【摘要】：短期負荷預測(Short-term load forecasting,STLF)是對未來若干小時、1天至幾天的電力負荷預報,作為安排發(fā)購電計劃,經(jīng)濟分配負荷及安排機組出力的基礎,精準的負荷預測是保證電網(wǎng)安全可靠運行的前提條件。隨著居民生活水平的提高,能源消耗加大,調(diào)溫負荷占總用電負荷的比重日益增長,導致電網(wǎng)氣象敏感負荷不斷上升,從而構(gòu)成用電峰荷,拉大了電網(wǎng)峰谷差,現(xiàn)有的短期負荷預測技術在應對復雜氣象條件時預測精度難以滿足電網(wǎng)要求。為落實電網(wǎng)對負荷精細化管理的要求,進一步提高電網(wǎng)負荷預測的精細化水平,確保電網(wǎng)安全穩(wěn)定運行,研究能真實反映負荷變化規(guī)律的負荷預測模型,對于提高短期負荷預測精度十分有必要。自邁入電力大數(shù)據(jù)時代以來,系統(tǒng)原始運行數(shù)據(jù)的存量增加,電力負荷預測技術與相關科學領域技術,如氣象、經(jīng)濟等的交叉滲透。不可置否,大數(shù)據(jù)將是未來電網(wǎng)的生產(chǎn)力,因此在短期負荷預測領域深度挖掘氣象、負荷大數(shù)據(jù)的價值,是融合大能源思維與大數(shù)據(jù)思維研究考慮多因素氣象的負荷預測建模,實現(xiàn)電力負荷精細化管理,提高短期負荷預測精度不可或缺的一部分。本文在電力大數(shù)據(jù)的基礎上,本文首先分析了考慮多因素氣象的負荷特性,從年周期、季周期、日周期等時間維度以及氣象的特殊性方面剖析了氣象對負荷的影響。針對氣象變化時負荷曲線預測精度低,預測模型不能完全適應氣象變化的情況,本文提出了一種提出了完全氣象因子序列的概念,基于數(shù)據(jù)挖掘的方法建立氣象�；�;采用空間多元回歸及滯后模型結(jié)合多策略靈敏度分析法,建立了針對復雜氣象條件下的曲線拐點預測模型;基于改進的K-means聚類分析法查找并獲取氣象特征日,計算初步預測曲線,主動判斷預測曲線畸變概率并進行優(yōu)化修正,得到最佳預測負荷曲線。為應對氣象突變對負荷曲線的影響提出了基于多粒度氣象信息匹配的曲線修正模型,針對突變氣象進行曲線修正。最后利用動態(tài)數(shù)據(jù)流對模型參數(shù)進行更新,實現(xiàn)精細化預測。最后采用本文方法對我國南方某地區(qū)全年負荷曲線進行預測,驗證了模型在多種氣象條件下的預測準確性,尤其適用于短期內(nèi)氣象存在復雜變化的情形。
[Abstract]:Short-term load forecasting (STLF) is a power load forecast for the next few hours or days, which serves as the basis for arranging generation and purchase plans, economic load distribution and generating units. Accurate load forecasting is the precondition to ensure the safe and reliable operation of power grid. With the improvement of residents' living standard, energy consumption is increasing, and the proportion of temperature adjustment load to total power load is increasing day by day, which leads to the rising of meteorological sensitive load of power grid, which forms the peak load of electricity consumption and widens the difference between peak and valley of power grid. The existing short-term load forecasting technology is difficult to meet the requirements of power grid when dealing with complex meteorological conditions. In order to meet the requirement of fine load management, to improve the precision of load forecasting, to ensure the safe and stable operation of power network, a load forecasting model which can truly reflect the law of load change is studied. It is necessary to improve the accuracy of short-term load forecasting. Since entering the era of electric power big data, the stock of the original operation data of the system has increased, and the interpenetration of power load forecasting technology and related scientific fields, such as meteorology, economy, etc. Big data will be the productivity of power grid in the future. Therefore, in the field of short-term load forecasting, the value of load big data is a combination of large energy thinking and big data thinking research, considering multi-factor meteorological load forecasting modeling. It is an indispensable part to realize the fine management of power load and improve the precision of short-term load forecasting. Based on the big data of electric power, this paper first analyzes the load characteristics of multi-factor meteorology, and analyzes the influence of meteorology on the load from the time dimension of annual cycle, season period, daily period and the particularity of meteorology. Because the forecasting precision of load curve is low and the forecasting model can not adapt to the situation of meteorological change, a concept of complete meteorological factor series is put forward in this paper, and the meteorological granulation set is established based on data mining method. By using spatial multivariate regression and lag model combined with multi-strategy sensitivity analysis, the curve inflection point prediction model for complex meteorological conditions is established, and the weather feature days are found and obtained based on improved K-means clustering analysis. The preliminary prediction curve is calculated, the distortion probability of the prediction curve is judged and the optimal load forecasting curve is obtained. In order to deal with the influence of meteorological catastrophe on load curve, a curve correction model based on multi-granularity meteorological information matching is proposed. Finally, the dynamic data stream is used to update the model parameters to achieve fine prediction. Finally, the method of this paper is used to forecast the annual load curve in a certain area of southern China, which verifies the accuracy of the model under various meteorological conditions, especially in the case of complex meteorological changes in the short term.
【學位授予單位】：廣西大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TM715

【參考文獻】

相關期刊論文 前10條

1 李培強;李慧;李欣然;;基于靈敏度與相關性的綜合負荷模型參數(shù)優(yōu)化辨識策略[J];電工技術學報;2016年16期

2 薛禹勝;賴業(yè)寧;;大能源思維與大數(shù)據(jù)思維的融合 (二)應用及探索[J];電力系統(tǒng)自動化;2016年08期

3 吳茵;張智光;苗增強;龔利武;黃柳強;覃芳璐;李濱;;基于標幺化負荷靈敏度的夏季空調(diào)負荷計算[J];電網(wǎng)與清潔能源;2016年02期

4 吳茵;張智光;楊小衛(wèi);龔利武;苗增強;覃芳璐;李濱;;考慮氣象因素的冬季取暖負荷計算[J];電網(wǎng)與清潔能源;2016年01期

5 薛禹勝;賴業(yè)寧;;大能源思維與大數(shù)據(jù)思維的融合(一)大數(shù)據(jù)與電力大數(shù)據(jù)[J];電力系統(tǒng)自動化;2016年01期

6 王繼業(yè);;大數(shù)據(jù):電網(wǎng)企業(yè)創(chuàng)新發(fā)展驅(qū)動力[J];國家電網(wǎng);2015年12期

7 黃歸蘭;趙宇;馬繼華;王慶國;陳陽;;廣西氣象災害預警信號分布特征及發(fā)布[J];氣象科技;2015年02期

8 計鹿飛;江琦;唐昊;譚琦;;基于半馬爾可夫控制過程的智能電網(wǎng)最優(yōu)儲能控制[J];電力系統(tǒng)自動化;2015年06期

9 肖勇;楊勁鋒;馬千里;闕華坤;王家兵;秦州;;基于模塊化回聲狀態(tài)網(wǎng)絡的實時電力負荷預測[J];電網(wǎng)技術;2015年03期

10 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

相關重要報紙文章 前2條

1 李婕茜;;讓大數(shù)據(jù)物盡其用[N];國家電網(wǎng)報;2016年

2 唐金生;;精細化管理是這樣煉成的[N];國家電網(wǎng)報;2007年

相關碩士學位論文 前3條

1 余威;氣象相似性網(wǎng)絡構(gòu)建及缺失氣象要素數(shù)據(jù)的插補[D];西南大學;2015年

2 石雪;基于數(shù)據(jù)挖掘的短期電力負荷預測[D];華南理工大學;2014年

3 劉凱;基于改進BP神經(jīng)網(wǎng)絡的短期負荷預測研究[D];河海大學;2005年

，

本文編號：2126880