Research Of Wind Power Forecast Based On Empirical Dynamic Modeling

China has No.l installed capacity of wind power all around the world,and its newly installed capacity of wind power is also the largest.In China,wind power has played an important role to realize energy structure adjustment and clean fuel substitution.However,because of the intermittency and volatility of wind,the high penetration of wind generation brings a series of challenges to grid operators.Recently,finding the techniques to improve the ability of the grid to bear large-scale wind power penetration has attracted wide attention.Reliable wind power forecast results can provide meaningful forecast information for the power system dispatch,and help to reduce the wind power curtailment.According to the different forms of forecast results,wind power forecasting techniques can be divided into the deterministic forecast and the probabilistic forecast.The output of the former is the expectation of wind power,while the latter considers the forecast error into the results.The result of the probabilistic forecast can be expressed as the prediction interval(PI)or the probability density function(PDF).Currently,the deterministic forecast has been wildly applied and its skill is relatively mature,while the probabilistic forecast can quantify the uncertainty of forecast results and provide more abundant information for grid regulators.So far there are several problems lie in the short-term and ultra-short-term wind power forecast methods.On the one hand,most of the current methods describe the dynamic characteristics of the wind power system by a set offixed equations or functions.The inherent deviation of the equations does harm to the mining of effective information hidden in the data,resulting in unavoidable model errors.On the other hand,the majority of the existing wind power forecast approaches set their sights on one or more wind farms,and seldom cater for a single wind turbine.Because of the critical role of WTPF in wind farm economic dispatch and wind turbine,there is a pressing need for the probabilistic WTPF approaches.In this paper,a novel wind power forecast framework based on empirical dynamic modeling(EDM)is proposed.EDM is a time series analysis method based on attractor reconstruction.A unique advantage of the proposed approach is that its effectiveness depends solely on the potential context hidden in the sequences of the state variables rather than the correctness of the predetermined models.Consequently,this approach can faithfully depict the dynamics of wind farms and improve forecast accuracy.Firstly,this paper analyzes the wind power time series to verify the nonlinearity of the wind power system and further reveals its chaotic nature,indicating the feasibility of using chaotic theory to analyze the dynamic behavior of the system.Then,according to Takens’theorem,a deterministic wind power forecast approach based on empirical dynamic modeling is proposed.In the deterministic forecast,the time series of wind power and the explanatory variables are applied for attractor reconstruction.By this means,the dynamics of the original wind power system can be accurately captured by the trajectories of the state variables in the reconstructed space.Then,using a simplex projection approach,wind power can be directly forecasted with the trajectories.Furthermore,this paper extends the forecasting framework based on dynamic empirical modeling and proposes a non-parametric probability method for wind turbine power forecast.In the probabilistic forecast,the divergence of adjacent trajectories in the reconstructed space indicates the forecasting uncertainty.An enhanced simplex projection algorithm is developed to forecast the probability distribution as well as the prediction interval of wind turbine power with respect to the trajectories.And the real-time updated state vector set in the reconstructed space as well as the particle swarm optimization algorithm further mernts the reliability and flexibility of the proposed approach.In this paper,data from the Global Energy Forecasting Competition 2014 and the Penglai Wind Farm are used for case studies.The results illustrate that the proposed approach can obtain advanced forecast results comparing with the benchmark methods.