Research On Short-term Wind Power Forecasting Based On Modal Decomposition And Optimization Algorithm

The shortage of traditional energy has always been a problem plaguing countries all over the world,and the world has begun to accelerate the transition from traditional energy structure to new energy structure.Although affected by the epidemic,carbon emissions have declined slightly,but with the acceleration of the global economic recovery,countries are still very dependent on traditional fossil energy.Vigorously developing new energy is an important way to solve the global energy problem.Among them,wind energy has become the fastest growing type of renewable energy and is the focus of the development of the new energy supply system.However,due to the high volatility of wind energy itself and the influence of regional environment,there are still certain problems in wind power generation systems.my country is still immature in the forecasting technology of new energy power generation,and the forecasting of the frequent occurrence of high intermittent events in the wind power system is still not accurate enough.Therefore,it is still necessary to increase the research on wind power forecasting and provide an effective,reliable and stable high-precision forecasting scheme.Therefore,this paper divides multiple links of wind power prediction layer by layer,improves the algorithm and model from different levels,and constructs multiple sets of systematic effective and complete wind power prediction schemes from two aspects of wind power point prediction and interval prediction.The research contents are as follows:(1)Aiming at the problem that the traditional decomposition method is easy to lose the information sequence,a set of method for constructing wind power interval prediction based on the lost information and kernel density estimation is proposed.At the same time,based on the wind power energy theory,the optimization method is adopted to optimize the modal number problem in the decomposition process.(2)In view of the poor convergence of traditional neural network,this paper proposes to adopt different mathematical methods to enrich the diversity of algorithm population,avoid local convergence to a great extent,and optimize the neural network based on the improved algorithm.At the same time,taking advantage of the advantages of MCMC method in dealing with random sequences,a wind speed interval prediction scheme based on MCMC sampling and error optimized neural network is proposed.(3)Aiming at the problem of modal aliasing that is easy to occur in the modal decomposition method,it is proposed to use the CEEMDAN method to process the wind speed sequence.This method can effectively solve the problem of noise residue,and at the same time effectively separate the modes,and optimize the limit based on the particle swarm algorithm.The learning machine model uses four seasons data for simulation,and builds a set of high-precision wind speed point forecasting scheme based on data decomposition combined with algorithm optimization of artificial intelligence model.