作者：李练兵1, 李佳祺2, 刘汉民3, 李明3, 任杰3, 王阳3, 马步云3, 田云峰3
作者单位：1. 省部共建电工装备可靠性与智能化国家重点实验室(河北工业大学), 天津 300130;
2. 河北工业大学人工智能与数据科学学院, 天津 300130;
3. 国网冀北张家口风光储输新能源有限公司, 河北 张家口 075000
关键词：短期电力负荷预测;堆叠的反向双层高低级门控循环单元;差分分解;误差补偿
摘要：
为解决短期电力负荷预测模型中迭代训练过程的误差累积，预测结果精度低的问题，基于双向门控循环单元（bigru）在短期电力负荷预测中的理论基础，在bigru的底层结构上进行改进，并对预测过程中产生的误差进行补偿，提出一种基于srdhlgru神经网络和差分误差补偿的短期电力负荷预测方法。第一阶段，建立堆叠的反向双层高低级门控循环单元（srdhlgru）网络模型，得到模型预测过程中产生的误差序列。第二阶段采用差分分解（dd）方法将第一阶段产生的误差序列进行一阶前向差分得到误差变化量序列，再次建立srdhlgru模型进行训练和预测，从而对第一阶段结果进行误差补偿。结合西部某市负荷数据集，基于python算法开展短期电力负荷预测的仿真，对比几种主流预测算法，仿真结果表明该组合模型的预测精度和稳定性比传统模型都有一定提升。
short-term load forecasting based on srdhlgru neural network and differential error compensation
li lianbing1, li jiaqi2, liu hanmin3, li ming3, ren jie3, wang yang3, ma buyun3, tian yunfeng3
1. state key laboratory of electrical equipment reliability and intelligence (hebei university of technology), tianjin 300130, china;
2. school of artificial intelligence and data science, hebei university of technology, tianjin 300130, china;
3. state grid zhangjiakou, hebei wind-light storage new energy limited company, zhangjiakou 075000, china
abstract: in order to solve the problem of error accumulation in the iterative training process and low prediction accuracy in the short-term power load forecasting model, based on the theoretical basis of bi-directional gated recurrent unit (bigru) in short-term power load forecasting, the underlying structure of bigru is improved, and to compensate the errors generated in the forecasting process, a short-term power load forecasting method based on srdhlgru neural network and differential error compensation is proposed. in the first stage, a stacked reverse double-layer high-low-level gated recurrent unit (srdhlgru) network model is established to obtain the error sequence generated during the model prediction process. in the second stage, the difference decomposition (dd) method is used to perform the first-order forward difference of the error sequence generated in the first stage to obtain the error change sequence, and the srdhlgru model is established again for training and prediction, so as to compensate the error of the first stage results. combined with the load data set of a city in the west, a short-term power load forecasting simulation is carried out based on the python algorithm. compared with several mainstream forecasting algorithms, the simulation results show that the forecasting accuracy and stability of the combined model are improved compared with the traditional models.
keywords: short-term electric load forecast;srdhlgru;difference decomposition;error compensation
2023, 49(6):143-150 收稿日期: 2021-10-17;收到修改稿日期: 2021-12-23
基金项目: 河北省省级科技计划资助 (20312102d)
作者简介: 李练兵(1972-),男,天津市人,教授,硕士生导师,博士,研究方向为电力电子技术、电力系统分析、新能源发电与微电网技术
参考文献
[1] farrokh r, aii i. demand response as a market resource under the smart grid paradigm[j]. ieeetransaction on smart grid, 2010(1): 82-88
[2] wang z, li h, tang z, et al. user-level ultra-short-term load forecasting model based on optimal feature selection and bahdanau attention mechanism[j]. journal of circuits systems and computers, 2021, 30(15): 215-279
[3] zhu z w, zhou m, hu f, et al. a day-ahead industrial load forecasting model using load change rate features and combining fa-elm and the adaboost algorithm[j]. energy reports, 2023(7): 971-981
[4] 肖白, 赵晓宁, 姜卓, 等. 利用模糊信息粒化与支持向量机的空间负荷预测方法[j]. 电网技术, 2021, 45(1): 251-260
[5] zhao h s, tian t. power system dynamic state estimation based on adaptive tracelesss kalman filter[j]. power system technology, 2014(1): 3790-3794
[6] 吴潇雨, 和敬涵, 张沛, 等. 基于灰色投影改进随机森林算法的电力系统短期负荷预测[j]. 电力系统自动化, 2015, 39(12): 50-55
[7] 陈艳平, 毛弋, 陈萍, 等. 基于eemd-样本熵和elman神经网络的短期电力负荷预测[j]. 电力系统及其自动化学报, 2016, 28(3): 59-64
[8] xia t, yu d, qi l, et al. application of bidirectional recurrent neural network combine with deep belife network in short-term load forecasting[j]. ieee access, 2019, 7: 160660-160670
[9] jiao r h, zhang t m,jiang y z, et al. short-term non-residential load forecasting based on multiple sequences lstm recurrent neural network[j]. ieee access, 2018, 6: 59438-59448
[10] 曾囿钧, 肖先勇, 徐方维. 基于小波变换与bigru-nn模型的短期负荷预测方法[j/ol]. 电测与仪表: 1-8[2021-10-06]. http://kns.cnki.net/kcms/detail/23.1202.th.20210113.1548.002.html.
[11] 邓带雨, 李坚, 张真源, 等. 基于eemd-gru-mlr的短期电力负荷预测[j]. 电网技术, 2020, 44(2): 593-602
[12] 王晓燕, 郎贺, 王品, 等. 基于stl分解的平均故障间隔时间组合预测[j]. 机床与液压, 2021, 49(17): 196-200
