International Journal of Innovative Technology and Exploring Engineering (IJITEE) ISSN: 2278-3075, Volume-9 Issue-5, March 2020 389 Published By: Blue Eyes Intelligence Engineering & Sciences Publication Retrieval Number: E2252039520/2020©BEIESP DOI: 10.35940/ijitee.E2252.039520 Abstract: Due to the stochastic nature of wind speed, accurate wind power prediction plays a major challenge to power system operators for unit commitment and load dispatching. To predict wind power production with great accuracy, wind speed forecasting in different time horizons is gaining importance nowadays. This paper explores the application of Adaptive Neuro-Fuzzy Inference Systems (ANFIS) to forecast the wind speed in Logan international airport, USA for one year in every one hour time interval. ANFIS with different structures and membership functions are trained to find out the best model to do short term wind forecasting. Simulation with the best model is performed in MATLAB and the results show that the three input model with wind speed, direction and air pressure as inputs using Gaussian bell membership function provides the smallest errors. Keywords : ANFIS, Logan airport, MAPE, Wind speed I. INTRODUCTION As wind is intermittent in nature, Independent Power Producers (IPP) need to forecast wind power upto a few hours ahead with reasonable accuracy for proper scheduling of electrical power in a power system[1]. Accurate prediction of wind is necessary to help the Load Dispatch Centers to manage the grid operations in an optimal fashion and power trading. Because of high penetration of wind energy into electric grids, bulk intermittent generation affects grid security, system operation and market economics. Further, deregulation of electric industry as well as power trading increase the importance of accurate forecasting. Operation and maintenance of wind turbine can also be planned based on the information of wind data at the specific location. Different techniques such as statistical methods, persistence method, numerical weather prediction method, artificial intelligent methods and hybrid methods[2] are used for accurate wind forecasting. This paper focuses on wind speed forecasting in Logan international airport by Adaptive Neuro-Fuzzy Inference Systems (ANFIS)[3], which is a combination of Artificial Neural Network (ANN) and Fuzzy Inference System (FIS). ANN has the capability of looking for patterns Revised Manuscript Received on February 06, 2020. * Correspondence Author V.Vanitha*, Department of Electrical and Electronics Engineering, Sri Krishna College of Technology, Coimbatore, India. Email:vvanitha55@gmail.com D.Magdalin Mary,Department of Electrical and Electronics Engineering, Sri Krishna College of Technology, Coimbatore, India. Email: magdalinmary.d@skct.edu.in G.Sophia Jasmine, Department of Electrical and Electronics Engineering, Sri Krishna College of Technology, Coimbatore, India. Email:sophiajasmine.g@skct.edu.in Akhil Balagopalan, Department of Electrical and Electronics Engineering, Amrita School of Engineering, Coimbatore, India. Email: akhilb09cc@gmail.com in the information presented to it and thus learns about the system. FIS produces output by analyzing previous experience. Self decision making capacity makes FIS suitable for prediction purpose. Combination of ANN and Fuzzy logic approaches perform better than individual ANN and Fuzzy forecasts, and ANN-Fuzzy approach provides excellent performance. When added with ANN, FIS acts as a feedback and gains more experience and produces accurate output[4][5]. II. METHODOLOGY OF PRESENT STUDY In the present study, wind speed prediction based on fuzzy logic and artificial neural network is used, which provides significantly less rule base but also increased estimated wind speed accuracy when compared to traditional fuzzy logic alone. It involves the following steps:(i)Data collection (ii) Data normalizing (iii) ANFIS structure selection (iv) Data training and (v) Data testing. Fig.1 shows the flow chart of methodology for the proposed study. Fig. 1. Flow chart of the methodology. A. Data Collection Good Hourly average values of wind speed, wind direction, air pressure and temperature have been used for the study from MIT site. The data was collected from a site that had a weather monitoring system having sensors for the four parameters. The hourly average values over a period of one year were available for the study. The annual data obtained from the site had missing and out of range values to the tune of 22% due to poor site conditions and communication errors. These deformities have been rectified by manual extrapolation using the values at the nearest time step and the refined data are used for generating the FIS system. These measured data have been averaged for every 30 min interval. Adaptive Neuro-fuzzy Inference System Based Short Term Wind Speed Forecasting V.Vanitha, D.Magdalin Mary, G.Sophia Jasmine, Akhil Balagopalan