Research On Wind Power Prediction And Green Generation Scheduling Problems

Based on the development and utilization of clean energy and the pollutant discharge of fossil fuels,this thesis studies on the wind power prediction problem and green power generation scheduling.Wind power prediction refers to using data from wind towers and weather forecasts to predict power output of a wind farms during a period in the future.Green power generation scheduling problem decides the start-stop status and generating capacity of each generating unit in the decision-making cycle to maximize the profit of a thermal power plant.Precisely predicting the wind power and properly formulating the power generation scheduling scheme can improve the utilization rate of wind energy and reduce the emission of fossil fuels,which is of great theoretical and practical significance for the green and sustainable development of the power industry.The research of this thesis focuses on the problem that low accuracy of wind power prediction causes the difficulties of wind generation’s grid integration.For ultra-short-term real-time and short-term wind power prediction,a designed batch strategy is used to reduce the prediction complexity and to predict wind power by using the Least Squares Support Vector Machine(LSSVM)method.For short-term wind power prediction which is affected by accumulated error,a hybrid modeling approach based on LSSVM and scenario tree which is generated based on actual data is proposed under the batch strategy.In addition,a wind power prediction system is developed to meet the needs of the actual wind farm.A mathematical model of mixed integer nonlinear programming is set up for the green power generation scheduling problem in consideration of emission penalty and randomness of electricity price.The Lagrangian Relaxation algorithm is used to solve the problem.The main research contents are as follows:1)Considering the long time-consuming and high difficulty coefficient of solving a single turbine prediction,the polymerization batch strategy is designed to divide the units into batches,taking the influential elements of operating parameters,environmental factors and geographical location into consideration.With the sample machine being selected,LSSVM is used to predict the wind power of a batch prototype to improve the efficiency of solving the problem.The experimental results show the effectiveness of the batch strategy and LSSVM.2)Aiming at the problem of accumulated error in short-term wind power prediction,based on the prediction of wind power by using batch polymerization strategy,this thesis proposes the idea of combining LSSVM with scenario tree which is based on the prediction of wind power by using batch strategy.The numerical experiments that are carried out through actual data from wind farm verify the effectiveness of the proposed method.3)By embedding the above models and methods,a wind power prediction system is developed.The system realizes the friendly interaction of the man-machine interface and provides data reference for the overhaul,scheduling department and the integration of wind power into power grid.4)In view of the high emission of power industry and the randomness feature of the price in electricity market,this thesis proposes a generation scheduling problem which concerns the emission penalties and stochastic electricity prices.Using robust optimization method to characterize the stochastic electricity price,a max-min mixed integer nonlinear programming model is established.According to the characteristics of the model,the Lagrangian relaxation algorithm is designed to solve the problem.The numerical experiments based on stochastic examples verify the effectiveness of the model and the algorithm.