Wind Power Forecasting Technology And Its Operation Analysis

The exacerbated relationship between energy and environment accelerates the development of renewable energy. With the mature generation technology, low generation cost, and convenient large-scale industrialization development, wind power will be a main part in energy and environment sustainable development. The accuracy of wind speed forecasting is related to the wind power scheduling. When large-scale wind power connected into the grid, it also affects the security and stability of the grid. Wind speed forecasting is the correct mean to solve the problem. This paper based on random fluctuation and intermittency of wind discuses the wind speed forecasting correction and economic scheduling. The main contents include that wind speed forecasting, prediction bias correcting, wind power fluctuation, energy storage systems, economic dispatch and auxiliary decision, and so on.(1) This paper applies time series model and Back Propagation (BP) neural network model to predict wind speed. Finally, a combination model of time series and BP neural network is proposed. In the combination model, the inputs of BP neural network are made up of historical data and residual errors calculated by time series model. The main causes of speed prediction error is analysed in this paper. A new wind speed forecasting bias correction method on Empirical Orthogonal Function (EOF) is proposed. Wind speed prediction error can be decomposed by EOF, by which the main components of error are got. Then, bias correction mode can be built by regression analysis. The model can be more accurately in the short-time wind speed forecasting. And then shows an actual example.(2) The fluctuation of wind power is the main researches in recently, and it is the key problem in wind power integration. The symbolic time series analysis method is applied in wind power fluctuation analysis in the paper. Due to the non-uniform distribution of engineering data, an adaptive partition method is proposed, according to the intensity of data sequence distribution, which can achieve non-uniform segmentation, and make the area sensitive to data. Based on the symbol sequence histograms, the key location of original data is caught by inversion of symbol sequence histograms. Finally, the validity of this method is verified by a wind farm wind power measured data, and it can provide a reference for wind power dispatching.(3) It is not accurate that normal distribution and Laplace distribution are used to fit the wind power prediction error. The partition fitting method is put forward in this paper.(3distribution is used to fit in every section of prediction error. Error distribution function in the whole region is got by weighted summation. The mathematical model of energy storage systems capacity considered the prediction error is established. The energy storage capacity is expressed as the function of unserved energy, and detail relationships about the energy storage capacity, error cumulative distribution function and charge state are introduced. Finally, a new index of energy shortage is proposed. As a consequence, the proposed method permits the sizing of energy storage systems as a function of the desired remaining forecast uncertainty, reducing simultaneously power and energy capacity.(4) As the development of wind power, grid-connected wind power becomes more important. Based on taking a random duplicate sample for wind speed probability distributions, calculate the corresponding wind farm output power in the paper. Then, analyzes system state variables by applying the former power flow method. Lastly, the probability descriptions of variables are obtained by probability and statistics methods. The WSCC-3-9case are shown in the paper aimed at verifying the accuracy and utility of the mentioned method. The conclusions are that the method is convenient, and provides the basis for further analysis of wind power.(5) Based on the aerodynamic model of variable speed wind turbine and hill-climbing search this is one of Maximum Power Point Tracking (MPPT) algorithms, definition formula of probability function used in simulated annealing algorithm is referenced in the paper. It is used in defined adaptive step. The purposes are that dynamic characteristics is fine beginning of searching stage, and steady state characteristics is stratifying at the end of searching stage. The model of variable speed wind turbine and the effects of output power when wind speed changed are realized in Matlab. The conclusions are that improved algorithm can make output power stabilized in new state, dynamic characteristics and steady state characteristics are stratifying.(6) It provides an efficient solution for wind power grid-connected operation, that study of economic dispatch. Duo to the limitation of wind power forecasting, the model of wind power prediction credibility is built, then, wind power forecasting error model is analyzed by partition fitting method. Based on those models, the model of storage energy cost considering forecasting credibility is discussed. Combining with thermal units cost, the economic dispatch model of power system containing wind farm is completed. The model considers the probability of loss of load and abandoned wind. And, in order to reduce storage energy cost is showed as the probability of wind power forecasting error. Simultaneously, the storage energy cost adjustment factor is added into the objective function, which is used to auxiliary decide scheduling wind power. The feasibility of the proposed method is verified by typical10-machine system, Monte Carlo simulation technology and fuzzy clustering is used.