Current Perspective on the Accuracy of Deterministic Wind Speed and Power Forecasting

Abstract

The intermittent nature of wind energy raised multiple challenges to the power systems and is the biggest challenge to declare wind energy a reliable source. One solution to overcome this problem is wind energy forecasting. A precise forecast can help to develop appropriate incentives and wellfunctioning electric markets. The paper presents a comprehensive review of existing research and current developments in deterministic wind speed and power forecasting. Firstly, we categorize wind forecasting methods into four broader classifications: input data, timescales , power output, and forecasting method. Secondly, the performance of wind speed and power forecasting models is evaluated based on 634 accuracy tests reported in twenty-eight published articles covering fifty locations of ten countries. From the analysis, the most significant errors were witnessed for the physical models, whereas the hybrid models showed the best performance. Although, the physical models have a large normalized root mean square error values but have small volatility. The hybrid models perform best for every time horizon. However, the errors almost doubled at the medium-term forecast from its initial value. The statistical models showed better performance than artificial intelligence models only in the very short term forecast. Overall, we observed the increase in the performance of forecasting models during the last ten years such that the normalized mean absolute error and normalized root mean square error values reduced to about half the initial values.

Wind Speed Forecasting Based on Extreme Gradient Boosting

“…The statistical model mainly combines the characteristics of time-series data to forecast wind speed and optimize the model parameters by using the forecast errors [5]. Common statistical approaches include Kalman filter and Markov chain [14], auto regressive model, autoregressive moving average [15][16][17], autoregressive conditional heteroscedasticity model [18], exponential smoothing [19] and autoregressive integrated moving average [20,21]. Nevertheless, among these models, assumptions need to be made with regard to the data distribution before modeling, and the obtained results are not satisfactory for nonlinear time series forecasting [22].…”

Genetic Algorithm Based Optimized Feature Engineering and Hybrid Machine Learning for Effective Energy Consumption Prediction

“…The observed MSE for the proposed model is 192.32. The mean absolute error (MAE) describes the difference between the initial value and the forecasted value and is extracted as the mean absolute difference in the data set [44]. The MAE for the proposed model is 10.05, calculated using the Equation 15.…”

Realistic Optimal Power Flow of a Wind-Connected Power System With Enhanced Wind Speed Model

“…Wind speed forecast models aim to estimate wind speed for a future time. They can be categorized into two types: time-scale horizon and model source types [23], [24].…”