| 引用本文: | 李永飞,张 耀,林 帆,等.基于气候特征分析及改进XGBoost算法的中长期光伏电站发电量预测方法[J].电力系统保护与控制,2024,52(11):84-92. |
| LI Yongfei,ZHANG Yao,LIN Fan,et al.Medium- and long-term power generation forecast based on climate characterisation and an improved XGBoost algorithm for photovoltaic power plants[J].Power System Protection and Control,2024,52(11):84-92 |
|
| 摘要: |
| 光伏发电在能源结构中的重要性不断凸显，而提高光伏发电量预测的准确性成为当前研究的关键问题。针对中长期光伏发电量预测问题，提出一个综合利用气候预测数据的中长期光伏发电量预测方法。首先，在基于气候预测数据的发电量预测框架中，根据气候预测数据特点和预测周期划分多重子模型以充分利用气候预测数据信息。其次，在进行数据预处理后，通过对气候特征衍生与交叉、特征筛选和选择，充分挖掘气候特征的高价值信息。然后，采取一种两重多阶段超参数寻优策略，对极端梯度增强(extreme gradient boosting, XGBoost)超参数进行调整以优化预测模型。最后，在真实光伏发电量数据上，以MAPE为标准评估预测水平，验证所提中长期光伏发电量预测方法的有效性。相关实验结果表明该方法可以有效提高光伏发电量预测精度。 |
| 关键词: 气候预测数据 XGBoost 中长期预测 光伏发电量预测 特征工程 |
| DOI：10.19783/j.cnki.pspc.231605 |
| 分类号: |
| 基金项目:国家重点研发计划项目资助(2022YFB2403500) |
|
| Medium- and long-term power generation forecast based on climate characterisation and an improved XGBoost algorithm for photovoltaic power plants |
|
LI Yongfei, ZHANG Yao, LIN Fan, ZHAO Yingjie, CHEN Yuxuan, ZHAO Hanting, HUO Wei
|
|
Shaanxi Key Laboratory of Smart Grid (School of Electrical Engineering, Xi’an Jiaotong University), Xi’an 710049, China
|
| Abstract: |
| The importance of photovoltaic (PV) power in the energy structure is constantly highlighted, and improving the accuracy of PV power prediction has become a key issue in current research. To address the PV prediction problem, a medium- and long-term PV power generation prediction method using climate prediction data is proposed. First, multiple sub-models are divided according to the characteristics of climate prediction data and prediction period to make full use of the data. After data pre-processing, the high-value information of climate features is fully exploited through the derivation and crossover and selection of climate features. A two-fold multi-stage hyper-parameter optimization strategy is adopted to optimize the prediction model by adjusting the XGBoost hyper-parameters. Using real photovoltaic generation data, the prediction level is evaluated by MAPE, and the effectiveness of the proposed medium- and long-term PV power generation prediction method is verified by experiment. The results show that the method can effectively improve the prediction accuracy of PV power generation. |
| Key words: climate prediction data XGBoost medium- and long-term forecasts photovoltaic power generation forecasts feature engineering |
