基于集成模型的光伏发电预测精度对比研究
广东工业大学, 土木与交通工程学院, 广东广州 510006
| 摘 要： | 短期光伏发电预测对改善电力系统并网的稳定性具有重要意义。为了提高短期光伏发电量预测的准确性，提出了一种基于集成模型的预测方法。首先，使用四分位法检测原始数据是否存在离群值，然后结合特征贡献度和Pearson相关系数，筛选出影响短期光伏发电量的多个特征变量。其次，构建集成模型的结构，并采用k-fold交叉验证方法对初级学习器和次级学习器中的模型进行训练。将初级学习器的结果作为次级学习器的新特征。最后，使用实际数据集验证了该方法的有效性。仿真结果表明，对离群值具有较高兼容性的初级学习器，有利于提高光伏发电量的预测精度。两种特征筛选方法的结合使用，比单一的特征提取方法，可以更有效地提取原始数据中的信息，提高模型的泛化能力。并且集成模型可以实现比单个模型更高的预测精度。 |
| 关 键 词： | 光伏发电; 太阳辐射因素; 特征工程; 集成模型; 交叉验证 |
| DOI： | 10.57237/j.res.2023.02.007 |
View:
Downloads:
收稿日期: 2023-07-15
接受日期: 2023-09-06
在线出版日期: 2023-09-08
Comparative Study on the Precision of Photovoltaic Power Generation Prediction Based on Integrated Models
School of Civil and Transportation Engineering, Guangdong University of Technology, Guangzhou 510006, China
| Abstract: | Short-term photovoltaic power generation prediction is of great significance for improving the stability of power system grid connection. In order to improve the accuracy of short-term photovoltaic power generation prediction, a prediction method is proposed based on integrated models. Firstly, the quartile method is used to detect whether there are outliers in the raw data, and then multiple feature variables that affect short-term photovoltaic power generation are selected by combining feature contribution and Pearson correlation coefficient. Secondly, construct the structure of the integrated model and use the k-fold cross validation method to train the models in the primary and secondary learners. It is a new feature of the secondary learner with the results of the primary learner. Finally, the effectiveness of this method was verified using actual datasets. The simulation results show that a primary learner with high compatibility with outliers is beneficial for improving the prediction accuracy of photovoltaic power generation. The combination of two feature selection methods can more effectively extract information from the original data and improve the generalization ability of the model compared to a single feature extraction method. Moreover, the integrated model can achieve higher prediction accuracy than a single model. |
| Keywords: | Photovoltaic Power Generation; Solar Radiation Factors; Feature Engineering; Integrated Model; Cross Validation |
参考文献
| [1] | 仲仕晶, 赵书杰. 多因素耦合对光伏发电性能影响的试验研究 [J]. 科学技术与工程, 2020, 20 (07): 2727-2732. |
| [2] | 李芬, 马年骏, 刘邦银, et al. 上海地区太阳能资源评估与散射辐射推算方法研究 [J]. 水电能源科学, 2015, 33 (05): 207-210. |
| [3] | Ma T, Yang H, Lu L. Solar photovoltaic system modeling and performance prediction [J]. Renewable and Sustainable Energy Reviews, 2014, 36. |
| [4] | Lorenz E, Hurka J, Heinemann D, et al. Irradiance Forecasting for the Power Prediction of Grid-Connected Photovoltaic Systems [J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2009, 2 (1): 2-10. |
| [5] | 张静, 褚晓红, 黄学安, et al. 一种基于加权马尔科夫链修正的SVM光伏出力预测模型 [J]. 电力系统保护与控制, 2019, 47 (19): 63-68. |
| [6] | Yin P-Y, Tsai C-C, Day R-F. PSO active learning of XGBoost and spatiotemporal data for PM2.5 sensor calibration [J]. IOP Conference Series: Earth and Environmental Science, 2019, 227 (5): 052048. |
| [7] | 曹颖超. 改进的GDBT迭代决策树分类算法及其应用 [J]. 科技视界, 2017 (12): 105+149. |
| [8] | 焦李成, 杨淑媛, 刘芳, et al. 神经网络七十年:回顾与展望 [J]. 计算机学报, 2016, 39 (08): 1697-1716. |
| [9] | 王国薇, 黄浩, 周刚, et al. 集成学习在短文本分类中的应用研究 [J]. 现代电子技术, 2019, 42 (24): 140-145. |
| [10] | 赵珂雨, 陈婉莹. 一种基于stacking集成学习的DGA域名检测方法 [J]. 数据通信, 2020 (06): 19-24. |
| [11] | 唐科, 秦敏, 赵星, et al. 基于Stacking集成学习模型的气态亚硝酸预测 [J]. 中国环境科学, 2020, 40 (02): 582-590. |
| [12] | 蒋洪迅, 田嘉, 孙彩虹. 面向PM_(2.5) 预测的递归随机森林与多层神经网络集成模型[J]. 系统工程, 2020, 38 (05): 14-24. |
| [13] | Xiao J, Li Y, Xie L, et al. A hybrid model based on selective ensemble for energy consumption forecasting in China [J]. Energy, 2018, 159. |
| [14] | 杨荣新, 孙朝云, 徐磊. 基于Stacking模型融合的光伏发电功率预测 [J]. 计算机系统应用, 2020, 29 (05): 36-45. |
| [15] | 史佳琪, 张建华. 基于多模型融合Stacking集成学习方式的负荷预测方法 [J]. 中国电机工程学报, 2019, 39 (14): 4032-4042. |
| [16] | 徐耀松, 段彦强, 王雨虹, et al. 基于相似日选择与改进Stacking集成学习的短期负荷预测 [J]. 传感技术学报, 2020, 33 (04): 537-545. |
| [17] | Mallor F, León T, Boeck L D, et al. A method for detecting malfunctions in PV solar panels based on electricity production monitoring [J]. Solar Energy, 2017, 153. |
| [18] | 俞娜燕, 李向超, 费科, et al. 基于改进高斯过程回归的光伏电站监测数据修复研究 [J]. 自动化与仪器仪表, 2020 (05): 56-58, 62. |
| [19] | 彭曙蓉, 郑国栋, 黄士峻, et al. 基于XGBoost算法融合多特征短期光伏发电量预测 [J]. 电测与仪表, 2020, 57 (24): 76-83. |
| [20] | 邓威, 郭钇秀, 李勇, et al. 基于特征选择和Stacking集成学习的配电网网损预测 [J]. 电力系统保护与控制, 2020, 48 (15): 108-115. |
| [21] | 崔书华, 李果, 刘军虎, et al. 基于偏度与峰度的数据质量评估 [J]. 弹箭与制导学报, 2015, 35 (06): 98-100, 105. |
