Wind Power Ultra-short-term Prediction Model Of RF-TCN-SA Based On Error Correction

Due to its volatility,randomness and intermittency,wind power generation brings a great test to the safety and stable operation of power system.Wind power forecasting technology can provide estimates of wind power output in advance,help power plants and dispatching departments of power system to make reasonable generation,maintenance and dispatching plans,reduce the influence of massive wind power grid connected to the power system,improve the efficiency of wind energy utilization and promote the massive development and utilization of wind power in the future.In order to improve the accuracy of ultra-short-term wind power prediction,this paper proposes a wind power ultra-short-term prediction model combining Random Forest(RF),Temporal Convolutional Networks(TCN)and Self Attention(SA).The details are as follows:Firstly,to ensure the accuracy and completeness of the input data,the anomalous data in the original data are cleaned by the boxplot method,and the null values in the cleaned data set are filled by the Lagrangian interpolation method,and the raw data were normalized by the deviation normalization method in order to unify the data magnitudes.Secondly,the TCN is chosen as the wind power prediction model to avoid the long-term dependence and gradient disappearance and gradient explosion of traditional recurrent neural networks.In addition,in view of the Adam optimizer’s tendency to miss the global optimal solution and the neuron death of ReLU activation function in the original TCN structure,The Lookahead optimizer and the PReLU activation function were used to mitigate the problem.The results of example verification show that the improved TCN model can improve the learning and convergence effect of the model,and improves the prediction performance of the model.Finally,the random forest algorithm was used to reduce the model input redundancy and improve the model learning and training efficiency by filtering features on the original data sample set.The self-attention mechanism was used to assign different weights to the input information at different time to improve the prediction accuracy of the model.In addition,in order to further reduce the wind power prediction error,an error correction model was established to correct the preliminary predicted value of wind power.The measured data of two wind farms,Xinjiang Biesitiereke and Guohua Jingxia North Wind Power are used for example analysis.The results of the analysis and comparison show that the proposed wind power ultra-short-term prediction model of RF-TCN-SA based on error correction can effectively improve the ultra-short-term prediction accuracy of wind power.