Wind Power Data Storage And Wind Power Prediction Based On Hadoop

The capacity of wind power installation is showing a gradual upward trend in our country.The wind farm centralized control center has widely promoted the use of intelligent monitoring technology,which has generated massive amounts of wind power data,so the highly efficient storage of wind power data has become an issue that has to be considered.At the same time,the wind speed and the wind direction change at any time which causes the output power of the fan has obvious randomness and uncontrollability,so it has become a crucial part of wind power generation that reads valid data from massive data to accurately predict wind power.In response to the above problems,the main work of this thesis is as follows:(1)The current mainstream big data technology products are analyzed.HDFS and Spark are selected to store wind power data and compute in parallel.Aiming at the characteristics of current wind power data,a small file merging algorithm is proposed based on the optimal grouping volume in this thesis.According to a certain merging strategy,small files are merged into a large file as much as possible,and then upload the large file to HDFS to solve the problem of Name Node memory consumption caused by too many small files.(2)Multiple prediction models are trained based on the data set containing time,wind speed,wind direction and other characteristics obtained from HDFS,and the random forest and the LSTM prediction models is selected because of better prediction effects.A combined prediction model is built based on the random forest prediction model and the LSTM prediction model through the covariance optimization method.According to the prediction error information matrix,the most optimal weight in a single prediction model is obtained based on the principle of the smallest square sum of errors.The experiments of the random forest,LSTM and combined forecasting models for wind power forecasting are completed.From the experiment results,the proposed small file merging algorithm based on the optimal grouping volume can achieve highly efficient storage of wind power data,and the combined prediction model can predict wind power more accurately than a single prediction model.