风电场风速预测组合模型研究

发布时间：2018-02-01 03:31

本文关键词： 风力发电风速预测时间序列 Elman神经网络 BP神经网络组合模型　出处：《华北电力大学》2014年硕士论文　论文类型：学位论文

【摘要】：风能以其可再生、无污染的特性越来越受到人们的关注。但由于风速的波动性和随机性,风机出力很不稳定。随着风力发电在电网中所占的比重逐渐增加,其对电力系统的安全稳定运行一定会造成一些不利影响。风电并网及电力调度中,风速预测的准确性可以提供非常重要的参考,大大的消除风速波动对电网的影响。鉴于这些原因,对风电场风速进行预测研究是非常有意义的。本文针对风速数据的非线性特性,利用改进的Elman神经网络修正ARIMA模型预测结果的方法,运用时间序列与神经网络的组合模型对短期风速预测进行研究。先利用ARIMA模型对风速进行预测,其线性规律信息包含在时间序列预测结果中,非线性规律包含在预测误差中。再将ARIMA模型的预测误差及历史风速一阶差分序列作为改进的Elman神经网络输入变量,将ARIMA模型的风速预测误差作为输出变量。最后将ARIMA模型预测结果与Elman神经网络的误差预测结果叠加,得到最终修正后的预测风速。为证明方法的有效性,分别与单一ARIMA模型、ARIMA-BP神经网络组合模型进行对比,对实际历史风速数据进行仿真预测。经验证,利用改进Elman神经网络修正ARIMA模型预测误差,比单一ARIMA模型能够更好的减小预测滞后性,提高预测精度、减小预测误差；比ARIMA-BP神经网络组合模型训练速度提高了30%以上。通过以上对风速预测问题的研究,运用组合模型进行了较为深入的探讨,并进行了数据处理及仿真,可以发现ARIMA-Elman神经网络组合模型比单一模型有更大的优越性,为解决实际工程问题提供了一定的参考。
[Abstract]:Wind energy has attracted more and more attention due to its renewable and pollution-free characteristics. However, due to the volatility and randomness of wind speed, the wind power output is very unstable. With the increasing proportion of wind power generation in power grid, wind power generation is becoming more and more important. It will inevitably cause some adverse effects on the safe and stable operation of power system. The accuracy of wind speed prediction can provide a very important reference in wind power grid connection and power dispatching. The influence of wind speed fluctuation on power grid is greatly eliminated. For these reasons, it is very meaningful to predict the wind speed of wind farm. According to the nonlinear characteristics of wind speed data, the improved Elman neural network is used to modify the prediction results of ARIMA model. Using the combined model of time series and neural network, the short-term wind speed prediction is studied. First, the ARIMA model is used to predict the wind speed, and the linear law information is included in the prediction results of time series. The nonlinear law is included in the prediction error, and then the prediction error of ARIMA model and the first order difference sequence of historical wind speed are taken as the input variables of the improved Elman neural network. The wind speed prediction error of ARIMA model is taken as the output variable. Finally, the prediction result of ARIMA model is superposed with the error prediction result of Elman neural network, and the final modified predicted wind speed is obtained. In order to prove the effectiveness of the method, the actual historical wind speed data are simulated and forecasted by comparing with the single ARIMA model ARIMA-BP neural network combination model. The improved Elman neural network is used to correct the prediction error of ARIMA model, which can reduce the prediction lag, improve the prediction accuracy and reduce the prediction error better than the single ARIMA model. The training speed of the combined model is more than 30% higher than that of the ARIMA-BP neural network. Through the above research on wind speed prediction, the combined model is used to conduct a more in-depth discussion, and the data processing and simulation are carried out. It can be found that the combined model of ARIMA-Elman neural network has more advantages than the single model, which provides a certain reference for solving practical engineering problems.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM614

【参考文献】