基于深度置信网络的短期风电功率预测

为解决海量数据用作预测模型训练样本导致信息冗杂的问题,提出一种基于深度置信网络的短期风电功率预测方法.该方法首先使用历史数据作为训练样本,通过深度置信网络无监督学习提取出其相应特征,随后采用K均值算法对提取出的特征进行聚类分析,将历史数据分作几类,并通过判别分析确定待测日所属类别,以该类别所属的历史数据对设置了误差反馈...

SHORT-TERM WIND POWER PREDICTION BASED ON DEEP BELIEF NETWORK

The current massive wind power historical data provides a very good foundation for the use of deep learning for wind power prediction. In order to solve the problem of information redundancy due to the massive data used as the training sample of the prediction model, a short-term wind power prediction method based on Deep Belief Network is proposed. The method first uses historical data as training samples, extracts its corresponding features through deep belief network unsupervised learning, and then uses the K-means algorithm to perform cluster analysis on the extracted features. The historical data is divided into several categories, and the category of the days to be measured is determined by discriminant analysis. Using the historical data belong to this category to supervised train the Deep Belief Network with the error feedback layer, and inputting the weather information of the day to be measured into the trained Deep Belief Network model, the predicted power is finally obtained. The effectiveness of the proposed method is verified by an example of the actual operation data of a wind farm in Yunnan province.