| 基于相似性识别的短期负荷动态预测方法 |
| 点此下载全文 |
| 引用本文：陈杰尧,黄炜斌,马光文,陈仕军,谢荻雅.基于相似性识别的短期负荷动态预测方法[J].电网与清洁能源,2020,36(4):1~7 |
| 摘要点击次数: 389 |
| 全文下载次数: 89 |
|
| 基金项目:基金项目：国家重点研发计划项目（2016YFC0402205） |
|
| 中文摘要:统调日负荷的精确预测对电力充裕性保障、电力系统规划有重要指导作用。相似性识别是数据挖掘技术的重要部分，基于相似性识别原理，提出一种短期负荷预测的新方法。首先对原始数据进行属性和重复记录清洗，清洗后得到实验数据；其次在考虑数据大小相似性的同时，引入了数据趋势相似性度量。基于2种度量从历史序列中识别出与查询序列信息高度重合的序列，建立备选相似序列集；考虑时间间隔与季节因素，从备选相似序列集中选取历史最相似序列，最终实现日负荷预测；随着新信息的进入，实现高峰负荷动态预测。利用该方法对重庆统调日负荷进行模拟预测，并与BP神经网络和支持向量机方法进行对比，证明了所提方法的可行性与有效性。 |
| 中文关键词:短期负荷预测 动态预测 相似性识别 数据趋势相似 数据清洗 |
| |
| A Short-Term Load Dynamic Prediction Method Based on Similarity Recognition |
|
|
| Abstract:The adjustment of daily load is an important manifestation of the energy consumption level of various sectors of the national economy and urban and rural residents. The accurate prediction of the load plays an important guiding role in power adequacy guarantee and power system planning. The development of the power spot market puts higher requirements on the accuracy of load prediction. Similarity recognition is an important part of data mining technology. Based on the principle of similarity recognition， this paper proposes a new method for short-term load prediction. A large number of data records always have errors. In this paper， the original data is firstly cleaned by attributes and duplicate records， and the experimental data is obtained after cleaning. Then， while the similarity and size of the data are being considered， a data trend similarity measure is introduced. A standby similar sequence set is established by identifying sequences that are highly coincident with the query sequence information from the historical sequence based on the two measures. Considering the time interval and seasonal factors， the most similar sequence is selected from the candidate similar sequence set， and the daily load forecast is finally realized. With the entry of new information， dynamic load forecasting of peak loads is achieved. This method is used to simulate and predict the daily load of Chongqing， and is compared with BP neural network and support vector machine method. The feasibility and effectiveness of the proposed method are proved. |
| keywords:short term load forecasting dynamic forecasting similarity identification data trend similarity data cleaning |
| 查看全文 查看/发表评论 下载PDF阅读器 |
