Imprecise Probability Prediction Method Of Solar Power Ramp Events

With the development of prominent resource crisis and environmental problems,the world has reached a consensus on the gradual replacement of traditional energy by renewable energy like photovoltaic power generation.With the gradual expansion of photovoltaic grid-connection scale,its volatility and randomness brings challenges to the safe and stable operation of power system,which also brings difficulties to the further improvement of photovoltaic permeability.Therefore,the fluctuation of photovoltaic output power which have great influence on the stability of power system has attracted wide attention of scholars.Among them,the large change of solar power output in a short time is called solar power ramp event.Correct identification and accurate prediction of solar power ramp events are of great significance in promoting solar power consumption,reducing operation cost and improving the safe and stable operation capability of power grid.Different from other energy,solar energy resources on which photovoltaic power generation depends have obvious diurnal variation due to the rotation and revolution of the earth,which is only related to time and geographical location.The accurate diurnal model can reflect the maximum power output of the photovoltaic power station in a certain day at each moment.At the same time,the diurnal trend of solar energy will bring the diurnal variation of photovoltaic output.For solar power ramp events,diurnal fluctuations are not the focus of attention as regular expected changes,but the irregular power changes caused by sudden changes of meteorological factors.In this context,if the effect of regular changes caused by daily periodicity is not taken into account in the identification of solar power ramp events,the obtained forewarning signal of ramp event may lose its warning significance to some extent,and even increase the operation cost and damage the reliability of the power system.Most solar power ramp events are caused by extreme meteorological factors and the frequency of such meteorological conditions is low,so there are few historical samples available.For the prediction of solar power ramp events,the limited historical observation is also the key problem that troubles the accurate prediction.Under this condition,if the deterministic prediction results are arbitrarily given based on the limited samples,the prediction may lose the practical application value,even get the wrong prediction and lead to the system instability.This paper first analyzes the diurnal trend of photovoltaic output and proposes a method to identify the solar power ramp event considering the diurnal trend.In this method,a new characteristic quantity of solar power ramp event is defined to effectively eliminate the diurnal variation of solar power.Then,based on the definition of solar power ramp event considering the daily periodic influence,an imprecise probabilistic prediction method of solar power ramp event is proposed.In order to avoid the possible prediction error caused by the limited historical sample,an imprecise probability prediction result is used to reflect the uncertainty caused by the limited sample.The optimal network structure matching the sample data is constructed by the structure learning algorithms.Among them,the imprecise conditional probability parameters reflecting the correlation between the network nodes are obtained by the imprecise Dirichlet model.Finally,according to the given meteorological conditions,the probability interval of each ramp state is obtained by deducing the credal network structure.In the further study of solar power ramp events in the regional power grid,the power ramp events of the photovoltic cluster in the regional power grid are graded according to the allowable power fluctuation range corresponding to the system frequency fluctuation.At the same time,considering the large amount of calculation and many influencing factors in the prediction of power ramp events of photovoltaic power clusters,the data compression and dimension-reduction processing of many meteorological factors of photovoltaic power stations in the cluster area are carried out by principal component analysis.After depression,the principal components of each order obtained are independent of each other.By establishing the direct mapping between meteorological components and each climbing grade,the imprecise probabilities corresponding to different ramp grades are obtained.Finally,the credal classifier is used to further analyze the imprecise probability results of each ramp grade,and output the graded forewarning results of solar power ramp events in the regional power grid.Method in this paper introduces the concept of imprecise to fully describ the inaccurate prediction caused by limited historical samples.The sample-based network structure can achieve the extraction of sample information.The parameters in the prediction model can also be selected based on different' risk attitudes,interval probability results can be obtained correspondingly,which provide a new train of thought in solving the problem limited historical observations.The graded forewarning method can not only achieve the graded imprecise probability prediction of the solar power ramp event in regional power grid,but also can output the graded warning result based on the imprecise probability result.The graded forewarning signal can provide more accurate and more direct warning information for dispatching operation of power grid.