Wind Power Forecasting using VARMAX

Wind power being one of the sustainable, renewable energy sources still must compete with the conventional sources of energy on a cost basis. With the use of IIOT data from the turbine, its maintenance requirement, expected production, failure rate and how long it will last can be determined. This paper explores the usability of a turbine's IIOT time series data collected over a period of one year to forecast the future power generation that can be expected for next three months. Ten features were taken into consideration after determining their causal effect on production using Granger causality test. Since some features can affect the production of turbine independently, Vector Auto Regression with Exogenous variable model is used for forecasting. The MAPE rate of 17% was achieved in the forecast of next three months.