短期风速统计预报方法的开发研究

发布时间：2018-06-21 22:04

本文选题：河西地区 + 风力发电　；参考：《兰州大学》2014年博士论文

【摘要】：面对煤炭、石油等传统能源资源的日益枯竭,以及日益严峻的环境问题,风能、太阳能等可再生能源已在世界范围内受到重视。其中风能作为重要的可再生能源资源,具有蕴藏量丰富、可再生、分布广、无污染等特性,经过近些年的发展,风力发电在电力发展中占据着不可忽视的地位。风电具有波动性和间歇性特点,大规模风电的接入对电力系统的安全稳定运行带来了挑战。风电功率预测对于电力调度部门根据风电功率变化及时调整调度计划、保证电能质量、减少系统的备用容量、降低系统运行成本都是至关重要的。而风电场风速预测是风电功率预测的基础,因此,提高风电场风速预测的精度,对于风力发电的发展起着十分关键的作用。根据预测周期的不同,风速预测通常可以分为长期风速预测、中期风速预测和短期风速预测。长期风速预测主要用于风电场规划设计;中期风速预测主要用于电力系统的功率平衡和调度、交易、暂态稳定评估;短期风速预测主要用于发电系统的控制,其对于及时纠正电网并网计划中的偏差,完善电网并网计划,充分利用风能,减少因中长期预测中的偏差而限发的电量,并保证电网安全,有着重要的意义。但目前短期风速预测精度依然不足,提高短期风速预测精度成为目前亟待解决的问题。本文的重点集中在短期风速统计预测方法的开发研究上。本文以河西地区的风速为研究对象,系统分析了该地区不同站点的风速和风向的统计规律,并探讨了其变化特征。根据其变化特征,开发了三类较高精度的短期风速统计预报方法,分别为基于周期矫正(SAM)的短期风速预测模型、基于经验模式分解(EMD)的风速预测模型和模型重组的新预测方法。这就为风力发电系统的控制和风电场的短期风功率预测系统的开发提供指导。其主要结果如下:1)开发的第一类研究方法,针对实际风速的复杂周期变化,将SAM应用于风速预测模型中。这类方法提出了两种基于SAM的短期风速预测模型,一种是将SAM和指数平滑法(ESM)相结合,我们称之为SAM-ESM模型,另一种是将SAM和小波神经网络(WNN)相结合,并用遗传算法(GA)对WNN进行了学习训练,我们称之为SAM-GA-WNN模型,利用这两种模型对河西地区的风速进行了短期预测,并将预测结果与传统的持续法(PM)预测结果进行了对比分析,结果表明,基于SAM的短期风速预测方法是一类较优的预测方法,能够提高预测精度。2)开发的第二类研究方法,针对风速数据序列的非平稳性,将处理非平稳信号的EMD方法应用于风速预测模型中。这类方法提出了两种基于EMD分解的风速预测模型,一种是将EMD分解和自回归移动平均(ARMA)模型相结合,我们称之为EMD-ARMA模型,另一种是将EMD分解和BP神经网络(BPNN)相结合,并用粒子群优化算法(PSO)对BPNN进行了学习训练,我们称之为EMD-PSO-BPNN模型,利用这两种模型对河西地区的风速进行了短期预测,并将预测结果与PM模型的预测结果进行了对比分析,结果表明,本研究所开发的第二类短期风速预测方法是一类较优的预测方法,能够提高预测精度。3)开发的第三类研究方法,针对风速变化的不同模式,将模型重组的思想应用到风速预测模型中。这类方法提出了两种模型重组的新预测模型,一种是将ESM模型和WNN模型相结合,ESM模型主要是用来捕获风速变化的线性模式,WNN模型是来捕获非线性模式,并考虑到WNN建模预测的复杂性,采用GA对WNN进行学习训练,我们称之为ESM-GA-WNN模型,另一种是ARMA和BPNN模型相结合,并用PSO对BPNN进行学习训练,我们称之为ARMA-PSO-BPNN模型,利用这两种组合模型对河西地区的风速进行了短期预测,并将预测结果与PM模型的预测结果进行了对比分析,结果表明,本研究所开发的第三类短期风速预测方法是一类较优的预测方法,能够提高预测精度。4)基于上述开发的三类短期风速统计预测方法,对比分析了它们的预测结果,并对它们的适用性进行了研究,整体上来说,SAM-GA-WNN模型和EMD-PSO-BPNN模型是两种较优的模型。
[Abstract]:In the face of the increasingly exhaustion of traditional energy resources such as coal and oil, and the increasingly severe environmental problems, the renewable energy, such as wind and solar energy, has been paid much attention in the world. Wind energy is an important renewable energy resource, which has the characteristics of rich, renewable, widely distributed, and no pollution. After recent development, wind energy has been developed. Power generation plays an important role in the development of electric power. Wind power has the characteristics of volatility and intermittence. The access of large-scale wind power brings challenges to the safe and stable operation of the power system. The prediction of wind power is timely adjusted for the adjustment plan according to the change of wind power, and the quality of power is guaranteed and the system is reduced. It is very important to use capacity to reduce the operating cost of the system. The wind speed prediction is the basis of wind power prediction. Therefore, improving the precision of wind speed prediction is very important for the development of wind power generation. According to the different forecast period, the wind speed prediction can be divided into long term wind speed prediction and medium wind speed. Prediction and short-term wind speed prediction. Long term wind speed forecast is mainly used for wind farm planning and design; medium wind speed forecast is mainly used for power balance and scheduling, transaction, transient stability assessment, short-term wind speed prediction is mainly used for power generation system control, which can correct the deviation in grid connection plan in time and improve grid grid plan. It is of great significance to make full use of wind energy to reduce the limit of electricity in the medium and long term prediction and to ensure the safety of the power grid. However, the accuracy of short-term wind speed prediction is still insufficient and the accuracy of short-term wind speed prediction is an urgent problem to be solved at present. On the basis of the wind speed in Hexi area, this paper systematically analyzes the statistical laws of wind speed and wind direction of different stations in this area, and discusses its change characteristics. According to the characteristics of the wind speed, three kinds of high precision short-term wind speed statistical forecasting methods are developed, which are the short-term wind speed prediction models based on SAM. The wind speed prediction model of empirical mode decomposition (EMD) and the new prediction method of model reengineering. This provides guidance for the control of wind power system and the development of short-term wind power prediction system for wind farms. The main results are as follows: 1) the first kind of research method developed is applied to the wind speed according to the complex periodic changes of the real wind speed. In the prediction model, this method proposes two SAM based short-term wind speed forecasting models. One is combining SAM with exponential smoothing (ESM), which we call SAM-ESM model. The other is combining SAM with wavelet neural network (WNN), and using genetic algorithm (GA) to train WNN, which we call the SAM-GA-WNN model. The two models predict the wind speed in Hexi area, and compare the prediction results with the traditional PM prediction results. The results show that the short-term wind speed prediction method based on SAM is a better prediction method and can improve the prediction precision.2) of the second kinds of research methods for wind speed data sequence. The non stationarity of the column is used to apply the EMD method dealing with non-stationary signals to the wind speed prediction model. Two kinds of wind speed prediction models based on EMD decomposition are proposed. One is combining the EMD decomposition and the autoregressive moving average (ARMA) model. We call it the EMD-ARMA model, the other is the EMD decomposition and the BP neural network (BPNN) phase. Combining and using the particle swarm optimization algorithm (PSO) for learning and training of BPNN, we call it the EMD-PSO-BPNN model, using these two models to predict the wind speed in the Hexi region, and compare the prediction results with the prediction results of the PM model. The results show that the second kinds of short-term wind speed predictor developed by this research have been developed. The method is a kind of better prediction method, which can improve the prediction precision.3). In view of the different modes of wind speed, the thought of model reorganization is applied to the wind speed prediction model. This kind of method puts forward two new prediction models of model reorganization, one is to combine the ESM model with the WNN model, and the ESM model is the main model. It is a linear mode used to capture wind speed change. The WNN model is to capture the nonlinear model and take into account the complexity of the WNN modeling prediction. GA is used to learn and train WNN. We call it the ESM-GA-WNN model. The other is the combination of ARMA and BPNN model, and the PSO is used to train BPNN. We call it the ARMA-PSO-BPNN model. The two combination models are used to predict the wind speed in Hexi area, and the prediction results are compared with the prediction results of PM model. The results show that the third kinds of short-term wind speed prediction methods developed by this study are a kind of better prediction method and can improve the prediction precision of.4) based on the three types of short-term winds developed above. The prediction results of fast statistical prediction are compared and analyzed, and their applicability is studied. On the whole, the SAM-GA-WNN model and the EMD-PSO-BPNN model are two better models.
【学位授予单位】：兰州大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TM614

【参考文献】