| 引用本文: | 梁志峰,康重庆,孙大雁,钟海旺,王建平,周海松,杨登宇.基于重大天气过程温度等势面波动的省级负荷预测[J].电力自动化设备,2026,46(2):176-184. |
| LIANG Zhifeng,KANG Chongqing,SUN Dayan,ZHONG Haiwang,WANG Jianping,ZHOU Haisong,YANG Dengyu.Provincial load forecasting based on temperature isosurface fluctuation during major weather process[J].Electric Power Automation Equipment,2026,46(2):176-184. |
|
| 本文已被：浏览 130次 下载 63次 |
码上扫一扫！ |
|
|
| 基于重大天气过程温度等势面波动的省级负荷预测 |
|
梁志峰1,2, 康重庆1,3, 孙大雁2, 钟海旺1,3, 王建平4, 周海松4, 杨登宇4
|
|
1.清华大学 电机工程与应用电子技术系，北京 100084;2.国家电网有限公司 国家电力调度控制中心，北京 100031;3.清华大学 清华四川能源互联网研究院，四川 成都 610213;4.南瑞集团有限公司(国网电力科学研究院有限公司)，江苏 南京 211106
|
|
| 摘要: |
| 重大天气过程期间省级系统负荷的精准预测对于新型电力系统的电力电量平衡至关重要。针对这一问题，提出了一种使用温度等势面图替代多站点温度的方法，将传统温度数据与负荷的关系转变为温度等势面图与负荷的关系，利用深度神经网络的图形识别和时序记忆能力，分析温度等势面图集合与负荷集合的复杂对应关系，构建了能够捕捉温度等势面波动与系统负荷变化规律的通用预测模型。在此基础上，采用迁移学习方法建立了寒潮、高温两类重大天气过程的日前96点系统负荷预测专项模型，分析了这两类重大天气过程的典型温度等势面图特征。基于某省2021年至2023年寒潮、高温期间的历史数据开展算例分析，结果表明负荷预测精度显著提升，验证了所提方法的有效性。 |
| 关键词: 重大天气过程 省级电网 负荷预测 等势面图 卷积神经网络-长短期记忆神经网络 人工智能 |
| DOI：10.16081/j.epae.202509022 |
| 分类号: |
| 基金项目:国家电网有限公司科技项目(4000-202355381A-2-3-XG) |
|
| Provincial load forecasting based on temperature isosurface fluctuation during major weather process |
|
LIANG Zhifeng1,2, KANG Chongqing1,3, SUN Dayan2, ZHONG Haiwang1,3, WANG Jianping4, ZHOU Haisong4, YANG Dengyu4
|
|
1.Department of Electrical Engineering, Tsinghua University, Beijing 100084, China;2.State Grid Dispatching Control Centre, State Grid Corporation of China, Beijing 100031, China;3.Sichuan Energy Internet Research Institute, Tsinghua University, Chengdu 610213, China;4.State Grid Electric Power Research Institute, NARI Group Corporation, Nanjing 211106, China
|
| Abstract: |
| The accurate forecasting of provincial load during major weather process is pivotal to balance the electric power and energy for new electric power system. Aiming at this problem, a method using temperature isosurface diagram instead of multi-station temperature data is proposed. The traditional relationship between temperature data and load is transformed into the relationship between temperature isosurface diagrams and load. Utilizing the graphical recognition and temporal memory abilities of deep neural network, the complex correspondence between temperature isosurface diagram set and load set are analyzed, a general prediction model enabled to capture the law of temperature isosurface fluctuation and system load change is established. Based on this general model, the day-ahead 96-point load forecasting models during cold wave and high-temperature weather are established using transfer learning methods, the typical temperature isosurface characteristics of these two kinds of weather events are analyzed. Based on the history data during cold wave and high-temperature periods in a certain province from 2021 to 2023, the load prediction accuracy is significantly enhanced, which validated the effectiveness of the proposed method. |
| Key words: major weather process provincial power grid load forecasting isosurface diagram CNN-LSTM neural network artificial intelligence |
