Research On Wind Speed And Wind Power Forecasting Related Issues

Along with the increased efforts on advocation for using the clean energy andpromoting energy saving and emissions reduction， the challenge faced by research andapplication of Grid-connected wind power is also increased. Due to the intermittency ofwind power which brings great influence on wind speed and wind power forecasting andtime sequential model of wind power is the basis analysis of scheduling and projectingarrangement in Grid-connected, so it has important research significance. Based on therandom characteristics of wind power, the paper studied a variety of algorithms and methodfor wind speed and wind power forecasting; Established the confidence relationship witha certain degree according to the forecast error which made out to the result of the windpower interval forecasting;On the choice of prediction methods, comparing with the method of time seriesARIMA model and the Elman neural network, used the data from jiugong mountain.While for the wind power, it was used the direct method and indirect method to getresults,then has analogy the load precision index to build wind speed prediction accuracyindex to do the evaluation.Based on the research of the historical data found that it was not enough to justanalysis the algorithm modeling, wind speed and wind power data had lots impact factorsof non-stationary series, which made of different frequencies of tendency item, periodicitem and random item. So used the method of wavelet analysis to carry on the trend ofmulti-scale decomposed for the original data into low frequency and high frequencyrandom item, then according to the different characteristics of the decomposition sequence,the paper used different method to do the forecasting. The objective function and constraintconditions had put forward to solve out the optimum wavelet decomposition scale, makethe decomposition scale value had the certain numerical basis, and can be programmed toachieve.According to the error of wind speed data, divided it into seven kind of wind speedrange then made the range error distribution curve fitting with the improved gaussiandistribution model combined with least-squares method to solving parameters. Based oninterval prediction and wind power forecast error probability distribution, will be based onthe error probability density curve of interval prediction of expanding into the temporalcharacteristics of probability distribution of forecast error, to achieve the purpose of windtrend forecasting. Established error probability statistical model under different single condition according to a large number of historical data, the two kinds of condition,respectively, one was based on the measured error of wind speed（or power）fluctuations inquantity△v=v_t+1-v_t and the other was based on the forecasting error of wind speed（orpower） ξ=v_t+1-v_p ξ=v_t+1-v_p, through the weighting factor correction the integrated errorprobability model and wind parameters trend predictionunder the condition of correlationcondition could be solved out.This article has also put forward a series of improvement measures, in terms ofprobability modeling, it could also considered factors such as seasonal factors and differentfroms in day and night, then more precise model could be set up; Meanwhile it cancombine load forecasting model, go into the research of power system reserve capacityoptimization problem, which meet the expectation of a certain objective function such asoperation cost minimum, under the condition of solving optimal unit combination problem,realize the coordination between the prediction and real-time scheduling.