Scribd Logo
Upload

Welcome to Scribd!

  • Energy Consumption Forecasting Model For Puerto Princesa Distribution System Using Multiple Linear Regression

    Uploaded by

    International Journal of Innovative Science and Research Technology
    63 views4 pages
    Power system engineers widely consider electric load forecasting because of its vital role in economically optimizing and securing the efficient operation of the power system

    Original Title

    Energy Consumption Forecasting Model for Puerto Princesa Distribution System Using Multiple Linear Regression

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    Power system engineers widely consider electric load forecasting because of its vital role in economically optimizing and securing the efficient operation of the power system

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Flag for inappropriate content
    63 views4 pages

    Energy Consumption Forecasting Model For Puerto Princesa Distribution System Using Multiple Linear Regression

    Uploaded by

    International Journal of Innovative Science and Research Technology
    Power system engineers widely consider electric load forecasting because of its vital role in economically optimizing and securing the efficient operation of the power system

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Flag for inappropriate content
    of 4

    Volume 5, Issue 11, November – 2020 International Journal of Innovative Science and Research

    Technology
    ISSN No:-2456-2165

    Energy Consumption Forecasting Model for


    Puerto Princesa Distribution System Using
    Multiple Linear Regression
    Alfred Rey G. Vasquez1, Michael Ernie F. Rodriguez2, Roy C. Dayupay3
    1
    Palawan State University, Philippines
    2
    Palawan State University, Philippines
    3
    Palawan State University, Philippines

    Abstract:- Power system engineers widely consider the short-term, medium-term, and long-term forecasts.
    electric load forecasting because of its vital role in Short-term forecast is used for hourly and weekly predictions,
    economically optimizing and securing the efficient the medium-term forecast is for monthly predictions, and the
    operation of the power system. A forecast can be utilized long-term forecast is for yearly predictions. Short-term
    by electric utilities to upgrade and improve the existing forecast is utilized by Dmitri et al. [2] and Srivastava et al. [3]
    distribution facilities. Also, through this prediction, and Singla et al. [4] while medium-term forecast is utilized in
    future developments could be planned concerning the study of Tay et al. [5]. Long-term forecast is used in the
    generation and transmission facilities. In this paper, the studies on [6]-[10].
    annual energy consumption of the Puerto Princesa
    Distribution System for the year 2019-2028 was Regression analysis is the modeling technique utilized
    forecasted using multiple linear regression. The peak in load forecasting to analyze the relationships of the different
    demand and the number of consumers were the variables variables [6]-[8]. Simple linear regression is used by
    considered for the regression analysis. From the error Khamaira et al. [1] and Ade-Ikuesan et al. [8], while multiple
    performance test, the results indicate that multiple linear linear regression is used in the studies on [6]-[7], and
    regression is a useful technique for long-term load [11]-[14]. Different variables are considered in performing
    forecasting, having a minimum percent error. Based on regression analysis, such as population, gross domestic
    the regression results, the energy consumption by 2028 is product (GDP), load demand and electricity cost [6]-[7],
    expected to be 566,078,019.1 kWh. The error [11]-[13].
    performance test demonstrates that the mean average
    percent error of 0.74% which indicates that the multiple In this paper, the energy consumption of the Puerto
    linear regression model is a good fit. Princesa Distribution System for the year 2019-2028 was
    forecasted using a multiple linear regression model. This
    Keywords:- Distribution System, Energy Consumption mathematical model considered variables such as peak
    Forecasting, Long-Term Forecast, Multiple Linear demand and the number of consumers.
    Regression.
    II. METHODOLOGY
    I. INTRODUCTION
    The historical data was collected from the utility
    Electricity is one of the fundamental needs and an company to forecast the energy consumption in the next 10
    essential resource in sustaining life that people utilize every years. The historical data provided by Palawan Electric
    day. The electricity demand in the world is anticipated to Cooperative (PALECO) are the number of consumers, peak
    grow due to the dependence on electricity of humanity to demand, and energy consumption for the year 2014-2018.
    perform different tasks. To prepare for the electricity demand
    growth, load forecasting is conducted to estimate the future
    demand for electricity.

    In [1], there are three types of load forecasting, which is

    IJISRT20NOV062 www.ijisrt.com 37
    Volume 5, Issue 11, November – 2020 International Journal of Innovative Science and Research
    Technology
    ISSN No:-2456-2165
    2.2 Error Performance Test
    For validation purposes, the error performance test of
    the forecast model is conducted [6]. Once all the independent
    variables are correctly identified, the error ε, sum of squares
    error (SSE), and the total sum of squares (TSS) are calculated
    as shown in Eqs. (3-5).

      yi  yi
    ˆ (3)

    SSE  i 1 ( yi  yi
    n
    ˆ )2 (4)

    TTS  i 1 ( yi  yi)2
    n
    (5)

    where yi is the actual value, ŷi is the predicted value, and yi


    is the average value.

    After solving the SSE and TSS, the coefficient of


    determination (R2) and the adjusted R2 (AdjR2) is then solved.
    The coefficient of determination measures the regression
    model as a whole, and this determines the acceptability of the
    model. The closer R2 to 1, the better the model is, and it tells
    how well the estimated regression is. The adjusted R2 is the
    calculated R2 from those variables, which is significant to the
    Fig 1: Process Flowchart model only.

    As defined earlier, regression is a modeling technique to SSE


    R2  1  (6)
    analyze the relationship between a dependent and one or more TTS
    independent variables. It aims to identify a function that
    describes the relationship between these variables as close as  SSE  n  1 
    possible. Using multiple linear regression (MLR), the energy AdjR 2  1     (7)
     TTS  n  k  1 
    consumption was found in terms of the independent variables
    that affect the energy consumption.
    where n is the number of data points, and k is the
    number of independent variables. The values of R2 and AdjR2
    2.1 Forecasting Using the MLR Model
    The MLR model can be expressed as ranges from 0-1, wherein a value closer to one means that the
    data fits better with the estimated function.
    Y  A  B1 X1  B2 X 2  B3 X 3  ...  Bn X n (1)
    Next, the t-statistic and P-value is determined.
    T-statistic shows the significance of each explanatory
    where Y refers to the energy consumption, the A, B1, B2, variable in predicting the dependent variable, and it has a
    B3, and Bn are the unknown regression coefficients, and X1, generally accepted value of greater than 2 or less than −2 for
    X2, X3, and Xn are the historical variables. The unknown each variable. P-value is the indicator for the probability if the
    coefficient in Eq. (1) can be calculated using a multiple parameter of population is equal to zero, and if it is equal to
    regression approach by minimizing the sum of the squares of 0.1, it indicates a significant regression [10].
    the projected errors. Equation (1) can be expressed for two
    historical variables, and a matrix is used to determine Finally, in conducting an error performance test,
    coefficients: determining the Mean Absolute Percentage Error (MAPE) is
    needed. A MAPE with a value of less than 5% indicates
    Y  A  B1 X1  B2 X 2 (2) excellent accuracy [15].

    The Analysis ToolPak of Microsoft Excel® is used in yactual  yapprox



    n
    determining the unknown regression coefficients A, B1 and i 1
    yactual
    B2. To forecast the energy consumption, the peak demand X1 MAPE  (8)
    n
    and the number of consumers X2 is predicted for 10 years by
    calculating for the average annual growth rate (AAGR) from
    the year 2014-2018.
    III. RESULTS AND DISCUSSION

    IJISRT20NOV062 www.ijisrt.com 38
    Volume 5, Issue 11, November – 2020 International Journal of Innovative Science and Research
    Technology
    ISSN No:-2456-2165
    The historical data available are shown in Table 1 were B2 can be calculated, and the forecasting model can be
    increasing peak load demand was observed. The peak written as:
    demand and number of consumers were considered as
    independent variables X1 and X2, respectively, while energy Y  102364679.6650  3634.1012( X1 )  3732.0936( X 2 ) (9)
    consumption represents the dependent variable Y. Using
    Equations 1 and 2, regression model coefficients A, B1, and

    Table 1: Historical Data of Energy Consumption,


    Peak Demand, and Number of Consumers
    Peak demand (kW) No. of Consumers Energy Consumption (kWh)
    Year
    X1 X2 Y
    2014 29,310 38,885 149,258,474
    2015 30,440 41,353 161,749,102.1
    2016 31,580 44,350 180,486,933.6
    2017 34,530 46,710 194,850,933.6
    2018 39,120 48,336 221,079,815.6

    The average annual growth rate (AAGR) is obtained peak demand is found using the calculated AAGR, future
    from the historical data to predict the peak demand and energy consumption can be forecasted using Eq. (9). Results
    number of consumers. The calculated AAGR was 7.56% and obtained are shown in Table 2 and Fig. 2, which imply an
    5.60% for peak demand and the number of consumers, increasing energy demand through the succeeding years from
    respectively. When the predicted number of consumers and 2019 to 2028.

    Table 2: Forecasted Energy Consumption for the Year 2019-2028


    Forecasted Energy Consumption
    Year Predicted Peak Demand (kW) Predicted No. of Consumers
    (kWh)
    2019 42,076.93957 51,989.54374 244,577,019.3
    2020 45,257.38352 55,919.24565 270,801,089.7
    2021 48,678.22576 60,145.97955 299,007,343.0
    2022 52,357.63710 64,692.19701 329,345,605.2
    2023 56,315.16187 69,582.04663 361,977,027.1
    2024 60,571.82164 74,841.50233 397,074,940.3
    2025 65,150.22697 80,498.50129 434,825,778.1
    2026 70,074.69742 86,583.09239 475,430,065.4
    2027 75,371.39081 93,127.59576 519,103,484.3
    2028 81,068.44213 100,166.77451 566,078,019.1

    Fig 2: Historical and Forecasted Energy


    Consumption for the Year 2014-2028

    IV. CONCLUSION to energy consumption of Puerto Princesa Distribution


    system from 2014 to 2028. The results obtained are
    The utilization of multiple linear regression was applied summarized in Table 2 and presented as a graphical form in

    IJISRT20NOV062 www.ijisrt.com 39
    Volume 5, Issue 11, November – 2020 International Journal of Innovative Science and Research
    Technology
    ISSN No:-2456-2165
    Figure 2. The annual peak demand and the number of [10]. S. R. Khuntia, J. L. Rueda, and M. A. M. M. van der
    consumers recorded from the electric utility between the Meijden. Long term load forecasting considering
    years 2014 and 2018 were the variables used. Based on the volatility using multiplicative error model, in
    regression results, the energy consumption by 2028 is Energies, vol. 11, 2018, 3308 (3-5).
    expected to be 566,078,019.1 kWh. Moreover, the error [11]. R. Akan, S. N. Keskin, and S. Uzundurukan. Multiple
    performance test demonstrates a coefficient of determination regression model for the prediction of unconfined
    (R2) and an adjusted R2 value of 0.995 and 0.991, compressive strength of jet grout columns, in
    respectively, with a mean average percent error of 0.74%, Procedia Earth and Planetary Science, Elsevier Ltd.,
    indicating that the multiple linear regression model is a good vol. 15, 2015, pp. 299-303.
    fit. Furthermore, results obtained from this study may be used [12]. N. Amral, C. S. Özveren, and D. King. Short term load
    to future studies applying different forecasting techniques forecasting using multiple linear regression, in Proc.
    such as exponential smoothing, artificial neural network and Univ. Power Eng. Conf., 2007, pp. 1192–1198.
    MatLab. [13]. B. Dhaval and A. Deshpande. Short-term load
    forecasting with using multiple linear regression,”
    REFERENCES Int. J. Electr. Comput. Eng., vol. 10, no. 4, 2020, pp.
    3911–3917.
    [1]. M. Y. Khamaira, A. S. Krzma, and A. M. Alnass. Long [14]. N. A. B. Kamisan, M. H. Lee, S. Suhartono, A. G.
    term peak load forecasting for the libyan network, in Hussin, and Y. Z. Zubairi. Load forecasting using
    Proceedings of First Conference for Engineering combination model of multiple linear regression
    Sciences and Technology, vol. 1, 2018, pp. 186-189. with neural network for Malaysian City in Sains
    [2]. K. Dmitri, A. Maria, and A. Anna. Comparison of Malaysiana, vol. 47, no. 2, 2018, pp. 419–426.
    regression and neural network approaches to [15]. R. B. Patel, M. R. Patel, and N. A. Patel. Electrical load
    forecast daily power consumption, in Proc. - 2016 forecasting using machine learning methods, RNN
    11th Int. Forum Strateg. Technol. IFOST 2016, 2017, and LSTM, in Journal of Xidian University, vol. 14 no.
    pp. 247–250, 2017. 4, 2020, pp. 1376-1386.
    [3]. A. K. Srivastava, D. Singh, and A. S. Pandey. Short
    term load forecasting using regression trees:
    random forest, bagging and M5P, in Int. J. Adv.
    Trends Comput. Sci. Eng. vol. 9, no. 2, 2020, pp.
    1898–1902.
    [4]. M. K. Singla, J. Gupta, P. Nijhawan, and A. S. Oberoi.
    Electrical load forecasting using machine learning,
    in Int. J. Adv. Trends Comput. Sci. Eng., vol. 8, no. 3,
    2019, pp. 615–619.
    [5]. K. G. Tay, A. Huong, W. K. Tiong, and S. Saleh.
    Electricity consumption forecasting using fuzzy
    inference system, in Int. J. Adv. Trends Comput. Sci.
    Eng., vol. 9, no. 1, 2020, pp. 249–254.
    [6]. Z. Mohamed and P. Bodger. Forecasting electricity
    consumption in New Zealand using economic and
    demographic variables, in Energy, Elsevier Ltd., vol.
    30, 2005, pp. 1833–1843.
    [7]. M. Krutsyak. Forecasting demand on the Ukrainian
    electricity market using socio-economic variables, in
    Economics, Management and Sustainability, vol. 4, no.
    1, 2009, pp. 46-57.
    [8]. O. Ade-Ikuesan, A. Oyedeji, and M. Osifeko. Linear
    regression long-term energy demand forecast
    modelling in Ogun State, Nigeria, in Journal of
    Applied Sciences and Environmental Management, vol.
    23, no. 4, 2019, 753-757.

    [9]. K. R. M. Supapo, R. V. M. Santiago, and M. C. Pacis.


    Electric load demand forecasting for
    Aborlan-Narra-Quezon distribution grid in
    Palawan using multiple linear regression, in 2017
    IEEE 9th International Conference on Humanoid,
    Nanotechnology, Information Technology,
    Communication and Control, Environment and
    Management (HNICEM), Manila, 2017, pp. 1-6.

    IJISRT20NOV062 www.ijisrt.com 40

    You might also like