| 潘国兵,余方吉,陈坚,欧阳静.基于Word2vec-LSTM与聚类修正的海上风电出力预测方法[J].高技术通讯(中文),2025,35(1):102~112 |
| 基于Word2vec-LSTM与聚类修正的海上风电出力预测方法 |
| A method for predicting offshore wind power output based on Word2vec-LSTM and cluster correction |
| |
| DOI：10. 3772 / j. issn. 1002-0470. 2025. 01. 011 |
| 中文关键词: 海上风电； 功率预测； 特征提取； 聚类修正 |
| 英文关键词: offshore wind power, power prediction, feature extraction, cluster correction |
| 基金项目: |
| 作者 | 单位 | | 潘国兵 | （浙江工业大学机械工程学院杭州 310012） | | 余方吉 | | | 陈坚 | | | 欧阳静 | |
|
| 摘要点击次数: 168 |
| 全文下载次数: 156 |
| 中文摘要: |
| 针对目前海上风电出力预测方法精度较低的问题，提出一种基于词向量化和长短期记忆网络（word to vector long short-term memory，Word2vec-LSTM）与聚类修正的海上风电出力预测方法。对Word2vec方法进行改进来提取时间序列数据特征，实现数据信息的高效利用；在长短期记忆神经网络的预测模型基础上，研究了一种基于k-shape聚类结果的预测结果修正算法，对预测结果距离聚类中心超过阈值的数值判定为预测误差偏大的数据并向簇中心进行修正。最后，基于江苏某海上风电场的真实数据进行测试，结果表明，基于Word2vec-LSTM与聚类修正的海上风电出力预测方法的平均绝对误差（mean absolute error，MAE）和均方根误差（root mean square error，RMSE）达到5.04和5.42，相比传统LSTM预测模型的误差平均降低了11.10%和12.25%，为海上风电并网与电网调控提供了技术支持。 |
| 英文摘要: |
| Aiming at the problem that the accuracy of mainstream prediction methods is low, a prediction method of offshore wind power output based on word to vector long short-term memory (Word2vec-LSTM) and cluster correction is proposed. The Word2vec method is improved to extract features from time series data and efficient utilization of data information is achieved. Based on the prediction model of long short-term memory neural network, a prediction result correction algorithm based on k-shape cluster results is studied. Finally, based on real data from an offshore wind farm in Jiangsu, the results showed that the average mean absolute error（MAE）and root mean square error（RMSE）of the Word2vec-LSTM and cluster correction based offshore wind power output prediction method reached 5.04 and 5.42, respectively. Compared with traditional LSTM prediction models, the average error decreases by 11.10% and 12.25%, providing technical support for offshore wind power grid connection and grid regulation. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
