本文選題：負荷聚類 + 負荷預測�。� 參考：《華北電力大學(北京)》2017年碩士論文

【摘要】：負荷預測一直以來都是電力系統(tǒng)的一項重要工作,準確的負荷預測可以經(jīng)濟合理地安排電力系統(tǒng)發(fā)電機組啟停,對于保持電網(wǎng)的安全穩(wěn)定運行、保持社會的正常生產(chǎn)秩序、有效降低發(fā)電成本有著重要作用。隨著智能電網(wǎng)技術的發(fā)展,高級量測體系和各種監(jiān)控系統(tǒng)大規(guī)模的部署。智能電表是高級量測體系的重要組成部分,能夠獲得用戶一定時間間隔內精確的用電負荷。智能電表產(chǎn)生數(shù)據(jù)的速度快、體量大,產(chǎn)生的數(shù)據(jù)沒有進行深入分析,造成了數(shù)據(jù)的浪費。因此充分挖掘用電數(shù)據(jù)的價值,研究智能用電大數(shù)據(jù)環(huán)境下的短期負荷預測具有重要意義。針對智能電表能獲取用戶級詳細用電數(shù)據(jù)的特點,本文從少量用戶數(shù)據(jù)入手,首先通過負荷聚類,分析了用戶用電行為之間的相似性;在此基礎上提出了基于OS-ELM的短期負荷預測模型,并通過仿真實驗驗證了所提模型能夠提升負荷預測精度且進一步展示了聚類結果與預測精度之間的關系;之后為適應智能用電大數(shù)據(jù)環(huán)境,進一步提出了基于Spark的并行OS-ELM短期負荷預測模型,并在實驗中驗證了模型能在保證預測精度的前提下具有較高的效率。本文具體工作如下:1.研究了用于負荷聚類的日期類型因素。針對不同的日期類型(普通工作日、節(jié)假日前一天、節(jié)假日)分別計算用戶的典型日負荷,把三種典型日負荷曲線拼接起來作為用戶的典型負荷曲線;然后對用戶典型負荷曲線進行聚類操作,挖掘用戶用電行為之間的相似性。2.針對少量用戶數(shù)據(jù),提出了基于OS-ELM的短期負荷預測模型。在負荷聚類基礎上,對不同的用戶類分別采用該負荷預測模型進行負荷預測并匯總得到系統(tǒng)級的負荷預測。在MATLAB平臺上進行仿真實驗,驗證了所提模型的有效性,并進一步展示了預測精度隨聚類數(shù)目變化的關系。3.針對海量用戶數(shù)據(jù),進一步提出了基于Spark的并行OS-ELM短期負荷預測模型。為了適應智能用電大數(shù)據(jù)環(huán)境,在所提基于OS-ELM的預測模型基礎上,提出了基于Spark的并行OS-ELM短期負荷預測模型,并在仿真實驗中驗證了所提并行預測模型在保證負荷預測精度的前提下仍具有較高的運行效率。
[Abstract]:Load forecasting has always been an important work of the power system. Accurate load forecasting can reasonably arrange power system generator set up and stop, which plays an important role in maintaining the safe and stable operation of the power grid, maintaining the normal production order of the society and reducing the cost of power generation effectively. The intelligent meter is an important part of the advanced measurement system. The intelligent meter is an important part of the advanced measurement system. It can obtain the accurate power load of the user in a certain time interval. The speed of the data is fast and the volume is large. The data produced is not deeply analyzed, and the data is wasted. Therefore, it is fully dug. It is of great significance to study the value of electrical data and to study the short-term load forecasting in the intelligent data environment. In view of the characteristics of the intelligent meter, which can obtain the detailed user level data, this paper begins with a small amount of user data, and first analyzes the similarity between the user's electricity use behavior through a small amount of user data. On this basis, the basis is put forward. In the short term load forecasting model of OS-ELM, the simulation experiments show that the proposed model can improve the load forecasting precision and further demonstrate the relationship between the clustering results and the prediction accuracy. Then, the parallel OS-ELM short-term load forecasting model based on Spark is further proposed to adapt to the intelligent data environment. It is proved that the model can have high efficiency on the premise of guaranteeing the prediction accuracy. The specific work of this paper is as follows: 1. the date type factors for load clustering are studied. For different date types (ordinary working day, holiday day, holiday), the typical daily load of users is calculated respectively, and three typical daily load curves are spliced. As a typical load curve of the user, the user's typical load curve is clustered, and the similarity between the user's electrical behavior is excavated for a small amount of user data, and a short-term load forecasting model based on OS-ELM is proposed. On the basis of load clustering, the load forecasting model is used for the different user classes to carry out the load forecasting model respectively. The load forecast is predicted and summarized. The simulation experiments on the MATLAB platform verify the validity of the proposed model, and further demonstrate the relationship between the prediction accuracy and the number of cluster numbers..3. is further proposed for the parallel OS-ELM short-term load forecasting model based on Spark. On the basis of the prediction model based on OS-ELM, a parallel OS-ELM short-term load forecasting model based on Spark is put forward on the basis of the forecast model based on the electric big data. It is proved that the proposed parallel prediction model still has a high operating efficiency under the premise of guaranteeing the precision of load forecasting.
【學位授予單位】：華北電力大學(北京)
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP311.13;TM715

【參考文獻】

相關期刊論文 前10條

1 鄧萬宇;楊麗霞;;基于Spark的OS-ELM并行化算法[J];西安郵電大學學報;2016年02期

2 吳哲夫;張彤;肖鷹;;基于Spark平臺的K-means聚類算法改進及并行化實現(xiàn)[J];互聯(lián)網(wǎng)天地;2016年01期

3 劉念;張清鑫;劉海濤;;基于核函數(shù)極限學習機的微電網(wǎng)短期負荷預測方法[J];電工技術學報;2015年08期

4 王繼業(yè);季知祥;史夢潔;黃復鵬;朱朝陽;張東霞;;智能配用電大數(shù)據(jù)需求分析與應用研究[J];中國電機工程學報;2015年08期

5 高賜威;李倩玉;蘇衛(wèi)華;李揚;;短期負荷預測中考慮積溫效應的溫度修正模型研究[J];電工技術學報;2015年04期

6 王德文;孫志偉;;電力用戶側大數(shù)據(jù)分析與并行負荷預測[J];中國電機工程學報;2015年03期

7 張素香;趙丙鎮(zhèn);王風雨;張東;;海量數(shù)據(jù)下的電力負荷短期預測[J];中國電機工程學報;2015年01期

8 張東霞;苗新;劉麗平;張焰;劉科研;;智能電網(wǎng)大數(shù)據(jù)技術發(fā)展研究[J];中國電機工程學報;2015年01期

9 劉科研;盛萬興;張東霞;賈東梨;胡麗娟;何開元;;智能配電網(wǎng)大數(shù)據(jù)應用需求和場景分析研究[J];中國電機工程學報;2015年02期

10 趙騰;張焰;張東霞;;智能配電網(wǎng)大數(shù)據(jù)應用技術與前景分析[J];電網(wǎng)技術;2014年12期

相關碩士學位論文 前1條

1 羅伶;基于模糊c均值聚類的分類分時電價研究[D];山東大學;2013年

，

本文編號：2113621