葛磊蛟 ,刘航旭 ,赵 康 ,李 华 ,张 波 ,李元良
AuthorsHTML:葛磊蛟 ¹ ，刘航旭 ¹ ，赵 康 ² ，李 华 ³ ，张 波 ⁴ ，李元良 ¹
AuthorsListE:Ge Leijiao,Liu Hangxu,Zhao Kang,Li Hua,Zhang Bo,Li Yuanliang
AuthorsHTMLE:Ge Leijiao¹，Liu Hangxu¹，Zhao Kang ²，Li Hua³，Zhang Bo⁴，Li Yuanliang¹
Unit:1. 天津大学智能电网教育部重点实验室，天津 300072；
2. 山东科技大学电气与自动化工程学院，青岛 266590；
3. 国网辽宁省电力有限公司经济技术研究院，沈阳 110015；
4. 中国人民解放军 32654 部队，济南 250024

Unit_EngLish:1. Key Laboratory of Smart Grid of Ministry of Education，Tianjin University，Tianjin 300072，China；
2. School of Electrical and Automation Engineering，Shandong University of Science and Technology，Qingdao 266590， China；
3. Economic and Technological Research Institute，State Grid Liaoning Electric Power Co.，Ltd.，Shenyang 110015，China；
4. 32654 PLA Troops，Jinan 250024，China

Abstract_Chinese:针对商业和居民混合的配电网负荷预测影响因素多样和随机性强的特点，提出了一种基于灰狼郊狼混合优 化算法(HGWOACOA)和长短时记忆网络(LSTM)相结合的配电网短期电力负荷预测方法．为实现对电力负荷的多 输入特征进行筛选和降维，本文采用最大信息系数法(MIC)对负荷与输入特征之间的非线性相关性进行量化，剔除 无效特征，再采用因子分析(FA)法对历史数据、气象、温度等输入特征进行信息提取，实现对有效影响因素降维； 综合考虑灰狼优化算法(GWOA)具有全局搜索能力强、收敛速度快的特点，郊狼优化算法(COA)具有区域探索能力 强、求解精度高的特点，为突出两种算法优点并弥补两种算法的不足，提出 HGWOACOA 对 LSTMN 的隐藏层神经 元个数和一次训练样本个数两种参数进行优化，克服了参数选取的困难．在保证预测精度的同时提升运行效率，实 现基于 HGWOACOA-LSTMN 的配电网短期电力负荷高精度预测．以配电网某商业和居民混合负荷为例，验证了文 中所述预测方法具有较好的精度和鲁棒性，同时提升了运行效率和简化模型输入特征的维数，证明本文所提方法的 精确度和适用性．
Abstract_English:For the commercial and residential mixed distribution network load forecasting，the influence factors are diverse and randomness is high. A short-term power load forecasting method of distribuion network based on the hy\u0002drid grey wolf optimization algorithm with coyote optimization algorithm(HGWOACOA) and long short-term mem\u0002ory network (LSTMN)is proposed. To select and reduce multiple input characteristics of power load，in this paper， we use the maximum information coefficient(MIC) method to quantify the nonlinear correlation between the load and input characteristics，eliminate the invalid features，and then use the factor analysis(FA) method to extract the information of historical data，meteorology，temperature and other input features，so as to reduce the dimension ofeffective influencing factors. Results show that the grey wolf optimization algorithm(GWOA) has strong global search ability and fast convergence speed，and the coyote optimization algorithm(COA) has strong global search ability and fast convergence speed. To highlight the advantages of the two algorithms and make up for their corre\u0002sponding shortcomings，a hybrid GWOA with COA(HGWOACOA)is proposed to solve long short-term memory networks. The number of hidden layer neurons and the number of training samples are optimized to overcome the dif\u0002ficulty of parameter selection. While ensuring the prediction accuracy，the operation efficiency is also improved， which realized the high-precision short-term power load forecasting of the distribution network based on HGWOACOA-LSTMN. Taking a commercial and residential mixed load in a distribution network as an example，the accuracy and robustness of the proposed forecasting method are verified. Meanwhile，the operation efficiency is im\u0002proved and the dimension of the input characteristics of the model is simplified，which proves the accuracy and applicability of the proposed method.
Keyword_Chinese:配电网；短期负荷预测；因子分析；灰狼优化算法；郊狼优化算法；长短时记忆网络
Keywords_English:distribution network；short-term load forecasting；factor analysis；grey wolf optimization algorithm； coyote optimization algorithm；long short-term memory network

PDF全文下载地址:http://xbzrb.tju.edu.cn/#/digest?ArticleID=6717