| 王力成* **,王子非*,邓宝华*,凌锋*,张有兵*.基于知识-数据混合驱动的综合能源系统多元负荷预测方法[J].高技术通讯(中文),2023,33(8):791~801 |
| 基于知识-数据混合驱动的综合能源系统多元负荷预测方法 |
| Multivariate load forecasting method of integrated energy system based on hybrid knowledge-data driven |
| |
| DOI：10. 3772/ j. issn. 1002-0470. 2023. 08. 002 |
| 中文关键词: 综合能源系统(IES)； 知识-数据混合驱动； 能量耦合特性； 多元负荷预测； 随机森林算法； Dropout技术 |
| 英文关键词: integrated energy system (IES), hybrid knowledge-data driven, energy coupling characteristics, multivariate load forecasting, random forest algorithm, Dropout technology |
| 基金项目: |
| 作者 | 单位 | | 王力成* ** | （*浙江工业大学信息工程学院杭州 310023）
（**之江实验室人工智能研究院杭州 311121） | | 王子非* | | | 邓宝华* | | | 凌锋* | | | 张有兵* | |
|
| 摘要点击次数: 1229 |
| 全文下载次数: 1269 |
| 中文摘要: |
| 当前的综合能源系统(IES)负荷预测方法几乎都是单一的数据驱动方法，忽略了IES中的能量耦合关系。此外，在现有的研究中，数据驱动方法的训练数据主要集中于历史负荷、气象等影响因素，较少考虑可再生能源出力以及不同能源供应给IES负荷预测结果带来的影响。针对上述问题，本文提出一种知识-数据混合驱动的IES多元负荷预测方法。该方法首先通过解析模型对IES中的能量耦合特性知识进行描述，并利用该知识模型对原始样本数据进行重构。然后将重构后的新样本数据作为数据驱动模型的训练样本，并使用基于随机森林算法的特征选择方法和Dropout技术提高模型的泛化能力。最后采用某IES工业园区的实际数据对本文所提方法的有效性进行了验证。仿真结果表明，该方法相较于传统单一数据驱动模型具有更好的预测效果和较高的可靠性。 |
| 英文摘要: |
| The current integrated energy system (IES) load forecasting methods are almost all single data-driven methods, ignoring the energy coupling relationship in IES. In addition, in the existing research, the training data of the data-driven method is mainly concentrated on historical load, weather and other influencing factors, and less consideration is given to the volatility of renewable energy output and the impact of various energy supplies on the IES load forecast results. In response to the above problems, this paper proposes a hybrid knowledge-data driven comprehensive energy system multiple load forecasting method. This method first describes the knowledge of energy coupling characteristics in IES through an analytical model, and uses the knowledge model to reconstruct the original sample data. Then use the reconstructed new sample data as the training sample of the data-driven model, and use the feature selection method based on the random forest algorithm and the Dropout technology to improve the generalization ability of the model. Finally, the effectiveness of the method is verified by the actual data of an IES industrial park. The simulation results show that this method has better prediction effect and higher reliability than the traditional single data-driven model. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
