Wind power prediction based on high-frequency SCADA data along with isolation forest and deep learning neural networks

Abstract

Wind power plays a key role in reducing global carbon emission. The power curve provided by wind turbine manufacturers offers an effective way of presenting the global performance of wind turbines. However, due to the complicated dynamics nature of offshore wind turbines, and the harsh environment in which they are operating, wind power forecasting is challenging, but at the same time vital to enable condition monitoring (CM). Wind turbine power prediction, using supervisory control and data acquisition (SCADA) data, may not lead to the optimum control strategy as sensors may generate non-calibrated data due to degradation. To mitigate the adverse effects of outliers from SCADA data on wind power forecasting, this paper proposed a novel approach to perform power prediction using high-frequency SCADA data, based on isolate forest (IF) and deep learning neural networks. In the predictive model, wind speed, nacelle orientation, yaw error, blade pitch angle, and ambient temperature were considered as input features, while wind power is evaluated as the output feature. The deep learning model has been trained, tested, and validated against SCADA measurements. Compared against the conventional predictive model used for outlier detection, i.e. based on Gaussian processes, the proposed integrated approach, which coupled IF and deep learning, is expected to be a more efficient tool for anomaly detection in wind power prediction.

A Critical Review of Wind Power Forecasting Methods—Past, Present and Future

“…Lin et al [36] implemented isolation forest (IF) along with deep learning neural network to detect outliers for more accurate wind power forecasting. Wind speed, wind direction, air temperature, etc., were extracted from a supervisory control and data acquisition (SCADA) dataset of an offshore wind turbine to be used as inputs while employing wind power as the output in the predictive model.…”

“…Manobel et al [47] applied the Gaussian process (GP) for detecting and removing outliers from SCADA data, where RMSE was improved by 25% in comparison with the standard forecasting methods. Besides, Lin et al [30,36] used IF to deal with outliers to increase accuracy. The results showed that preprocessing the SCADA data would develop more accurate forecasting.…”

Comparison of New Anomaly Detection Technique for Wind Turbine Condition Monitoring Using Gearbox SCADA Data

“…The score will tend to 1 when the expected path length for an Isolation Tree instance tends to 0. Isolation Forest has been used for pre-processing for one paper [28] before a power prediction technique is done with neural networks. The IF detected outliers in the power curve to clean it, then the neural network predicted power output for OREC's 7MW Levenmouth demonstrator turbine.…”

“…The number of outliers expected in the training is based on an expected contamination parameter provided to the model by the user. To the best of the authors' knowledge Elliptical Envelope has not been considered for fault detection, but as shown previously was used for power curve cleaning [28].…”

A Novel Deep Learning Approach for Wind Power Forecasting Based on WD-LSTM Model

“…et al integrated IF with deep learning and proposed a novel approach to perform power prediction using high-frequency SCADA data. Compared with the conventional predictive model used for outlier detection, the proposed deep learning prediction model shows superiority in wind power prediction [40].…”