Study Of Wind Farm Output Forecast Based On Random Forest Algorithm

In order to improve the increasingly deteriorating ecological environment and comply with the mainstream of international development,vigorous development of renewable energy has become the focus of energy policy in many countries.As an abundant,widespread and easy to develop renewable energy,wind power has been well qualified for large-scale development.However,large-scale incorporation of wind power into grid system brings serious power quality problems and threatens the safe operation of the grid,since it has strong intermittency,volatility and randomness.The accurate prediction of wind farm output allows power dispatch department to pre-arrange scheduling programs,so as to ensure the safe and economical operation of the grid.This paper aims at pursuing a more accurate short-term forecast of wind farm output,using the output prediction algorithm based on random forest wind farm model,which reasonably measures the influencing factors related to wind farm output forecast.The forecast mode of random forest algorithms is put forward based on the practical the actual wind farm output data.Compared with decision trees,support vector machine prediction model,this forecast model gets a good prediction on condition that the preconditions are similar.Firstly,pretreatment was done to the obtained data,methods such as horizontally or vertically adjacent,successive minus,least square method,similar days,normalization and so on are used for detection and correction in view of phenomena such as minimax value,glitches,and distortion amount difference level and so on.Then OOB is adopted to estimate the importance of each characteristic variables,excluding the impact of small property factors and selecting the best combination of input variables.At the same time,two important parameters,namely trees numbers Ntree and split node number Ntree,should be changed,thus the most exact prediction can be obtained based on the parameters under the model.Random forests,decision trees and support vector machine prediction model output are introduced to predict wind farm output based on absolute percentage error and root mean square error of the two models results of the evaluation method.Besides,practical examples are adopted to make simulation analysis.Experimental results show that the proposed model predicts accurately and computes efficiently,has some practical value and prospects.