基于多参数代价敏感系数学习及数据驱动模型的电力能耗预测

doi:10.3969/j.issn.1674-1951.2021.08.008

华电技术 ›› 2021, Vol. 43 ›› Issue (8): 54-60.doi: 10.3969/j.issn.1674-1951.2021.08.008

基于多参数代价敏感系数学习及数据驱动模型的电力能耗预测

施杰(), 张安勤^*()

上海电力大学计算机科学与技术学院,上海 200090

收稿日期:2020-09-11 修回日期:2021-05-20 出版日期:2021-08-25 发布日期:2021-08-24
通讯作者: 张安勤
作者简介:施杰（1992—）,女,河南驻马店人,在读硕士研究生,从事数据挖掘研究（E-mail: 1181540647@qq.com）。
基金资助:
国家自然科学基金项目(61772327)

Power consumption prediction based on multi-parameter cost-sensitive coefficient learning and data-driven model

SHI Jie(), ZHANG Anqin^*()

College of Computer Science and Technology,Shanghai University of Electric Power,Shanghai 200090,China

Received:2020-09-11 Revised:2021-05-20 Online:2021-08-25 Published:2021-08-24
Contact: ZHANG Anqin

摘要/Abstract

摘要：

企业节能减排是我国社会发展的前沿问题和研究热点,对企业用户的电力能耗现状进行综合分析与评估是进行节能改造或节能设计的前提和基础。阐述了现阶段常用的电力能耗预测方法,分析了分类与回归树（CART）算法作为弱学习器构建预测模型的缺点,针对原始AdaBoost算法只关注预测误差率最小的不足,在算法实质基础上研究并提出一种基于多参数代价敏感系数学习的改进AdaBoost算法。建立基于改进AdaBoost算法的回归预测模型,通过真实数据进行短期电力能耗预测,验证了改进算法对模型性能的提升。

关键词: 电力能耗预测, 代价敏感系数, 数据驱动, 分类与回归树算法, AdaBoost算法

Abstract:

Energy conservation and emission reduction in enterprises are the frontier issues and research hot-spots in the way of China´s development. Comprehensive analysis and evaluation on the current situation of power consumption in enterprises is the premise and foundation for energy-saving transformations or energy conservation designs. Having described the common forecasting methods for electricity consumption, the shortcomings of constructing a prediction model taking Classification and Regression Tree （CART）algorithm as the weak learner are analyzed. To deal with the deficiency of the traditional AdaBoost algorithm focusing on the minimum prediction error rate only, an improved AdaBoost algorithm based on multi-parameter cost-sensitive coefficient learning is studied and proposed based on the essence of the algorithm. A regression prediction model constructed based on the improved AdaBoost algorithm can make short-term power consumption prediction according to real data, which verifies the improvement of the model´s performance.

Key words: electricity consumption forecast, cost-sensitive coefficient, data-driven, Classification and Regression Tree algorithm, AdaBoost algorithm

中图分类号:

TM74

施杰, 张安勤. 基于多参数代价敏感系数学习及数据驱动模型的电力能耗预测[J]. 华电技术, 2021, 43(8): 54-60.

SHI Jie, ZHANG Anqin. Power consumption prediction based on multi-parameter cost-sensitive coefficient learning and data-driven model[J]. Huadian Technology, 2021, 43(8): 54-60.

图/表 7

表1

表2

表3

图1

表4

表5

表6

参考文献 13

[1]	梁玉红, 聂淑花. 产业升级背景下我国战略性新兴产业发展研究[J]. 中国党政干部论坛, 2017(1):22-24.
	LIAGN Yuhong, NIE Shuhua. Research on the development of strategic emerging industries in China under the background of industrial upgrading[J]. Chinese Cadres Tribune, 2017(1):22-24.
[2]	郑琪文, 程方晓. 电力系统继电保护不稳定原因及解决办法研究[J]. 科技风, 2020(18):201-202.
	ZHENG Qiwen, CHENG Fangxiao. Research on the reasons and solutions of the unstable relay protection of power system[J]. Technology Wind, 2020(18):201-202.
[3]	岳峰, 史志伟, 董金才, 等. 智能电网继电保护控制设备硬件可靠性设计及测试[J]. 华电技术, 2020, 42(2):50-57.
	YUE Feng, SHI Zhiwei, DONG Jincai, et al. Reliability design and test for relay protection devices applied in smart grid[J]. Huadian Technology, 2020, 42(2):50-57.
[4]	吴源梓. 厉行节约倡导低碳生活——基层央行节能减排工作思考[J]. 时代金融, 2016(23):315-316.
	WU Yuanzi. Strict economy and advocate low-carbon life:Thinking about the work of energy saving and emission reduction of the basic central bank[J]. Times Finance, 2016(23):315-316.
[5]	胡倩, 孙志达, 江坷滕, 等基于短期负荷打捆预测的售电公司偏差考核控制方法[J]. 华电技术, 2021, 43(4):47-55.
	HU Qian, SUN Zhida, JIANG Keteng, et al. Deviation assessment and control method for electricity sales companies based on short-term load bundling forecast[J]. Huadian Technology, 2021, 43(4):47-55.
[6]	卢芸. 短期电力负荷预测关键问题与方法的研究[D]. 沈阳:沈阳工业大学, 2007.
[7]	陈涛, 吕松, 任廷林, 等. 基于最小二乘支持向量机的周用电量预测方法[J]. 华电技术, 2020, 42(1):35-40.
	CHEN Tao, LYU Song, REN Tinglin, et al. Prediction method for weekly electricity consumption based on LSSVM algorithm[J]. Huadian Technology, 2020, 42(1):35-40.
[8]	JIN Y, RIVARD H, ZMEUREANU R. On-line building energy prediction using adaptive artificial neural networks[J]. Energy & Buildings, 2005, 37(12):1250-1259.
[9]	NETO A H, FIORELLI F A S. Comparison between detailed model simulation and artificial neural network for forecasting building energy consumption[J]. Energy & Buildings, 2008, 40(12):2169-2176.
[10]	李正方, 杜景林, 周芸. 基于改进CART算法的降雨量预测模型[J]. 现代电子技术, 2020(2):133-137.
	LI Zhengfang, DU Jinglin, ZHOU Yun. Rainfall prediction model based on improved CART algorithm[J]. Modern Electronics Technique, 2020(2):133-137.
[11]	FRIEDMAN J H, HASTIE T, TIBSHIRANI R. Additive logistic regression:A statistical view of boosting[J]. Annals of Statistics, 2000, 28(2):337-407.
[12]	FRIEDMAN J H. Greedy function approxi-mation: A gradient boosting machine[J]. Annals of Statistics, 2001, 29(5):1189-1232. doi: 10.1214/aos/1013203450
[13]	王学玲, 王建林. 基于代价敏感的AdaBoost算法改进[J]. 计算机应用与软件, 2013(10):123-125.
	WANG Xueling, WANG Jianlin. Improving AdaBoost algorithm based on cost-sensitive[J]. Computer Applications and Software, 2013(10):123-125.

record_date	power_consumption/（kW·h）
2015-01-01	2 900 575
2015-01-02	3 158 211
2015-01-03	3 596 487
2015-01-04	3 939 672
2015-01-05	4 101 790

record_date	power_consumption/（kW·h）	dow	dom	month	year	weekend	weekend_sat	weekend_sun	week_of_month	period_of_month	period2_of_month
2015-01-01	2 900 575	3	1	1	2015	0	0	0	1	1	1
2015-01-02	3 158 211	4	2	1	2015	0	0	0	1	1	1
2015-01-03	3 596 487	5	3	1	2015	1	1	0	1	1	1
2015-01-04	3 939 672	6	4	1	2015	1	0	1	1	1	1
2015-01-05	4 101 790	0	5	1	2015	0	0	0	4	3	2

dow_0	dow_3	dow_4	dow_5	...	period_of_month_1	period_of_month_2	period2_of_month_1	period2_of_month_2
0	1	0	0	...	1	0	1	0
0	0	1	0	...	1	0	1	0
0	0	0	1	...	1	0	1	0
0	0	0	0	...	1	0	1	0
1	0	0	0	...	0	1	0	1

预测模型		弱学习器数量
预测模型		100	200	300	400	500	600	700	800	900	1 000
原始算法	训练集	0.414 3	0.414 3	0.453 7	0.476 1	0.476 1	0.474 1	0.491 2	0.497 4	0.476 0	0.492 0
原始算法	测试集	0.581 4	0.681 4	0.661 4	0.663 2	0.563 2	0.514 8	0.661 1	0.561 1	0.544 1	0.514 1
参数组合1	训练集	0.460 5	0.460 5	0.468 1	0.488 1	0.492 7	0.499 6	0.460 5	0.493 6	0.470 9	0.501 2
参数组合1	测试集	0.673 4	0.773 4	0.753 4	0.755 2	0.655 2	0.606 8	0.753 1	0.653 1	0.636 1	0.606 1
参数组合2	训练集	0.493 6	0.483 6	0.474 2	0.482 2	0.496 1	0.519 4	0.501 1	0.446 8	0.514 2	0.521 6
参数组合2	测试集	0.755 4	0.855 4	0.835 4	0.837 2	0.737 2	0.688 8	0.835 1	0.735 1	0.718 1	0.688 1

预测模型		学习率
预测模型		0.001	0.003	0.010	0.030	0.100	0.300	0.500	1.000	3.000	10.000
原始算法	训练集	0.478 7	0.465 7	0.464 5	0.490 7	0.480 0	0.479 1	0.480 9	0.476 5	0.516 3	0.489 4
原始算法	测试集	0.478 2	0.552 8	0.590 2	0.616 6	0.510 0	0.452 2	0.437 3	0.572 7	0.480 3	0.435 1
参数组合1	训练集	0.513 8	0.495 2	0.508 7	0.469 6	0.475 5	0.470 1	0.475 6	0.473 0	0.523 7	0.488 3
参数组合1	测试集	0.568 2	0.642 8	0.680 2	0.706 6	0.600 0	0.542 2	0.527 3	0.662 7	0.570 3	0.525 1
参数组合2	训练集	0.505 4	0.558 6	0.555 3	0.569 7	0.552 8	0.581 2	0.482 9	0.528 6	0.461 6	0.525 1
参数组合2	测试集	0.649 2	0.723 8	0.761 2	0.787 6	0.681 0	0.623 2	0.608 3	0.743 7	0.651 3	0.606 1

基于多参数代价敏感系数学习及数据驱动模型的电力能耗预测

Power consumption prediction based on multi-parameter cost-sensitive coefficient learning and data-driven model

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 13

相关文章 3

编辑推荐

Metrics

本文评价

dow_0	dow_3	dow_4	dow_5	...	period_of_month_1	period_of_month_2	period2_of_month_1	period2_of_month_2
0	1	0	0	...	1	0	1	0
0	0	1	0	...	1	0	1	0
0	0	0	1	...	1	0	1	0
0	0	0	0	...	1	0	1	0
1	0	0	0	...	0	1	0	1

学习器数量	损失函数	原始算法		参数组合1		参数组合2
学习器数量	损失函数	训练集	测试集	测试集	测试集	测试集	测试集
100	线性	0.459 0	0.590 6	0.511 5	0.679 6	0.543 1	0.770 6
	平方	0.425 3	0.514 0	0.488 5	0.603 0	0.491 0	0.694 0
	指数	0.435 5	0.496 1	0.496 3	0.585 1	0.504 3	0.676 1
300	线性	0.466 9	0.495 8	0.525 4	0.584 8	0.565 4	0.675 8
	平方	0.453 5	0.597 7	0.468 2	0.686 7	0.519 8	0.777 7
	指数	0.443 0	0.193 1	0.545 9	0.282 1	0.509 2	0.373 1
500	线性	0.486 0	0.495 8	0.494 1	0.584 8	0.551 9	0.675 8
	平方	0.481 7	0.497 7	0.531 3	0.586 7	0.530 9	0.677 7
	指数	0.454 2	0.507 5	0.493 3	0.596 5	0.508 9	0.687 5
700	线性	0.485 5	0.492 7	0.560 8	0.581 7	0.565 5	0.672 7
	平方	0.487 0	0.451 0	0.535 4	0.540 0	0.584 4	0.631 0
	指数	0.447 2	0.693 4	0.546 0	0.782 4	0.513 7	0.873 4

[1]	李彩霞, 赵军, 李建伟, 王伟, 王杰, 于浩洋. 基于数据驱动的CFB机组变负荷工况SO₂质量浓度建模[J]. 综合智慧能源, 2022, 44(3): 63-69.
[2]	石立宝, 翟放. 考虑风-光-荷不确定性的数据驱动型机组组合模型[J]. 综合智慧能源, 2022, 44(1): 18-25.
[3]	韩旭东. 基于数据驱动的火电机组高压加热系统异常检测研究[J]. 华电技术, 2021, 43(8): 67-73.

dow_0	dow_3	dow_4	dow_5	...	period_of_month_1	period_of_month_2	period2_of_month_1	period2_of_month_2
0	1	0	0	...	1	0	1	0
0	0	1	0	...	1	0	1	0
0	0	0	1	...	1	0	1	0
0	0	0	0	...	1	0	1	0
1	0	0	0	...	0	1	0	1

dow_0	dow_3	dow_4	dow_5	...	period_of_month_1	period_of_month_2	period2_of_month_1	period2_of_month_2
0	1	0	0	...	1	0	1	0
0	0	1	0	...	1	0	1	0
0	0	0	1	...	1	0	1	0
0	0	0	0	...	1	0	1	0
1	0	0	0	...	0	1	0	1