Abstract
Markov chains (MC) are statistical models used to predict very short to short-term wind speed accurately. Such models are generally trained with a single moving window. However, wind speed time series do not possess an equal length of behavior for all horizons. Therefore, a single moving window can provide reasonable estimates but is not an optimal choice. In this study, a forecasting model is proposed that integrates MCs with an adjusting dynamic moving window. The model selects the optimal size of the window based on a similar approach to the leave-one-out method. The traditional model is further optimized by introducing a self-adaptive state categorization algorithm. Instead of synthetically generating time series, the modified model directly predicts one-step ahead wind speed. Initial results indicate that adjusting the moving window MC prediction model improved the forecasting performance of a single moving window approach by 50%. Based on preliminary findings, a novel hybrid model is proposed integrating maximal overlap discrete wavelet transform (MODWT) with auto-regressive integrated moving average (ARIMA) and adjusting moving window MC. It is evident from the literature that MC models are suitable for predicting residual sequences. However, MCs were not considered as a primary forecasting model for the decomposition-based hybrid approach in any wind forecasting studies. The improvement of the novel model is, on average, 55% for single deep learning models and 30% for decomposition-based hybrid models.
Highlights
Wind speed forecasting models can be classified from two perspectives: (a) time scale (b) applied methodology
Decomposition-based hybrid model: Even though decomposition models are extensively studied for wind forecasting [35], [72], very few studies are based on decomposition-based Markov chains (MC) models
To address the third problem, a novel decomposition hybrid model is proposed based on the maximal overlap discrete wavelet transform (MODWT), autoregressive integrated moving average (ARIMA), and modified Markov Chain model
Summary
Wind speed forecasting models can be classified from two perspectives: (a) time scale (b) applied methodology. Based on the applied methodology, wind speed forecasting models can be divided into five categories: i) Persistence, ii) Physical, iii) Statistical, iv) Artificial Intelligence/Machine Learning (AI/ML), and v) hybrid. The advanced learning methods tend to fall into local optima Such problems are not associated with statistical models that make them suitable for very short to short-term wind speed forecasting. The discretization of the data into a proper number of states still needs improvements Another parameter that limits the forecasting accuracy of the traditional model is the size of modelling data (hereafter called window size). He et al [54] constructed finite-state MC considering data of three hours and for each individual month In this way, the diurnal non-stationarity and the seasonality of wind time series are accounted without complex models. A balance is to be maintained between computational complexity and forecasting accuracy
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have