基于k-means聚类和变分位鲁棒极限学习机的短期负荷预测方法
|
摘要
随着售电侧市场的逐步开放,集中式的供售电模式被打破,为获取更精确的区域短期负荷预测值,提出一种基于k-means聚类和变分位鲁棒极限学习机的短期负荷预测方法。首先利用传统的k-means聚类算法对历史电力负荷数据进行负荷模式的提取,获取相同用电行为的用户负荷曲线。然后采用变分位鲁棒极限学习机对不同类负荷曲线分别建立预测模型,最后叠加单个的预测值形成最终的预测结果。通过设定不同的分位值来模拟不同的预测场景,以此得到所有可能性的预测值,即实现变分位-多场景的VQR-ORELM灵活预测。为验证所提方法的有效性,采用2个实际案例进行仿真分析。结果表明,相对于支持向量机、BP神经网络、极限学习机模型、鲁棒极限学习机模型,所提模型在聚类前后预测精度始终最高,进一步验证了所提方法的优越性和灵活性。通过k-means聚类后,所有模型预测性能都有较大提高。
With the development of power sales market, the centralized electricity trading pattern is broken up. In order to obtain more accurate results of short-term regional load forecasting, this paper proposes a short-term load forecasting method by combining kmeans clustering method with varied quantile outlier robust extreme learning machine(VQR-ORELM). Firstly, the k-means clustering algorithm is used to extract different load patterns from historical load data. Then VQR-ORELM is used to establish the forecasting model for all load patterns respectively, obtaining final load prediction result by means of the superposition of individual forecasting results.The different prediction scenarios are simulated by setting different quantile to obtain the predicted values of all possibilities, thus realizing the flexible forecasting of varied quantile and varied scenes. To verify the effectiveness of the proposed method, the simulation analysis on two cases is carried out. The results show that the proposed model outperforms all other models in predicting the short-term load with and without clustering all the time, such as support vector machine, BP neural network, extreme learning machine and outlier robust extreme learning machine, further verifying the superiority and flexibility of the proposed method. In the meanwhile, there is a noticeable improvement in the prediction accuracy of all these models after k-means clustering.
With the development of power sales market, the centralized electricity trading pattern is broken up. In order to obtain more accurate results of short-term regional load forecasting, this paper proposes a short-term load forecasting method by combining kmeans clustering method with varied quantile outlier robust extreme learning machine(VQR-ORELM). Firstly, the k-means clustering algorithm is used to extract different load patterns from historical load data. Then VQR-ORELM is used to establish the forecasting model for all load patterns respectively, obtaining final load prediction result by means of the superposition of individual forecasting results.The different prediction scenarios are simulated by setting different quantile to obtain the predicted values of all possibilities, thus realizing the flexible forecasting of varied quantile and varied scenes. To verify the effectiveness of the proposed method, the simulation analysis on two cases is carried out. The results show that the proposed model outperforms all other models in predicting the short-term load with and without clustering all the time, such as support vector machine, BP neural network, extreme learning machine and outlier robust extreme learning machine, further verifying the superiority and flexibility of the proposed method. In the meanwhile, there is a noticeable improvement in the prediction accuracy of all these models after k-means clustering.
引文
[1]刘会家,管鑫,陈波,等.考虑主动需求的主动配电网负荷预测[J].电力系统保护与控制,2018,46(10):68-74.LIU Huijia, GUAN Xin, CHEN Bo, et al.Load forecasting for active distribution network in the presence of active demand[J].Power System Protection and Control, 2018,46(10):68-74.
[2]庞鹏.电力市场化改革背景下电力需求响应机制与支撑技术[J].广东电力,2016,29(1):70-78.PANG Peng.Electric power demand response mechanism and support technology under electric power market reform[J]. Guangdong Electric Power,2016,29(1):70-78.
[3]白杨,李昂,夏清.新形势下电力市场营销模式与新型电价体系[J].电力系统保护与控制,2016,44(5):10-16.BAI Yang, LI Ang, XIA Qing.Electricity business marketing modes in the new environment and new electricity pricing systems[J].Power System Protection and Control,2016, V44(5):10-16.
[4]王友春,文闪闪,秦跃进,等.基于ARIMA-SVM模型的微电网短期负荷组合预测研究[J].陕西电力,2014,42(3):19-23.WANG Youchun, WEN Shanshan,QIN Yuejin,et al.Study on microgrid short-term load combination forecasting based on ARIMA-SVM model[J].Shaanxi Electric Power,2014,42(3):19-23.
[5]司徒友,李刚强,张水平.基于小波分解和深度信念网络的短期风速预测[J].广东电力,2017,30(4):25-30.SITU You, LI Gangqiang, ZHANG Shuiping. Short-term wind speed forecasting based on wavelet decomposition and deep belief network[J]. Guangdong Electric Power,2017,30(4):25-30.
[6]朱陶业,晏小兵,李应求.ARIMA模型在广西短期电力负荷预测中的应用[J].电力科学与技术学报,2000, 15(2):20-22.ZHU Taoye, YAN Xiaobing, LI Yingqiu. The ARIMA model in the forecasting of power load in guangxi[J].Journal of Electric Power Science and Technology, 2000,15(2):20-22.
[7]孟安波,胡函武,刘向东.基于纵横交叉算法优化神经网络的负荷预测模型[J].电力系统保护与控制,2016, V44(7):102-106.MENG Anbo,HU Hanwu,LIU Xiangdong.Short-term load forecasting using neural network based on wavelets and crisscross optimization algorithm[J].Power System Protection and Control,2016,V44(7):102-106.
[8]胡杨,常鲜戎.基于改进EMD-PSVM的短期负荷预测[J].智慧电力,2016, 44(3):29-33.HU Yang, CHANG Xianrong. Short-term load forecasting based on improved EMD-PSVM[J]. Smart Power, 2016,44(3):29-33.
[9]王宁,谢敏,邓佳梁,等.基于支持向量机回归组合模型的中长期降温负荷预测[J].电力系统保护与控制,2016,V44(3):92-97.WANG Ning, XIE Min, DENG Jialiang, et al.Mid-long term temperature-lowering load forecasting based on combination of support vector machine and multiple regression[J].Power System Protection and Control, 2016,V44(3):92-97.
[10]毛力,王运涛,刘兴阳等.基于改进极限学习机的短期电力负荷预测方法[J].电力系统保护与控制,2012(20):140-144.MAO Li, WANG Yuntao, LIU Xingyang, elt all. Shortterm power load forecasting method based on improved extreme learning machine[J]. Power Systerm Protectin and Control, 2012(20):140-144.
[11]董朕,殷豪,孟安波.基于混合算法优化的短期风功率预测[J].智慧电力,2017,45(11):24-30.DONG Zhen, YIN Hao, MENG Anbo.Short-term wind power forecasting based on kernel extreme learning machine optimized by hybrid algorithm[J].Smart Power,2017,45(11):24-30.
[12] ZHANG K, LUO M. Outlier-robust extreme learning machine for regression problems[J]. Neurocomputing,2015, 151:1519-1527.
[13] SATTADI S, SHAMSHIRBAND S, ALIZAMIR M, et al.Extreme learning machine for prediction of heat load in district heating systems[J]. Energy&Buildings, 2016,122:222-227.
[14] WANG Y. Kernel-based clustering for short-term load forecasting[C]. International Conference on Advances in Power System Control, Operation&Management, IET,2017.
[15]蔡秋娜,刘思捷,陆秋瑜.基于GMM聚类和SVM的用户负荷行业分类辨识方法[J].广东电力,2017,30(12):91-96.CAI Qiuna, LIU Sijie, LU Qiuyu. Identification method for user industry classification based on GMM clustering and SVM[J]. Guangdong Electric Power, 2017, 30(12):91-96.
[16]牟宏,王春义,顾洁,等.基于支持向量-分位数回归的短期负荷预测及风险分析[J].电力科学与技术学报,2016, 31(3):51-56.MOU Hong, WANG Chunyi, GU Jie, et al. Short-term load forecasting and risk analysis based on support vector-quantile regression[J]. Journal of Electric Power Science and Technology,2016, 31(3):51-56.
[17] GAN D, WANG Y, YANG S, et al. Embedding based quantile regression neural network for probabilistic load forecasting[J]. Journal of Modern Power Systems&Clean Energy, 2018, 6(2):1-11.
[18] ERIC W. Commercial and residential hourly load profiles for all TY3 location in the USA. https://openei.org/datasets/dataset/commercial-and-residential-hourly-loadprofiles-all-tmy3-locations
[2]庞鹏.电力市场化改革背景下电力需求响应机制与支撑技术[J].广东电力,2016,29(1):70-78.PANG Peng.Electric power demand response mechanism and support technology under electric power market reform[J]. Guangdong Electric Power,2016,29(1):70-78.
[3]白杨,李昂,夏清.新形势下电力市场营销模式与新型电价体系[J].电力系统保护与控制,2016,44(5):10-16.BAI Yang, LI Ang, XIA Qing.Electricity business marketing modes in the new environment and new electricity pricing systems[J].Power System Protection and Control,2016, V44(5):10-16.
[4]王友春,文闪闪,秦跃进,等.基于ARIMA-SVM模型的微电网短期负荷组合预测研究[J].陕西电力,2014,42(3):19-23.WANG Youchun, WEN Shanshan,QIN Yuejin,et al.Study on microgrid short-term load combination forecasting based on ARIMA-SVM model[J].Shaanxi Electric Power,2014,42(3):19-23.
[5]司徒友,李刚强,张水平.基于小波分解和深度信念网络的短期风速预测[J].广东电力,2017,30(4):25-30.SITU You, LI Gangqiang, ZHANG Shuiping. Short-term wind speed forecasting based on wavelet decomposition and deep belief network[J]. Guangdong Electric Power,2017,30(4):25-30.
[6]朱陶业,晏小兵,李应求.ARIMA模型在广西短期电力负荷预测中的应用[J].电力科学与技术学报,2000, 15(2):20-22.ZHU Taoye, YAN Xiaobing, LI Yingqiu. The ARIMA model in the forecasting of power load in guangxi[J].Journal of Electric Power Science and Technology, 2000,15(2):20-22.
[7]孟安波,胡函武,刘向东.基于纵横交叉算法优化神经网络的负荷预测模型[J].电力系统保护与控制,2016, V44(7):102-106.MENG Anbo,HU Hanwu,LIU Xiangdong.Short-term load forecasting using neural network based on wavelets and crisscross optimization algorithm[J].Power System Protection and Control,2016,V44(7):102-106.
[8]胡杨,常鲜戎.基于改进EMD-PSVM的短期负荷预测[J].智慧电力,2016, 44(3):29-33.HU Yang, CHANG Xianrong. Short-term load forecasting based on improved EMD-PSVM[J]. Smart Power, 2016,44(3):29-33.
[9]王宁,谢敏,邓佳梁,等.基于支持向量机回归组合模型的中长期降温负荷预测[J].电力系统保护与控制,2016,V44(3):92-97.WANG Ning, XIE Min, DENG Jialiang, et al.Mid-long term temperature-lowering load forecasting based on combination of support vector machine and multiple regression[J].Power System Protection and Control, 2016,V44(3):92-97.
[10]毛力,王运涛,刘兴阳等.基于改进极限学习机的短期电力负荷预测方法[J].电力系统保护与控制,2012(20):140-144.MAO Li, WANG Yuntao, LIU Xingyang, elt all. Shortterm power load forecasting method based on improved extreme learning machine[J]. Power Systerm Protectin and Control, 2012(20):140-144.
[11]董朕,殷豪,孟安波.基于混合算法优化的短期风功率预测[J].智慧电力,2017,45(11):24-30.DONG Zhen, YIN Hao, MENG Anbo.Short-term wind power forecasting based on kernel extreme learning machine optimized by hybrid algorithm[J].Smart Power,2017,45(11):24-30.
[12] ZHANG K, LUO M. Outlier-robust extreme learning machine for regression problems[J]. Neurocomputing,2015, 151:1519-1527.
[13] SATTADI S, SHAMSHIRBAND S, ALIZAMIR M, et al.Extreme learning machine for prediction of heat load in district heating systems[J]. Energy&Buildings, 2016,122:222-227.
[14] WANG Y. Kernel-based clustering for short-term load forecasting[C]. International Conference on Advances in Power System Control, Operation&Management, IET,2017.
[15]蔡秋娜,刘思捷,陆秋瑜.基于GMM聚类和SVM的用户负荷行业分类辨识方法[J].广东电力,2017,30(12):91-96.CAI Qiuna, LIU Sijie, LU Qiuyu. Identification method for user industry classification based on GMM clustering and SVM[J]. Guangdong Electric Power, 2017, 30(12):91-96.
[16]牟宏,王春义,顾洁,等.基于支持向量-分位数回归的短期负荷预测及风险分析[J].电力科学与技术学报,2016, 31(3):51-56.MOU Hong, WANG Chunyi, GU Jie, et al. Short-term load forecasting and risk analysis based on support vector-quantile regression[J]. Journal of Electric Power Science and Technology,2016, 31(3):51-56.
[17] GAN D, WANG Y, YANG S, et al. Embedding based quantile regression neural network for probabilistic load forecasting[J]. Journal of Modern Power Systems&Clean Energy, 2018, 6(2):1-11.
[18] ERIC W. Commercial and residential hourly load profiles for all TY3 location in the USA. https://openei.org/datasets/dataset/commercial-and-residential-hourly-loadprofiles-all-tmy3-locations