对一定地区进行科学合理的电网规划的基本前提是精准预测该地区未来用电负荷的规模布局、用电特性。光伏发电和风力发电等清洁电源以及柔性负荷和储能系统等新型电力要素的加入为电网的转型和发展提供了巨大的动力，实现对清洁能源、分布式能源、柔性负荷和储能系统大量接入的新型电力系统的精确负荷预测可以提高电力系统的运行可靠性和调度经济性，而传统的负荷预测已不能满足构建能源互联网和新型电力系统的要求。基于此，本文主要工作如下：

首先，通过对短期负荷预测工作的背景和意义、现状和主流预测模型和理论的研究得出设计应用取长补短的组合算法比单一算法更高效精确的结论，为下文建立复合预测模型奠定理论基础。同时，以缺乏除水电外的调峰调频电源，几乎无火电支撑，发电负荷波动性极大的甘肃某地区电网为例，分析了典型的“双高”外送型清洁能源密集型电网中各种内外界因素对不同性质和行业负荷的影响，得出各类用电负荷由于用电特性不同而对各种因素敏感程度不同的结论。选取甘肃某地区2023年全年的负荷数据及相关数据，为在保证预测精度的同时避免输入数据维度过大导致的预测模型难以收敛、运算速率大幅下降等问题，确定本文所建立的输入数据为包含历史负荷数据、气候特征值、日期类型特征值数据在内的21维结构，并对该数据集中的缺失、异常数据进行了处理。

其次，为了提升灰狼算法（GWO）的运算速度和抗干扰能力，本文建立了基于混沌映射及正余弦搜索策略和混沌扰动的改进灰狼优化算法（SCGWO）。其中，引入Sin混沌对输入数据集进行初始化处理以提高原始数据集的广泛性、代表性和遍布性，从而改进搜索空间的使用效果；使用正弦余弦搜索策略提高算法寻找全局最优解的精确度；采用混乱扰动策略避免了基本灰狼算法最佳解仅局限在局部最优的问题，从而提高了算法进行大规模搜索的能力。此外，为验证SCGWO的各方面性能分别选取7个单峰函数检验趋近能力、精确程度，选取6个多峰函数检验跳出局部极值的能力、总体稳定性和整体搜索能力，以及选取10个固定维度多峰函数检验协调能力。

最后，在GWO与SCGWO算法的基础上，对随机森林算法（RF）的相关理论进行了研究，构建了GWO-RF与SCGWO-RF短期负荷预测模型，利用SCGWO数据遍历性强、全局优化性强的特点，针对RF算法选取参数代表性差的局限，随机选取步长，并确定所得出的解为最优解。选取甘肃某地区2023年全年的负荷数据及相关数据，通过仿真与实验表明融合正余弦搜索和混沌扰动的灰狼优化算法与随机森林算法的结合有效提高了模型的精度，明显加快了迭代到最优解的进程。

The basic premise of scientific and reasonable power grid planning for a certain area is to accurately predict the scale layout and power characteristics of the future power load in the area. The addition of clean power sources such as photovoltaic power generation and wind power generation, as well as new power elements such as flexible load and energy storage systems, provides a huge impetus for the transformation and development of the power grid. The accurate load prediction of the new power system with a large number of clean energy, distributed energy, flexible load and energy storage systems can improve the operation reliability and dispatch economy of the power system. However, the traditional load forecasting can not meet the requirements of building energy Internet and new power system. Based on this, the main work of this paper is as follows:

First of all, through the study of the background, significance, current situation and mainstream forecasting models and theories of short-term load forecasting, it is concluded that the combination algorithm designed and applied to learn from each other's strengths is more efficient and accurate than a single algorithm, which lays a theoretical foundation for the establishment of the composite forecasting model. At the same time, taking a power grid in Gansu Province, which lacks peaking and frequency modulation power supply except hydropower, almost no thermal power support, and the power generation load fluctuates greatly as an example, the influence of various internal and external factors on different properties and industry loads in a typical "double-high" clean energy intensive power grid is analyzed. It is concluded that each type of power load is sensitive to different factors due to different power characteristics. The load data and related data of a region in Gansu province for the whole year of 2023 are selected. In order to ensure the prediction accuracy and avoid the problems such as the difficulty of convergence of the prediction model and the significant decrease of the operation rate caused by the excessive dimension of the input data, the input data established in this paper is determined to be a 21-dimensional structure including historical load data, climate characteristic value and date type characteristic value data. The missing and abnormal data in the data set are dealt with.

Secondly, in order to improve the operation speed and anti-interference ability of GWO, an improved Grey Wolf optimization algorithm (SCGWO) based on chaotic mapping, sine-cosine search strategy and chaos perturbation is established. Sin chaos is introduced to initialize the input data set to improve the universality, representativeness and ubiquity of the original data set, so as to improve the use effect of the search space. The sines and cosines search strategy is used to improve the accuracy of the algorithm to find the global optimal solution. The chaotic perturbation strategy avoids the problem that the optimal solution of the basic gray Wolf algorithm is only limited to the local optimal, thus improving the ability of the algorithm to carry out large-scale search. In addition, in order to verify the performance of SCGWO in various aspects, 7 unimodal functions were selected to test the reaching ability and accuracy degree, 6 multi-modal functions were selected to test the ability to jump out of local extreme values, the overall stability and the overall search ability, and 10 fixed dimension multi-modal functions were selected to test the coordination ability.

Finally, in order to solve the subjectivity and limitation caused by the selection of decision tree number n based on experience, which plays a decisive role in the prediction effect of RF algorithm, the relevant theories of random forest algorithm (RF) are studied on the basis of GWO and SCGWO algorithms, and the short-term load prediction models of GGO-RF and SCGWO-RF are constructed. Based on the strong ergodic and global optimization characteristics of SCGWO data, the step size is randomly selected to determine the optimal solution for the poor representation of parameters selected by RF algorithm. The load data and related data of a region in Gansu province in 2023 are selected. Simulation and experiment show that the combination of sine-cosine search and chaotic disturbance grey Wolf optimization algorithm and random forest algorithm can effectively improve the accuracy of the model and accelerate the process of iteration to the optimal solution.

[1]王康. 基于改进GRU神经网络的电力系统短期负荷预测研究[D]. 青岛大学, 2021.

[2]朱彤. 能源转型中我国电力能源的结构、问题与趋势[J]. 经济导刊, 2020(06): 48-53.

[3]李晖, 刘栋, 姚丹阳. 面向碳达峰碳中和目标的我国电力系统发展研判[J]. 中国电机工程学报, 2021, 41(18): 6245-6259.

[4]Gielen D., Boshell F., Saygin D., et al. The Role of Renewable Energy in The Global Energy Transformation[J]. Energy Strategy Reviews, 2019, 7(24): 38-50.

[5]Rafi S H, Deeba S R, Hossain E. A short-term load forecasting method using integrated CNN and LSTM network[J]. IEEE Access, 2021, 9: 32436-32448.

[6]韩富佳，王晓辉，乔骥等. 基于人工智能技术的新型电力系统负荷预测研究综述[J/OL]. 中国电机工程学报:1-24[2023-04-14].

[7]曾其权，马驰，冯彩梅等. 储能产业发展现状和机遇的研究与探讨[J]. 中国能源,2022,44(07):59-65.

[8]蔡黎，葛棚丹，代妮娜等. 电动汽车入网负荷预测及其与电网互动研究进展综述[J]. 智慧电力,2022,50(07):96-103.

[9]陈光宇，杨锡勇，江海洋等. 高比例新能源接入下计及工业负荷特性的电网需求响应调控策略[J/OL]. 电力自动化设备,2023(04):1-8[2023-04-14].

[10]朱继忠，董瀚江，李盛林等. 数据驱动的综合能源系统负荷预测综述[J]. 中国电机工程学报, 2021, 41(23): 7905-7924.

[11]Meyabadi A. F., Deihimi M. H.. A Review of Demand-side Management: Reconsidering Theoretical Framework[J]. Renewable and Sustainable Energy Reviews, 2017, 173(80): 367-379.

[12]Shi H., Wang L., Scherer R., et al. Short-term Load Forecasting Based on Ada Belief Optimized Temporal Convolutional Network and Gated Recurrent Unit Hybrid Neural Network[J]. IEEE Access, 2021, 8(9): 66965-66981.

[13]Kong W., Dong Z. Y., Jia Y., et al. Short-term Residential Load Forecasting based on LSTM Recurrent Neural Network[J]. IEEE Transactions on Smart Grid, 2019, 10(1): 841-851.

[14]赵智永. 基于深度学习的时序电力负荷预测方法研究[D]. 燕山大学,2022.

[15]王金睿. 基于电力市场的短期负荷预测研究与应用[D]. 华北电力大学(北京), 2019.

[16]毛李帆, 江岳春, 龙瑞华, 等. 基于偏最小二乘回归分析的中长期电力负荷预测[J]. 电网技术, 2008, 32(19): 71-77.

[17]廖峰, 刘清良, 贺辉, 等. 基于改进灰色模型与综合气象因素的母线负荷预测[J]. 电网技术, 2011, 35(10): 183-188.

[18]Sagheer A, Kotb M. Unsupervised pre-training of a deep LSTM-based stacked autoencoder for multivariate time series forecasting problems[J]. Scientific reports, 2019, 9(1): 1-16.

[19]Cheng L M, Bao Y Q, Tang L, et al. Very‐short‐term load forecasting based on empirical mode decomposition and deep neural network[J]. IEEJ Transactions on Electrical and Electronic Engineering, 2020, 15(2): 252-258.

[20]黄运有，詹剑锋. 基于深度学习的短期负荷预测综述[J]. 高技术通讯，2021，31(03): 240-248.

[21]Lu C, Li J, Zhang G, et al. A GRU-Based Short-Term Multi-energy Loads Forecast Approach for Integrated Energy System[C]//2022 4th Asia Energy and Electrical Engineering Symposium (AEEES). IEEE, 2022: 209-213.

[22]周彬彬，黄嘉昕，耿冉冉等. 基于灰色-ARIMA 耦合模型的中国发电量短期预测[J]. 科技和产业,2022,22(12):382-387.

[23]张子阳，王珂珂. 基于灰色关联和麻雀搜索算法的电力负荷预测[J]. 辽宁工程技术大学学报(自然科学版),2022,41(03):283-288.

[24]黄世超. 基于CEEMDAN多模型组合的综合能源系统负荷预测研究[D]. 贵州大学，2022.

[25]Wang Y., Chen J., Chen X., et al. Short-Term Load Forecasting for Industrial Customers Based on TCN-Light GBM[J]. IEEE Transactions on Power Systems, 2021, 36(3): 1984-1997.

[26]朱刘柱, 王绪利, 马静等. 基于小波包分解与循环神经网络的综合能源系统短期负荷预测[J]. 电力建设, 2020, 41(12): 131-138.

[27]何宏宇，龚泽玮，李诗颖等. 基于AM-LSTM模型的电力系统短期负荷预测[Z]//自动化与仪器仪表：No. 280. 2023：61-65.

[28]丁宏，陶晓峰，陆春艳等. 基于RWT-SVM的台区配电网日前负荷预测研究[J/OL].南京信息工程大学学报(自然科学版):1-12[2023-04-14].

[29]Xu X, Sun J, Wang C, et al. A novel hybrid CNN-LSTM compensation model against Do S attacks in power system state estimation[J]. Neural Processing Letters, 2022:1-25.

[30]鲁斌, 霍泽健, 俞敏. 基于LSTNet-Skip的综合能源系统多元负荷超短期预测[J/OL]. 中国电机工程报: 1-12[2023-02-15].

[31]Massaoudi M., Refaat S. S., Chihi I., et al. A novel stacked generalization ensemble-based hybrid LGBM-

[32]XGB-MLP model for Short-Term Load Forecasting[J]. Energy, 2021, 21(04): 74-88.

[33]刘岩，彭鑫霞，郑思达. 基于改进LS-SVM的短期电力负荷预测方法研究[J]. 电测与仪表,2021,58(05):176-181.

[34]Masood Z, Gantassi R, Choi Y. A multi-step time-series clustering-based seq2seq lstm learning for a single household electricity load forecasting[J]. Energies, 2022,15(7): 2623.

[35]赵彦，孙俊.一种并行化的改进型灰狼分簇算法[J].电讯技术，2020,60(10):1214-1221.

[36]王勇亮，王挺，姚辰. 基于Kent映射和自适应权重的灰狼优化算法[J]. 计算机应用研究，2020，37(S2):37-40.

[37]马详. 基于门控循环单元神经网络的短期电力负荷预测研究及应用[D]. 天津理工大学，2022.

[38]Aprillia H, Yang H T, Huang C M. Statistical load forecasting using optimal quantile regression random forest and risk assessment index[J]. IEEE Transactions on Smart Grid, 2020, 12(2): 1467-1480.

[39]Shen L, Wang Y. TCCT: Tightly-coupled convolutional transformer on time seriesforecasting[J]. Neurocomputing, 2022, 480: 131-145.

[40]Wang C, Wang Y, Ding Z, et al. A transformer-based method of multienergy loadforecasting in integrated energy system[J]. IEEE Transactions on Smart Grid, 2022,13(4): 2703-2714.

[41]Huang X, Huang X, Wang B, et al. Fault diagnosis of transformer based on modified grey wolf optimization algorithm and support vector machine[J]. IEEJ Transactions on Electrical and Electronic Engineering, 2020, 15(3): 409-417.

[42]Yu M, Xiao C Liu M J, et al. Motor fault diagnosis based on teaching and learning gray wolf algorithm optimized support vector machine[C].Journal of Physics: Conference Series. IOP Publishing, 2022, 2232(1): 012017.

[43]闫群民，董新洲，穆佳豪，马永翔. 基于改进多目标粒子群算法的有源配电网储能优化配置[J]. 电力系统保护与控制，2022，50(10): 11-19.

[44]莫艳红，聂慧，刘振丙，杨辉华. 基于莱维飞行的灰狼优化算法[J]. 微电子学与计算机，2019，36(04):78-83.

[45]伍乙杰，黄文灏，赖仕达等. 基于随机森林和双向长短期记忆网络的超短期负荷预测研究[J]. 电气自动化,2022,44(05):35-37+40.

[46]刘成龙，高旭，曹明. 基于VMD和BA优化随机森林的短期负荷预测[J]. 中国测试,2022,48(04):159-165.

[47]于军琪，聂己开，赵安军等. 基于特征挖掘的ARIMA-GRU短期电力负荷预测[J].电力系统及其自动化学报,2022,34(03):91-99.

[48]孟昭睿. 电力系统短期负荷的组合预测方法研究[D]. 武汉大学，2021.

[49]Chen K, Chen G, Xu D, et al. NAST: non-autoregressive spatial-temporal transformer for time series forecasting[J]. ar Xiv preprint ar Xiv: 2102.05624, 2021.

[50]Zhang H, Liu T, Ye X, et al. Differential evolution-assisted salp swarm algorithm with chaotic structure for real-world problems[J]. Engineering with Computers, 2022: 1-35.

[51]朱瑛，谢睿，郑若池. 基于节点匹配代价优化的随机森林算法[J]. 计算机工程与设计，2020，41(11):3106-3111

[52]王奕森. 随机森林和深度神经网络的若干关键技术研究[D]: [博士学位论文]. 北京: 清华大学，2018

[53]曹正凤. 随机森林算法优化研究[D]: [博士学位论文]. 北京: 首都经济贸易大学，2014

[54]李大海，詹美欣，王振东. 基于多个改进策略的增强麻雀搜索算法[J/OL]. 计算机应用:1-11[2023-04-14].

[55]Gharehchopogh F S, Namazi M, Ebrahimi L, et al. Advances in sparrow search algorithm: a comprehensive survey[J]. Archives of Computational Methods in Engineering, 2023, 30(1): 427-455.

[56]Zhang H, Liu T, Ye X, et al. Differential evolution-assisted salp swarm algorithm with chaotic structure for real-world problems[J]. Engineering with Computers, 2022: 1-35.

[57]Ma H, Shen S, Yu M, et al. Multi-population techniques in nature inspired optimization algorithms: A comprehensive survey[J]. Swarm and evolutionary computation, 2019, 44:365-387.

[58]盛子聿. 基于神经网络的短期负荷预测模型[D]. 西南大学，2021. 王雪菲，李勇，

[59]马骊. 随机森林算法的优化改进研究[D]. 广州：暨南大学，2016.

[60]黄青平，李玉娇，刘松，等. 基于模糊聚类与随机森林的短期负荷预测[J]. 电测与仪表，2017，54(23)：41-46.

[61]刘琪琛，雷景生，郝珈玮，等. 基于Spark平台和并行随机森林回归算法的短期电力负荷预测[J]. 电力建设，2017，38(10)：84-92.

[62]刘达，孙堃，黄晗. 基于EEMD和随机森林的月度负荷预测[J]. 智慧电力，2018，46(6)：12-18.