| With the continuous consumption of fossil fuels,it has become a consensus to gradually replace fossil energy with renewable energy in the world.Wind power is clean and renewable,and is currently one of the most mature power generation methods in the field of renewable energy.In my country,wind power has become the third largest energy source after thermal power and hydropower.However,the randomness and instability of the wind itself causes drastic fluctuations in wind power,which makes it difficult for the power grid to dispatch.Accurate prediction of wind power can make corresponding dispatching strategies for the power system in advance.Wind power interval prediction can help decision-makers better assess the uncertainty and potential risks brought by wind power grid integration by giving the confidence interval of wind power power.Wind power interval prediction is mainly divided into two categories: the first type is to build a dual-output model of wind power power based on neural networks to predict the wind powe r interval;the second type is to pre-suppose or estimate the probability distribution function of the wind power prediction error.Reverse operation to generate confidence in terval.This paper conducts an in-depth study of the first type of wind power interval prediction method,the main work is as follows:(1)Considering that the existing wind power interval prediction preprocessing technology has the problem of modal aliasing imagination,a wind power interval prediction preprocessing method combining CEEMDAN and FIG is proposed.First,the wind power time series CEEMDAN is decomposed,and then the sub-components are screened for fuzzy information granulation,and finally th e maximum sequence,minimum sequence,average sequence and other sub-components after the fuzzy information granulation are input into the prediction model;(2)Considering the hypothetical probability distribution function of the second type of wind power interval prediction method and the complicated calculation,a multi-step interval prediction method for wind power based on CEEMDAN and CNN-BiLSTM dual output models is proposed.First,the initial wind power time series are slightly fluctuated to generate the initial upper and lower power series.Then the upper and lower sequences are decomposed by CEEMDAN respectively.Secondly,all the decomposition components of the two sequences are used as the input of the prediction model,and the prediction results of the respective components of the two sequences are superimposed to obtain the initial prediction interval.Finally,the improved coverage width criterion CWCproposed is used as the objective function to optimize the interval,and the wind power prediction interval under a given confidence level is obtained.Use actual wind farm dat a to verify,and compare with CNN-GRU,CNN-LSTM,KELM,SVR four models,prove the effectiveness of CNN-BiLSTM prediction model;(3)Considering that the multi-step interval prediction method of wind power based on CEEMDAN and CNN-BiLSTM dual output models cannot introduce other factors that affect wind power,such as wind speed,wind direction,temperature,etc.,a combination based on CEEMDAN-FIG and CNN-BiLSTM is proposed.Model-based multi-step interval forecasting method for wind power.First,perform CEEMDAN decomposition on the wind power time series.Then,the optimal component is selected for fuzzy information granulation.Then input the granulated information component and the non-information granulated component into the prediction model to obtain the initial wind power prediction interval.Finally,the CWCproposed indicator is used as the objective function to optimize,and the confidence interval is generated.After simulation verification,compared with the multi-step interval prediction method of wind power based on the CEEMDAN and CNN-BiLSTM dual-output model,at the same confidence level,the multi-step interval prediction method of wind power based on the combined model of CEEMDAN-FIG and CNN-BiLSTM All of the interval forecast indicators are better than the former. |
