Short term wind speed prediction using artificial neural networks

Abstract

As an alternative to fossil fuels, wind is a plentiful, clean, and renewable natural resource for energy. Essentially, power generation from wind depends on wind speed; thus, wind speed prediction becomes increasingly important for modern wind farm management and supply-demand balancing in the Smart Grid. However, wind speed is generally very difficult to estimate, due to its non-stationary and intermittent nature.In this paper, an approach based on artificial neural network (ANN) is developed. The neural network is trained and tested using data from the National Wind Technology Center. Computer simulation results show that the proposed neural network model can successfully predict wind speed in real-time.Index Terms -Wind speed prediction, artificial neural networks.

Short‐Term Forecasting of Wind Gusts at Airports Across CONUS Using Machine Learning

“…Mean CSI values across the eight airports and both seasons increase by 0.05–0.25 when the AR term is included irrespective of whether the full predictor suite or those selected by the stepwise procedure are included. This is consistent with past research (Coburn & Pryor, 2022; Lodge & Yu, 2014) and re‐emphasizes the strong red noise characteristic of WS time series. Performance diagrams for logistic regression model fits at KDEN using the predictor suite and the KDEN real predictor suite suggest that inclusion of the AR term appreciably diminishes performance discrepancies between models developed using the full and “real” predictor suites (Figure 3).…”

“…AR term is included irrespective of whether the full predictor suite or those selected by the stepwise procedure are included. This is consistent with past research Lodge & Yu, 2014) and re-emphasizes the strong red noise characteristic of WS time series. Performance diagrams for logistic regression model fits at KDEN using the predictor suite and the KDEN real predictor suite suggest that inclusion of the AR term of wind gust magnitudes for each airport.…”

Summary of Wind Speed Time Series Prediction Algorithms

“…Single prediction method refers to the use of only one model or algorithm for wind speed prediction research, a single algorithm mainly includes the following several kinds, for example, a simple prediction method that directly takes the actual wind speed of the previous moment as the prediction result of the next moment [5], and a time series to analyze the trend of future changes based on the characteristics of wind speed changes in the past. Column method [6], Kalman filter algorithm [7][8], which solves the system state equation directly based on observation data; neural network algorithm [9], which has good global approximation ability and is suitable for non-stationary data; prediction methods such as support vector machine algorithm [10][11][12], which processes wind speed data based on statistical learning theory. In addition, the wavelet analysis method [13] which has good noise reduction characteristics and the spatial correlation method [14] which considers the spatial and temporal similarity of wind speed have also been applied to wind speed prediction.…”

An ultra‐short‐term wind speed prediction model using LSTM based on modified tuna swarm optimization and successive variational mode decomposition

“…Said this model exhibited a more powerful forecasting capacity for short-term wind speed prediction at wind farms. ANN methods [15][16][17] have the characteristics of parallel processing, distributed storage, fault tolerance, self-learning, strong self-adaptation, and are suitable for solving complex nonlinear problems. Huang et al 18 used the radial basis function (RBF) neural network to predict the wind speed of a certain wind farm in Beijing.…”