Research Of Wind Energy Resource Assessment In Wind Farm Macro-siting

Wind farm siting plays a very important role in wind power development and it includes macro-siting and micro-siting. Macro-siting is carried out in the planning stage and it aims to determine the wind farm site in a large area through the analysis and comparison of wind energy resource, weather, terrain, traffic and other construction conditions of several candidate wind farms. Wind energy resource is the most important consideration among all the factors, because the economic benefits of wind farm rely heavily on it. Accurate and effective wind energy resource assessment is the premise of wind farm macro-siting. Meanwhile, the long-term wind speed data derived from wind energy resource assessment in macro-siting is also the foundation of micro-siting. Wind energy resource assessment during the wind farm macro-siting is studied from difference perspectives in this thesis. The main content of the thesis are listed as follows.Representative year method is widely used to assess wind energy resource in China. The errors and the factors influencing the errors are systematically analysed with the hourly wind speeds and wind directions recorded over 20 year period at 8 weather stations. The metrics are selected from four aspects: long-term mean wind speed, Weibull distribution parameters, mean wind power density and mean annual power output of wind turbines. In order to further improve the accuracy, an improved representative year method is proposed. In the aspect of the correlation analysis, reduced major axis(RMA) regression is adopted. RMA regression is on the basis of the assumption that the error exists both in the independent and dependent variables, which makes the algorithm more realistic. In the aspect of data correction, the correction method of combining the wind direction and wind speed is employed to give full consideration to the seasonal variation of wind speed. Simulation results show that the error reductions have been achieved by using the improved algorithm, and in some cases the reduction of error is up to 40.8% compared with the unimproved one in the estimation of the long-term mean wind speed.Neural network based wind energy resource assessment method is able to make use of the ability of the nonlinear fitting of the neural network to establish the nonlinear mapping relationship between the wind farm and weather reference station. For the problem of easy to fall into the local optima, a wind energy resource assessment method based on adaptive particle swarm optimizated neural network is proposed. Adaptive particle swarm optimization receives the weights and thresholds passed from the neural network, and continue to train. Then, an elitist learning strategy is preformed when the evolutionary state is classified as convergence state to avoid falling into the local optima. Simulation results show that the proposed method gives fewer errors compared with the neural network. In some cases, the reduction of error in mean absolute relative error(MARE) and correlation coefficient of long-term hourly wind speed is up to 11.26% and 23.2% respectively.From the perspective of the probability distribution of wind speed, a wind energy resource assessment algorithm based on discrete random variable joint probability distribution is proposed. The hourly wind speeds of the wind farm and reference weather station are considered as discrete random variable, and the joint probability distributions and conditional probability distribution are used to calculate the probability characteristics of long-term wind speed at wind farms. The method has a wider application since there is no assumption about wind speed probability distribution, and the simulation results show that the proposed method is better than traditional linear regression.In addition to the wind speed, the wind direction is also an important factor in wind resource assessment, and it plays a key role in the wind turbines arrangement. The wind direction forecast is independent of the wind speed forecast under normal circumstances. The comparative analysis of three wind direction forecasting methods is carried out, and the three methods are linear regression, matrix method and direct method. The Chi-square goodness of fit is chosen as the evaluate metric. Simulation results indicate that the matrix method has best performance with the Chi-square goodness of fit less than 0.04, but only the frequency in each of the wind direction sectors is obtained. The direct method gives larger error than the matrix method, and the linear regression method gives the biggest error. This conclusion can provide reference for practical application.The impacts of climate change on wind power generation and wind extremes of 50-year are analysed using the meteorological data(1973-2011) in South Korea. Results uncover that the wind power resource and potential wind power generation could be notably impacted by climate change induced variation in mean wind speed, the way to assess wind resource according to observed historical data has inherent shortcoming due to its fail to consider the impact of future climate change. To reduce the impact of climate change on wind energy resource assessment, the whole life cycle of wind energy resource assessment method is proposed, which fully takes into account the future changes in wind speed. Under this idea, the rheumatoid wind index method and gray theory are used to correct the annual power output. The simulation results show that the corrected annual power output is much closer to the true value.