Research On Wind Power Forecasting Method Based On Deep-Learning Network And Its Application

Along with the wind power capacity increasing, the impact of wind power access to power grid become more and more prominent.Due to the volatility and intermittent of wind power,large-scale wind power connected to the grid will inevitably affect the grid.An effective prediction of wind power can reduce the adverse impact of the wind power access to power grid, and can be helpful for power grid scheduling.Therefore, the study of wind power prediction is of great importance.This paper, taking a wind farm as the object, based on the deep learning algorithmwe build a power prediction model. And we discussed the practicability of the model, to find a power prediction method with higher prediction precision.Based on the research and development of the model and the demand analysis,we build a wind power prediction system.In this paper, the main content can be concluded in the following four paragraphs.1. We did preprocessing and statistical analysis for numerical weather prediction data, the analysis results show that the deviation of the numerical weather forecast wind speed data and the actual wind speed is large, it is necessary for the correction.Through the correlation analysis we can know the relationship betweennumerical weather forecast wind speed and the actual wind speed, so it can be corrected.2. Based on the deep learning algorithm, the numerical weather forecast data correction model is established.The output of this model is used as input parameters in the power prediction model.Based on the deep learning algorithm,the wind power prediction model considering wind speed, wind direction, temperature, air pressure and history power outputis established.And then use the model to predict the output power of wind turbine within 24 hours. The results showed that the deep learning algorithm is much more accurate than BP algorithm, which verified the effectiveness of the proposed algorithm.3. We did research in the using of deep learning network in combination forecast. Respectively, we set up four combined forecasting models: combined forecasting model with support vector machine(SVM) model and auto-regressive and moving average(ARMA)model; combined forecasting model with deep learning network model and SVM model; combined forecasting model with deep learning network model and ARMA model; combined forecasting model with deep learning model, SVM model and ARMA model. Then we analyzed the simulation results of the four models, we found that the deep learning network model practical in the combination forecast, and combination forecast can improve the prediction accuracy to a certain extent.4. Wind power prediction system is designed based on the wind power prediction model which we proposed before. The system includes data collection module, main database module, algorithm module and the forecast data releasing platform module.