Time Series Prediction of Wind Power using Autoregressive- Moving Average Model

Abstract

Proper forecasting of wind is a significant requirement for an effective
way of incorporating renewable energy in contemporary power
systems. The nonlinear and highly changing nature of wind speed
is difficult to forecast, particularly when it comes to models of the
wind power curve, which is a union of wind speed and power output.
This paper uses the Auto Regressive Moving Average (ARMA)
model to predict the wind power curves based on the past wind speedpower
records. Being a time series model, the ARMA model is used
to represent the time variation in the data and offer predictions with
a statistical basis. The standard error measures that were used to
evaluate model performance with Mean Absolute Error (MAE) and
Root Mean Square Error (RMSE). Conclusions provided find that,
using the ARMA model can give credible predictions of wind power
curves; it is a simple, although effective method to forecast wind energy.
The results not only emphasize the Autoregressive Moving Average
Modeling as a statistical tool to integrate renewable energy, but
also establish the basis to do more research on it, which may include
some hybrid and ensemble approaches to represent an even better
forecasting pinnacle.