Detecting Icing on the Blades of a Wind Turbine Using a Deep Neural Network

The blades of wind turbines located at high latitudes are often covered with ice in late autumn and winter,where this affects their capacity for power generation as well as their safety.Accurately identifying the icing of the blades of wind turbines in remote areas is thus important,and a general model is needed to this end.This paper proposes a universal model based on a Deep Neural Network(DNN)that uses data from the Supervisory Control and Data Acquisition(SCADA)system.Two datasets from SCADA are first preprocessed through undersampling,that is,they are labeled,normalized,and balanced.The features of icing of the blades of a turbine identified in previous studies are then used to extract training data from the training dataset.A middle feature is proposed to show how a given feature is correlated with icing on the blade.Performance indicators for the model,including a reward function,are also designed to assess its predictive accuracy.Finally,the most suitable model is used to predict the testing data,and values of the reward function and the predictive accuracy of the model are calculated.The proposed method can be used to relate continuously transferred features with a binary status of icing of the blades of the turbine by using variables of the middle feature.The results here show that an integrated indicator systemis superior to a single indicator of accuracy when evaluating the prediction model.

Highway icing time prediction with deep learning approaches based on data from road sensors

In harsh climates,highway icing poses a hazard to traffic safety and increases road maintenance costs.It is of great significance to predict when the highway icing may occur and take a preventive plan.However,there are few studies on highway icing time prediction due to the scarcity and complexity of data.In this study,variables of icing temperature,friction,ice percentage,road surface temperature,water film height,saline concentration,and road condition were collected by road sensors distributed on a highway in China.A large-scale time series highway surface information dataset called HighwayIce is formed.Furthermore,a deep learning approach called IceAlarm,composed of long short-term memory neural network(LSTM),multilayer perceptron(MLP),and residual connection,has been developed to predict when the highway will ice.The LSTM is used to process dynamic variables,the MLP is used to process static variables,and the fully-connected layers with residual connections are used to make a deep fusion.The experimental results show that the average mean absolute error before icing using the IceAlarm model is about 6min and outperforms all baseline models.The HighwayIce dataset and IceAlarm model can help improve the prediction accuracy and efficiency of forecasting real-world road icing time,therefore reducing the impact of icy road conditions on traffic.