风电功率预测方法研究及其应用

发布时间：2018-03-01 18:25

本文关键词： 风电功率预测经验模式分解相空间重构支持向量机元学习组合预测　出处：《湖南大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着大规模风电接入电网，风电功率的波动性和间歇性给电力系统安全稳定运行带来了严峻的挑战。风电功率预测是解决大规模风电接入电网的关键技术之一，能够为电力系统调度控制提供技术支持。目前国内针对风电功率预测方法的研究取得了一定的成果但不够深入，预测系统的开发刚刚起步，缺乏规范和实践经验。因此，对风电功率预测方法研究具有重要的意义。本文以某风电场为研究对象，对风电场风速分布风电功率时间特征展开了系统的研究，以探求更高精度的风电功率预测方法。同时，，根据需求分析研发了风电功率预测系统。主要内容如下：首先，以某风电场为对象，对风速功率特性进行了系统的研究。结果表明：风速的概率分布呈现威布尔分布特征；风电功率具有随机性波动性和混沌性，随着时间尺度的降低风电功率的波动性减弱。然后，提出了一种以经验模式分解（Empirical Mode Decomposition,EMD）和相空间重构为核心的风电功率预测方法，将分解后的本征模函数（IntrinsicModel Function, IMF）分量和剩余分量进行相空间重构，将重构序列输入支持向量机（Support Vector Machine,SVM）模型预测风电功率。结果表明，经过EMD分解降低了建模复杂程度，相空间重构能够降低偏差较大分量对预测结果的影响，提高了预测的精确度。其次，提出了基于元学习的风电功率非线性组合预测模型。以灰色模型时间序列模型线性回归模型和神经网络模型预测结果和预测序列的特征属性作为元预测器的输入，从而发现并纠正基预测器的系统偏差。在元预测器中，采用门控网络函数确定各基预测器权重，保证了权重的时变性和非负性。将该算法应用于风电功率预测，预测结果表明：该算法预测精度高于单一预测算法和常用的组合预测算法。最后，根据需求分析开发了一套风电功率预测系统，实现了对风电功率的短期预测。通过某风电场实测数据测试表明，系统安全可靠，可操作性强，能够很好地实现风电功率预测。
[Abstract]:With the large-scale wind power, wind power fluctuation and intermittence has brought severe challenges to the safe and stable operation of power system. Wind power prediction is one of the key technologies to solve large scale wind power integration, can provide technical support for the dispatch of power systems. The wind power prediction method research some achievements but not deep enough, the prediction system has just started, the lack of standardized and practical experience. It has important significance for wind power prediction method research.
In this paper, a wind farm is taken as the research object, and the wind speed distribution and wind power time characteristics of wind farms are systematically studied, in order to explore a more accurate prediction method of wind power. Meanwhile, a wind power prediction system is developed based on demand analysis.
First of all, with a wind farm as the object, the wind speed? Power characteristics were studied. The results show that the wind speed probability distribution of Weibull distribution characteristics; wind power randomness? Volatility and chaos, with time scale to reduce the volatility of wind power weakened.
Then, put forward a kind of empirical mode decomposition to (Empirical Mode Decomposition, EMD) wind power forecasting methods and phase space reconstruction is the core of the decomposed intrinsic mode functions (IntrinsicModel, Function, IMF) phase space reconstruction and residual components, the reconstructed sequence input support vector machine (Support Vector Machine SVM), the model of wind power forecasting. The results showed that after EMD decomposition reduces the modeling complexity, phase space reconstruction can reduce the influence of large deviation component on the prediction results, improve the prediction accuracy.
Secondly, based on meta learning wind power nonlinear combination forecasting model. The grey model? Time series model? Linear regression model and neural network model to predict the characteristics of results and the prediction of the sequence as the input element predictor, to discover and correct deviation system based pre sensor. In the predictor, determine the the base predictor weighted by the gating network function, ensure the time-varying and non negative weights. This algorithm is applied to predict wind power. The prediction results show that the algorithm prediction accuracy is higher than the combination of single prediction algorithm and the commonly used prediction algorithm.
Finally, according to the demand analysis, a wind power prediction system is developed, and the short-term prediction of wind power is realized. The test data of a wind farm show that the system is safe, reliable and operable, and it can achieve the prediction of wind power very well.

【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM614

【参考文献】