Research On The Method Of Wind Farm Layout Optimization Under Uncertain Wind Condition

Wind energy is widely utilized and continuously developed as an environmentalfriendly and renewable energy resource.The utilizations of wind energy face the problem caused by the variability,stochastic and uncertainty of wind.The short-term wind forecasting is necessary to enhance the stability of wind power systems.The longterm establishment of the wind probability distribution is necessary for wind resource assessment.The wind farm layout optimization(WFLO)problem aims to maximize the energy production of a wind farm by the optimal placement of turbine based on the probability distribution of wind speeds and directions.Currently most of existing WFLO methods ignore the variation and uncertainty of wind conditions.The studies in this thesis include wind forecasting and uncertainty analysis,the grid-based WFLO algorithm and robust quantification method of the energy production of a wind farm.The detail of research works and contributions is as follows:1.A wind speed prediction algorithm is proposed based on the combined autoregressive and k-nearest-neighbor regression model.The regression parameters are trained to capture both recent correlation and variation pattern of wind speed series.The simulations verify that the prediction error of the algorithm is less than the traditional regression model in most wind conditions.A method for quantifying the uncertainty of discrete wind distribution is proposed.The discrete probability distribution for wind speed and direction is expressed by the wind probability vectors.The non-parametric test based on the wind distribution vectors is adopted to evaluate the difference between historical and future wind distributions.The probability distribution of the the wind distribution vectors is modeled based on the Dirichlet-multinomial model for estimating the variation range of future wind distributions.2.A flexible grid-based WFLO method is proposed to optimize the grid parameters for generating grid points in wind farm area and the bit string for representing the occupation of grid points by turbines.The grid parameters are designed to simplify the constraint handling about the boundary of the feasible area and the safe distance between turbines.An estimation model of the capacity factor for grid-based wind farm is proposed based on the relationship between grids occupation ratio and capacity factor.For bit string optimization,a customized binary-coded genetic algorithm is proposed to avoid the individual with unsatisfing constraints and duplication in population.The simulation results demonstrate that the optimization results and convergence of solution of the proposed method is better than other grid-based WFLO methods.3.The WFLO strategy is proposed to improve the robustness of energy production under uncertain wind distributions.Robustness quantification of the energy production is implemented by the Value-at-Risk assessment in stochastic optimization and the worst case estimation in robust optimization.In stochastic optimization,normal approximation of wind distribution vector with the Dirichlet distribution is applied to estimate the probability distribution of energy production and calculate value at risk of energy production.In robust optimization,maximization of minimum energy production within uncertainty set of wind probability vector is solved by nested optimization method.The simulations verify that the optimization strategy achieves the trade-off between maximizing the energy production and minimizing the potential loss of the energy production caused by uncertain wind conditions and can also reduce the fluctuation of power and annual energy production.