| With the world’s energy demand continues to grow,the global energy crisis is increasingly steep.Coupled with the environment problems caused by traditional energy become more severe,renewable energy has been vigorously developed.Photovoltaic generation,as a green renewable energy,has received widespread attention with its permanence,cleanliness and flexibility.However,distributed photovoltaic generations have features of randomness and intermittence,the changes of power generation will lead to voltage pulsation at the access point and make the distribution network’s control more difficult.Therefore,it is of great theoretical and practical significance to study the uncertainty of the output power of distributed photovoltaic power generation system.Around the subject,the main contents of this paper are as follows.(1)Probabilistic research of photovoltaic power based on orthogonal seriesIn existing PV power output probability modeling,parameter analysis method requires presuppositions parameter distribution and cannot consider the impacts of random factors.Kernel density estimation has different selections of bandwidth value.For this problem,the probability model of PV output power is directly established based on orthogonal series theory.in the paper.The accuracy and validity of proposed model are verified using measurement historical data of the PV output power,combining with goodness of fit test and error analysis.(2)Interval prediction of photovoltaic power based on hierarchical clustering and dual output PSO-ELMThe traditional PV output power prediction uses the indirect point value prediction.The model is based on the relationship between the irradiance and the photovoltaic output power,ignoring the diversity of the factors affecting the PV output power.Simultaneously,deterministic point value prediction has large error range and low stability.Aiming at this problem,the interval prediction model based on scene classification and extreme learning machine is proposed in this paper.According to the measured meteorological data and the photovoltaic power,the prediction intervals are given directly.Then using particle swarm optimization algorithm to optimize the output weight of the model,the prediction intervals,which taking into account reliability and clarity,are obtained.(3)Fast interval prediction of photovoltaic power based on boosting and ELMConsidering the optimization process of ELM takes a long time,a fast interval forecasting model of PV output power based on boosting is established in the paper.Taking the ELM model as the base learner,the sample with large error and the base learner with good learning ability are focused according to the prediction results of the validation set.by using Adaboost algorithm.In the process of learning,the weight of sample and learner are updated to get the combination prediction interval.It is verified by the example simulation that the proposed model improves the computation speed of interval prediction technique.(4)Interval prediction of photovoltaic power based on online sequential learning machineConsidering the fact that the data of the new distributed photovoltaic power generation system is small,the sampling information cannot be obtained at one time and the information is redundant,the correlation coefficient is used to extract the variables which are related to the output power as model inputs.Then using the characteristics that online sequential learning machines can batch samples without having to repeat historical data,online PV output power interval prediction model,which can be successive iterations,is established in the paper.Experiments show that the proposed model can meet the needs of online application in order to ensure certain accuracy. |
