In this study, neural network parameters namely data split, no of hidden layers, drop out, optimizer, function, learning rate and weight initializer are optimized with multi responses. Experiments were conducted on L8 Orthogonal array and analyzed the degree of influence of neural network process parameter on individual performance characteristic using Data Envelopment Analysis (DEA). This Neuro DOE – DEA model is unique in this research.