为进一步提高超短期光伏发电功率预测的精度,提出一种基于强化学习的多模型融合光伏发电功率预测方法。首先,采用局部离群因子算法检测、剔除异常点,并用多层感知机回归算法进行修补,解决数据异常问题;然后,将数据分为训练集、验证集与测试集,在训练集中训练支持向量机回归（SVR）、多元线性回归（MLR）、贝叶斯岭回归（BRR）、卷积-长短期记忆（CNN-LSTM）与基于粒子群算法优化的门控循环单元（PSO-GRU）模型,并在验证集对训练得到的模型进行验证,分别选出最佳的模型作为子模型;最后,在测试集中使用5个子模型进行预测,并将各预测结果用强化学习的方法进行融合,将融合值作为最终的预测结果。实验结果表明,该预测方法的平均绝对误差、均方误差、均方根误差与相对误差相比单模型方法以及其他传统的融合方法均有显著降低,验证了该方法的有效性。
Abstract
In order to further improve the accuracy of ultra-short-term photovoltaic power prediction, a multi-model fusion photovoltaic power prediction method based on reinforcement learning is proposed. Firstly, the local outlier factor（LOF） algorithm is used to detect and remove outliers, and a multilayer perceptron regression algorithm is employed to correct the data anomalies. Then, the data is divided into training, validation, and testing sets. In the training set, models such as support vector Regression （SVR）, multiple linear regression（MLR）, Bayesian ridge regression（BRR）, convolutional neural network-long short term memory （CNN-LSTM） and particle swarm optimization-gated recurrent unit （PSO-GRU） are trained. These trained models are validated on the validation set to select the best-performing models as sub-models. Finally, in the testing set, the five sub-models are used for forecasting, and their predictions are fused using a reinforcement learning method. The fusion value is taken as the final prediction result. Experimental results show that the proposed method significantly reduces the mean absolute error, mean squared error, root mean squared error, and relative error compared to single-model methods and other traditional fusion methods, verifying the effectiveness of this approach.
关键词
异常检测 /
机器学习 /
强化学习 /
多模型融合 /
光伏发电功率预测
Key words
anomaly detection /
machine learning /
reinforcement learning /
multi-model fusion /
photovoltaic power prediction
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 贾凌云, 云斯宁, 赵泽妮, 等. 神经网络短期光伏发电预测的应用研究进展[J]. 太阳能学报, 2022, 43(12): 88-97.
JIA L Y, YUN S N, ZHAO Z N, et al.Recent progress of short-term forecasting of photovoltaic generation based on artificial neural networks[J]. Acta energiae solaris sinica, 2022, 43(12): 88-97.
[2] 程泽, 李思宇, 韩丽洁, 等. 基于数据挖掘的光伏阵列发电预测方法研究[J]. 太阳能学报, 2017, 38(3): 726-733.
CHENG Z, LI S Y, HAN L J, et al.PV power generation forecast based on data mining method[J]. Acta energiae solaris sinica, 2017, 38(3): 726-733.
[3] 张家安, 郝峰, 董存, 等. 基于两阶段不确定性量化的光伏发电超短期功率预测[J]. 太阳能学报, 2023, 44(1): 69-77.
ZHANG J A, HAO F, DONG C, et al.Ultra-short-term power forecasting of photovoltaic power generation based on two-stage uncertainty quantization[J]. Acta energiae solaris sinica, 2023, 44(1): 69-77.
[4] 刘澄, 王辉, 李天慧, 等. 分布式新能源发电对配电网电压影响研究[J]. 可再生能源, 2019, 37(10): 1465-1471.
LIU C, WANG H, LI T H, et al.Study on influence of distributed new energy generation on distribution network voltage[J]. Renewable energy resources, 2019, 37(10): 1465-1471.
[5] 刘国海, 孙文卿, 吴振飞, 等. 基于Attention-GRU的短期光伏发电功率预测[J]. 太阳能学报, 2022, 43(2): 226-232.
LIU G H, SUN W Q, WU Z F, et al.Short-term photovoltaic power forecasting based on Attention-GRU model[J]. Acta energiae solaris sinica, 2022, 43(2): 226-232.
[6] 刘纯, 范高锋, 王伟胜, 等. 风电场输出功率的组合预测模型[J]. 电网技术, 2009, 33(13): 74-79.
LIU C, FAN G F, WANG W S, et al.A combination forecasting model for wind farm output power[J]. Power system technology, 2009, 33(13): 74-79.
[7] 忻俊杰, 徐良, 李永杰, 等. 基于集成学习的多模型融合光伏发电功率动态修正预测[J]. 传感器与微系统, 2021, 40(4): 117-121.
XIN J J, XU L, LI Y J, et al.Multi-model fusion dynamically correct prediction of PV power generation based on ensemble learning[J]. Transducer and microsystem technologies, 2021, 40(4): 117-121.
[8] 庄家懿, 杨国华, 郑豪丰, 等. 基于多模型融合的CNN-LSTM-XGBoost短期电力负荷预测方法[J]. 中国电力, 2021, 54(5): 46-55.
ZHUANG J Y, YANG G H, ZHENG H F, et al.Short-term load forecasting method based on multi-model fusion using CNN-LSTM-XGBoost framework[J]. Electric power, 2021, 54(5): 46-55.
[9] 杨锡运, 刘欢, 张彬, 等. 基于熵权法的光伏输出功率组合预测模型[J]. 太阳能学报, 2014, 35(5): 744-749.
YANG X Y, LIU H, ZHANG B, et al.A combination method for photovoltaic power forecasting based on entropy weight method[J]. Acta energiae solaris sinica, 2014, 35(5): 744-749.
[10] 邱瑞东, 何山, 董宁, 等. 基于LSTM-LGB模型的光伏电站辐照强度预测[J]. 安徽大学学报(自然科学版), 2021, 45(3): 66-71.
QIU R D, HE S, DONG N, et al.Irradiation intensity prediction of photovoltaic power station based on LSTM-LGB model[J]. Journal of Anhui University (natural science edition), 2021, 45(3): 66-71.
[11] SUTTON R S, BARTO A G.Reinforcement learning: an introduction[J]. IEEE transactions on neural networks, 1998, 9(5): 1054.
[12] SALLAB A E, ABDOU M, PEROT E, et al.Deep reinforcement learning framework for autonomous driving[J]. Electronic imaging, 2017, 29(19): 70-76.
[13] LI Z J, HUANG B, AJOUDANI A, et al.Asymmetric bimanual control of dual-arm exoskeletons for human-cooperative manipulations[J]. IEEE transactions on robotics, 2018, 34(1): 264-271.
[14] BREUNIG M M, KRIEGEL H P, NG R T, et al.OPTICS-OF: identifying local outliers[C]//European Conference on Principles of Data Mining and Knowledge Discovery. Berlin, Heidelberg: Springer Berlin Heidelberg, 1999: 262-270.
[15] ORTUÑO M F, CONEJERO W, MORENO F, et al. Could trunk diameter sensors be used in woody crops for irrigation scheduling? A review of current knowledge and future perspectives[J]. Agricultural water management, 2010, 97(1): 1-11.
[16] PAN H, YOU X M, LIU S, et al.Pearson correlation coefficient-based pheromone refactoring mechanism for multi-colony ant colony optimization[J]. Applied intelligence, 2021, 51(2): 752-774.
[17] 宋坤骏, 丁建明, 林建辉. 基于改进Fisher准则、VMD、距离相关系数和核极限学习机的轴承故障诊断[J]. 铁道机车车辆, 2018, 38(3): 22-28.
SONG K J, DING J M, LIN J H.Rolling bearing fault diagnosis with modified Fisher criterion, VMD, distance correlation coefficients and kernel extreme learning machine[J]. Railway locomotive & car, 2018, 38(3): 22-28.
[18] LIU H P, SUN F C, ZHANG X Y.Robotic material perception using active multimodal fusion[J]. IEEE transactions on industrial electronics, 2019, 66(12): 9878-9886.
[19] LIANG X Y, DU X S, WANG G L, et al.A deep reinforcement learning network for traffic light cycle control[J]. IEEE transactions on vehicular technology, 2019, 68(2): 1243-1253.
[20] ZHOU T L, CHEN M, YANG C G, et al.Data fusion using Bayesian theory and reinforcement learning method[J]. Science China information sciences, 2020, 63(7): 170209.
[21] RISSANEN J J.Fisher information and stochastic complexity[J]. IEEE transactions on information theory, 1996, 42(1): 40-47.
[22] 陈文颖, 林永君, 杨春来, 等. 基于SVM预测模型的光伏发电系统MPPT研究[J]. 太阳能学报, 2013, 34(2): 245-250.
CHEN W Y, LIN Y J, YANG C L, et al.Research on MPPT of PV systems based on SVM model[J]. Acta energiae solaris sinica, 2013, 34(2): 245-250.
[23] 余威, 韩艳彬, 种永刚, 等. 基于加权几何平均迭代的改进BESO法[J]. 南京航空航天大学学报, 2020, 52(3): 416-421.
YU W, HAN Y B, CHONG Y G, et al.Improved BESO method based on weighted geometric mean iteration[J]. Journal of Nanjing University of Aeronautics & Astronautics, 2020, 52(3): 416-421.
基金
浙江省自然科学基金（LR20F030004）
PDF(1938 KB)
