在競爭激烈的零售市場中,電源商正在尋求通過智能電表分析客戶的每一個用電數(shù)據(jù),這將為他們提供大量機會,以便在非常大量的智能電網(wǎng)中實現(xiàn)對客戶電力消耗需求的額外了解。通常,有大量的分析解決方案來呈現(xiàn)家庭的設(shè)施使用情況,但這些類型的解決方案并未提供準(zhǔn)確的信息。因此,我們嘗試對個體家庭能源消費模式進行全面分析,并設(shè)計一個家庭層面預(yù)測模型,利用歷史能耗數(shù)據(jù)預(yù)測未來有價值的實際需求和相應(yīng)的有關(guān)需求。本文提出了一種完全獨特的方法來預(yù)測配電系統(tǒng)中的輸電時間序列分析,該方法顯示了不同消費行為的比例,以及相鄰時段內(nèi)不同時段的消費水平。所提出的方法預(yù)測了客戶使用智能電表管理其家庭電力數(shù)據(jù)收集的合法性,并幫助客戶系統(tǒng)操作員檢測和控制負(fù)載需求。該模型在大型數(shù)據(jù)集中查找不同時期的各種功率趨勢。通過使用時間序列數(shù)據(jù)方法和預(yù)測模型的預(yù)測來進行評估。結(jié)果表明,具有良好準(zhǔn)確性的預(yù)測可以幫助公司和最終用戶通過將功耗從高峰時段轉(zhuǎn)移到非高峰時段來控制其負(fù)載需求。應(yīng)用知識發(fā)現(xiàn)回歸模型,在一周內(nèi)明顯改善了電力趨勢消費,并幫助用戶改善客戶需求,如節(jié)能,低價和管理。

【文章頁數(shù)】：72 頁

【學(xué)位級別】：碩士

【文章目錄】：
Abstract
摘要
Abbreviations
Chapter1 General Introduction
    1.1 Purpose and significance of research
    1.2 Related Technology
        1.2.1 Big data
        1.2.2 Analytics
        1.2.3 R and RStudio tools for programming
    1.3 Foundation of research and development
    1.4 Research Methods
    1.5 Work summary
Chapter2 Preliminaries
    2.1 Introduction
    2.2 Data Mining Theory
        2.2.1 Introduction
        2.2.2 Types of data mining
        2.2.3 Data mining process
    2.3 Time Series Analysis
        2.3.1 Time series classification
        2.3.2 Time series aim
        2.3.3 Times series components
        2.3.4 Time series forecasting
            2.3.4.1 Forecasting without external factors
            2.3.4.2 Forecasting with external factors
        2.3.5 Forecasting accuracy
        2.3.6 Data preprocessing
            2.3.6.1 Outliers detection
            2.3.6.2 Denoising and Smoothing
            2.3.6.3 Differencing
            2.3.6.4 Data scaling
            2.3.6.5 Normalization
    2.4 Forecasting Methods
        2.4.1 Regression models
        2.4.2 Autoregressive and moving average models
        2.4.3 Exponential smoothing models
        2.4.4 Artificial neural networks models
        2.4.5 Markov chain model
    2.5 Load forecast household electricity
Chapter3 Design of Analysis System
    3.1 Introduction
    3.2 Data collection
        3.2.1 Description of dataset
        3.2.2 Electrical smart meters
        3.2.3 Measurement description
        3.2.4 Data set attribute information
    3.3 Data preparation and preprocessing
    3.4 Feature selection and modeling
        3.4.1 Data visualization and transformation
        3.4.2 The annual household electricity consumption
    3.5 Time series data
        3.5.1 Time series concepts
        3.5.2 Decomposition of time series
            3.5.2.1 Change the data format
            3.5.2.2 Analytics data exploration
    3.6 Construction and time series forecasting model
        3.6.1 Automated model time series forecasting ETS(A,N,A)
        3.6.2 Forecasting method
            3.6.2.1 Exponential smoothing
            3.6.2.2 Simple exponential smoothing(SES)
            3.6.2.3 Forecasting for further more times points smoothing
        3.6.3 ARIMA models
        3.6.4 Advanced forecasting methods
        3.6.5 Prediction model evaluation
    3.7 Monthly trend and forecasting results
    3.8 Knowledge discovery regression model
        3.8.1 Case study
        3.8.2 Plotting power consumption
        3.8.3 Implementation over one week
        3.8.4 Determine the trend of weekly consumed energy
Chapter4 Discussion and Conclusion
    4.1 Discussion
    4.2 Conclusion
Reference
Acknowledgement
Appendices



本文編號：3969120