Research On Wind Power Ramp Characteristics And Prediction

In the case of the energy lacking and environment problems prominent, windpower as a kind of clean and renewable energy, showed its strong developmentprospects. However, with the wind power penetration constantly increasing, therandomness and uncertainty of wind power brought new risks to the safe operationof power grid, and became a significant constraint for wind power industrydevelopment. In order to reduce the effect of wind power integration on powersystem, general response is predicting the power output value of wind farm and thenplanning timely. But traditional predictions for wind power did not fully take intoaccount the ramp events, making it difficult to deal with risk events which affect thepower system most seriously and not easy to be predicted. Therefore grasping theramp event laws, strengthening the capabilities of ramp events early warning, andimproving the prediction accuracy of ramp information become the key to solve thelimitation of wind power development.With previous studies, this paper on the basis of the wind power rampdefinition, designed a wind power detecting method and its procedure by slidingwindow technology. The method dynamically detected the key information such asstarting time, amplitude, rate and duration of wind power ramp event by reasonablysetting the rate threshold and sliding window parameters. According to the rampevents detecting results for an actual wind farm output, this paper classified andcounted the historical data, then analyzed and counted the ramp characteristics,showed ramp amplitude and time distribution in different statistical cycles anddifferent direction. After above works, differences between time-power seriesfeatures before and during the ramp event were got and the reasons of severaltypical ramp events were found. Based on the analysis and statistical results forwind power ramp events characteristics, support vector machine (SVM) was used toconstruct the rolling progressive multivariate model for extracting the precursorcharacteristics of ramp events, early-warning the ramp events and predicting theamplitude or duration of the ramp events by selecting the specific parameters.Examples showed that the designed method could accurately complete the windpower quick ramp event warning and prediction. Based on the ramp prediction,ramp amplitude predicting error probability distribution was fitted as normaldistribution, and a technique to calculate the optimal amount of SR that the systemoperator should provide, at the same time being able to respond to not only errors inthe forecasts for load and wind power production but also the errors in the forecastsfor ramp events was proposed, and an example was given to test the method. This research work is supported by the National High Technology Research andDevelopment of China (863Program)(NO.2011AA05A105).