TONG Lin,HE Qingqing.Power load interval forecast based on weighted Markov modified fuzzy information granule[J].Journal of Yanbian University,2022,(02):151-157.
加权Markov修正模糊信息粒的电力负荷区间预测
- Title:
- Power load interval forecast based on weighted Markov modified fuzzy information granule
- 文章编号:
- 1004-4353(2022)02-0151-07
- Keywords:
- power load forecast; weighted Markov; fuzzy information granulation; long and short - term memory network; error correction
- 分类号:
- TP391
- 文献标志码:
- A
- 摘要:
- 针对电力负荷预测存在波动性且预测精度不高的问题,提出一种基于加权马尔可夫(Markov)修正模糊信息粒的电力负荷区间预测方法.该方法首先对电力负荷数据序列进行基于模糊信息粒化(FIG)的空间窗口重构,以此得到电力负荷模糊信息粒和电力负荷的各阶自相关系数; 然后建立由基于FIG和长短时记忆网络(LSTM)组合的模型(FIG - LSTM),以此获得能够预测不同模糊粒的3组LSTM模型; 最后建立加权Markov - FIG - LSTM模型,并通过消除3组LSTM模型中的预测误差得到电力负荷预测区间和趋势值.实例分析表明,Markov - FIG - LSTM模型的RMSE、MAE和MAPE指标比FIG - LSTM模型分别降低了4.78%、11.37%和11.72%,因此该方法可为电网调度提供有效的数据支撑.
- Abstract:
- Aiming at the problem of fluctuation and low precision in power load forecasting, a method of power load interval prediction based on weighted Markov modified fuzzy information granules was proposed.Firstly, the spatial window of power load data series is reconstructed based on fuzzy information granulation(FIG), and the order autocorrelation coefficients of fuzzy information granules and power load are obtained.Then, a combination model(FIG - LSTM)based on FIG and LSTM was established to obtain three groups of LSTM model that could predict different fuzzy particles.Finally, the weighted Markov - FIG - LSTM model is established, and the power load prediction interval and trend value are obtained by eliminating the prediction error of three groups of LSTM models.The example analysis shows that the RMSE, MAE and MAPE indexes of the Markov - FIG - LSTM model are reduced by 4.78%, 11.37% and 11.72% respectively compared with the FIG - LSTM model.Therefore, this method can provide effective data support for power grid dispatching.
参考文献/References:
[1] 庄世杰,於志勇,郭文忠,等.基于Zoneout的跨尺度循环神经网络及其在短期电力负荷预测中的应用[J].计算机科学,2020,47(9):105 - 109.
[2] 王克杰,张瑞.基于改进BP神经网络的短期电力负荷预测方法研究[J].电测与仪表,2019,56(24):115 - 121.
[3] 张淑清,杨振宁,张立国,等.基于弹性网降维及花授粉算法优化BP神经网络的短期电力负荷预测[J].仪器仪表学报,2019,40(7):47 - 54.
[4] 童林,官铮.改进鲸鱼优化支持向量机的交通流量模糊粒化预测[J].计算机应用,2021,41(10):2919 - 2927.
[5] 胡时雨,罗滇生,阳霜,等.基于多变量LS - SVM和模糊循环推理系统的负荷预测[J].计算机应用,2015,35(2):595 - 600.
[6] 俞颖,黄风华,阮奇.基于改进粒子群优化算法和CRNN的多类SVM分类[J].延边大学学报(自然科学版),2019,45(3):215 - 220.
[7] 张宇帆,艾芊,林琳,等.基于深度长短时记忆网络的区域级超短期负荷预测方法[J].电网技术,2019,43(6):1884 - 1892.
[8] 茆美琴,洪嘉玲,张榴晨.考虑光伏出力预测误差修正的储能优化配置方法[J].太阳能学报,2021,42(2):410 - 416.
[9] 童林,官铮,王立威,等.基于时序分解与误差修正的新能源爬坡事件预测[J].浙江大学学报(工学版),2022,56(2):338 - 346.
[10] 刘帅,朱永利,张科,等.基于误差修正ARMA - GARCH模型的短期风电功率预测[J].太阳能学报,2020,41(10):268 - 275.
[11] 张娜,王守相,葛磊蛟,等.一种光伏短期出力区间预测方法[J].太阳能学报,2020,41(8):173 - 179.
[12] ZADEH L A.Toward a theory of fuzzy informa - tion granulation and its centrality in human reasoning and fuzzy logic[J].Fuzzy Sets & Systems, 1997,90(2):111 - 127.
[13] 柯德营,殷业,杜军辉.基于模糊信息粒理论的模糊控制方法[J].计算机仿真,2016,33(1):391 - 395.
[14] PAN W, CHEN Q, ZHU M, et al.A data - driven fuzzy information granulation approach for battery state of health forecasting[J].Journal of Power Sources, 2020,475:228716.
[15] GREFF K, SRIVASTAVA R K, KOUTNÍK J, et al. LSTM: A search space odyssey[J]. IEEE Transactions on Neural Networks & Learning Systems, 2016,28(10):2222 - 2232.
[16] 杨梓鑫,薛源,徐浩军,等.基于RBF神经网络与Markov组合的飞行风险预测研究[J].系统工程理论与实践,2019,39(8):2162 - 2169.
[17] JIANG X C, CHEN S F.Application of weighted Markov SCGM(1,1)model to predict drought crop area[J]. Systems Engineering - Theory & Practice, 2009,29(9):179 - 185.
[18] 苗作华,刘耀林,王海军.耕地需求量预测的加权模糊-马尔可夫链模型[J].武汉大学学报(信息科学版),2005(4):305 - 308.
备注/Memo
收稿日期: 2021-08-22
作者简介: 童林(1995—),男,讲师,研究方向为机器学习、智能控制.