﻿ 基于BP神经网络的北京夏季日最大电力负荷预测方法
 气候与环境研究  2019, Vol. 24 Issue (1): 135-142 PDF

1 中国气象局北京城市气象研究院, 北京 100089;
2 北京市气象服务中心, 北京 100089;
3 京津冀环境气象预报预警中心, 北京 100089

A Method for Prediction of Daily Maximum Electric Loads in the Summer in Beijing Based on the BP Neural Network
LI Chen1,2, GUO Wenli2, WU Jin3, JIN Chenxi2
1 Institute of Urban Meteorology, China Meteorological Administration, Beijing 100089;
2 Beijing Meteorological Service Center, Beijing 100089;
3 Environment Meteorology Forecast Center of Beijing-Tianjin-Hebei, Beijing 100089
Abstract: Based on daily maximum electric loads and meteorological data in the summer (June-August) from 2006 to 2017 in Beijing, the relationship between electric load and meteorological factors is diagnosed. Using the BP (Back Propagation) neural network algorithm, two maximum electric power load prediction models are established and evaluated. The results indicate that (1) the basic electric load on weekends in Beijing in the summer is much less than that in working days, which should be distinguished when being removed; (2) the influence of meteorological factors on meteorological load has cumulative effect, and the correlation between them is the highest for two days of accumulation; (3) taking the actual situation into account, two different daily maximum electric load forecasting models are established based on different independent variables. Comparing the prediction results with actual data, both of the forecasting models show good prediction performance that can meet the actual demand of the power sector. The forecasting model with meteorological load of the previous day as an independent variable shows better prediction effect.
Keywords: BP neural network     Daily maximum electric load     Cumulative meteorological factors     Prediction model
﻿
1 引言

2 资料与方法 2.1 资料来源

2.2 研究方法 2.2.1 资料的处理

 ${I_{{\rm{TH}}}} = (1.8T + 32) - (0.55 - 0.55{H_{\rm{R}}})(1.8T - 26),$ (1)

2.2.2 BP神经网络基本原理

3 夏季最大电力负荷与气象因子的关系 3.1 气象负荷的分离

 $E = {E_{\rm{e}}} + {E_{\rm{m}}} + {E_0},$ (2)

 ${E_{\rm{m}}} = E - {E_{\rm{e}}}.$ (3)
3.2 气象负荷与气象因子的关系

3.3 气象因子的累积效应对夏季气象负荷的影响

 图 1 2006~2016年北京夏季气象负荷与累积气象因子的相关系数的变化 Fig. 1 Variation of correlation coefficients between the meteorological electric loads and cumulative meteorological factors in the summer in Beijing from 2006 to 2016
4 夏季最大电力负荷预测模型的建立与检验 4.1 最大电力负荷的周末效应

 图 2 2006~2016年北京夏季平均星期一至星期日最大电力负荷 Fig. 2 Averaged summertime daily maximum electric loads from Monday to Sunday in Beijing from 2006 to 2016

4.2 模型的建立与检验

4.2.1 方案1模型结果

 图 3 夏季最大电力负荷预测模型方案1模拟值与实际值的对比：（a）2006~2015年；（b）2010年；（c）2016年 Fig. 3 Comparison between simulation value by the scheme 1 model and actual value of daily maximum electric loads in the summer: (a) 2006−2015; (b) 2010; (c) 2016

4.2.2 方案2和方案3模型结果

 图 4 2016年夏季最大电力负荷预测模型方案2和方案3模拟结果对比 Fig. 4 Comparison of simulation results between scheme 2 and scheme 3 models of daily maximum electric loads in the summer of 2016

5 2017年实际预测试验

 图 5 2017年夏季最大电力负荷预测模型方案1和方案3的预测结果对比：（a）气象因子为预报值；（b）气象因子为实际值 Fig. 5 Comparison of prediction results between scheme 1 and scheme 3 models of daily maximum electric loads in the summer of 2017: (a) Meteorological factors are predictive values; (b) meteorological factors are actual values

6 结论与讨论

（1）北京夏季日最大电力负荷与气象因子关系密切，对独立气象因子中的气温变化最敏感。而闷热指数综合考虑了气温和相对湿度的共同作用，其与最大电力负荷的相关性要强于任何单独气象因子。

（2）气象因子对北京夏季日最大负荷的影响具有累积效应，这种效应在累积两天时最明显。

（3）北京夏季最大负荷具有明显的周末效应，工作日的基础负荷要明显高于周末，剔除时应当加以区分。

（4）考虑到最大电力负荷具有“记忆性”，加入了前一日气象负荷的预测模型效果要显著好于仅考虑气象因子的模型，但实际工作中在无法获取这一数据的情况下，采用累积2 d闷热指数作为预测因子建立模型也不妨为一种选择。

（5）通过实际预测试验表明，方案1和方案3模型均具有较好的预测能力，能够满足实际业务需求。

（6）本文在研究过程中也拟合了其他值（例如历年逐日最大电力负荷、历年夏季最大电力负荷极值等）来代表基础负荷，但是最后的模型效果欠佳，故最终确定利用历年夏季最大电力负荷平均值拟合得到基础负荷。但城市基础负荷的研究一直是个难题，要完美地剔除基础负荷几乎不可能，这是本文的不足之处，也是下一步工作的重点。

（7）鉴于日最大电力负荷的变化机制十分复杂，影响因素多种多样，加上本文中采用的BP神经网络算法仍有不足，所建立的模型对极值的预测存在一定误差，未来还应当深入研究，采用大数据挖掘等新的方法，进一步提高模型的预测能力。

 [] 胡江林, 陈正洪, 洪斌, 等. 2002a. 华中电网日负荷与气象因子的关系[J]. 气象, 28(3): 14–18, 37. Hu Jianglin, Chen Zhenghong, Hong Bin, et al. 2002. A relationship between daily electric loads and meteorological elements in central China[J]. Meteorological Monthly (in Chinese), 28(3): 14–18, 37. DOI:10.3969/j.issn.1000-0526.2002.03.003 [] 胡江林, 陈正洪, 洪斌, 等. 2002b. 基于气象因子的华中电网负荷预测方法研究[J]. 应用气象学报, 13(5): 600–608. Hu Jianglin, Chen Zhenghong, Hong Bin, et al. 2002. Forecast technique of electric network loads in central China based on meteorological elements[J]. Journal of Applied Meteorological Science (in Chinese), 13(5): 600–608. DOI:10.3969/j.issn.1001-7313.2002.05.009 [] 罗慧, 巢清尘, 李奇, 等. 2005. 气象要素在电力负荷预测中的应用[J]. 气象, 31(6): 15–18. Luo Hui, Chao Qingchen, Li Qi, et al. 2005. Application of meteorological factors to load forecasting based on ANN[J]. Meteorological Monthly (in Chinese), 31(6): 15–18. DOI:10.3969/j.issn.1000-0526.2005.06.003 [] 陶琳, 岳小冰. 2016. 一种新的粒子群算法优化支持向量机的短期负荷预测[J]. 电子设计工程, 24(16): 151–154. Tao Lin, Yue Xiaobing. 2016. A new power load chaotic predicting based on support vector machine and particle swarm optimization algorithm[J]. Electronic Design Engineering (in Chinese), 24(16): 151–154. DOI:10.3969/j.issn.1674-6236.2016.16.047 [] 吴兑, 邓雪娇. 2001. 环境气象学与特种气象预报[M]. 北京: 气象出版社: 170-171. Wu Dui, Deng Xuejiao. 2001. Environmental Meteorology and Special Meteorological Forecast (in Chinese)[M]. Beijing: China Meteorological Press: 170-171. [] 徐沐阳, 何钢建, 胡元, 等. 2015. 基于气象因素的SVR方法在温州电网负荷预测中的应用[J]. 中国科技信息(3): 62–64. Xu Muyang, He Gangjian, Hu Yuan, et al. 2015. Application of SVR method based on meteorological factors in load forecasting of Wenzhou power grid[J]. China Science and Technology Information (in Chinese)(3): 62–64. DOI:10.3969/j.issn.1001-8972.2015.03.024 [] 叶殿秀, 张培群, 赵珊珊, 等. 2013. 北京夏季日最大电力负荷预报模型建立方法探讨[J]. 气候与环境研究, 18(6): 804–810. Ye Dianxiu, Zhang Peiqun, Zhao Shanshan, et al. 2013. Research on meteorological forecast technique of daily maximum electric loads during summer in Beijing[J]. Climatic and Environmental Research (in Chinese), 18(6): 804–810. DOI:10.3878/j.issn.1006-9585.2013.12146 [] 尤焕苓, 丁德平, 王春华, 等. 2008. 应用回归分析和BP神经网络方法模拟北京地区电力负荷[J]. 气象科技, 36(6): 801–805. You Huanling, Ding Deping, Wang Chunhua, et al. 2008. Application of regression analysis and artificial neural network to Beijing daily power peak load forecast[J]. Meteorological Science and Technology (in Chinese), 36(6): 801–805. DOI:10.3969/j.issn.1671-6345.2008.06.026 [] 张自银, 马京津, 雷杨娜. 2011. 北京市夏季电力负荷逐日变率与气象因子关系[J]. 应用气象学报, 22(6): 760–765. Zhang Ziyin, Ma Jingjin, Lei Yangna. 2011. Beijing electric power load and its relation with meteorological factors in summer[J]. Journal of Applied Meteorological Science (in Chinese), 22(6): 760–765. DOI:10.3969/j.issn.1001-7313.2011.06.015 [] 郑贤, 唐伍斌, 贝宇, 等. 2008. 桂林电网日负荷与气象因素的关系及其预测[J]. 气象, 34(10): 96–101. Zheng Xian, Tang Wubin, Bei Yu, et al. 2008. A relationship between daily load and meteorological factors for Guilin power network and forecasting[J]. Meteorological Monthly (in Chinese), 34(10): 96–101. DOI:10.7519/j.issn.1000-0526.2008.10.013