Research On Very Short-Term Wind Power Forecasting Models

In the electric dispatching system, predictability is a critical factor to dealwith the uncertainty in wind power fluctuations. Accurate prediction of the windgenerator capacity would help grid dispatching personnel to arrange generatingplan in advance and reduce the difficulties in electric power planning anddispatching for operators. So very short term wind power prediction modelresearch is of the important practical significance.The adaptability of the Markov model and ARIMA model in wind powerprediction were studied. The ARIMA model was built using the Eviews. Thestability of the power data was analysed, and the results showed that power datahad obvious non-stationary so that the differencing scheme was used. Thenaccording to the model determining order principle, the appropriate ARIMAmodel was chosen to forecast dynamically. The four state-spaces were given interms of the coarse to fine equivalent method when studying the Markov model.The first order Markov chain models were established in each state-space torealize the mutli-step predictive.The influence of the different state-spacenumbers on the prediction accuracy was studied. The Markov model could gainreasonable forecast value by improving the extraction method of predicting thesingle point value.The prediction results of the benchmark model(PM model)、Markov model and ARIMA model were compared. The practical measured powerdata in a small wind farm was analysed. The results show that the Markov modelwith the more number of the high state-space can get better precision, and theMarkov model in power prediction has stronger adaption. The Markov modelprovides the uncertainty analysis (probability distribution) of the wind powerprediction, the results will provide for wind power management decision-makingbasis for further.The study wind power prediction theory can improve the matching relation between the electricity supply and demand.And this study is based on the actualoperation data of wind farms, so it meas a lot in both theory and practice forwind power interconnection and dispatching.