Modeling and Forecasting of Electricity Consumption Using Arima and Exponential Smoothing
CHAPTER ONE
Aim and Objectives of the study:
The aim of this research work is to fit a time series forecasting modelto monthly electricity consumption (mwh).
The objectives of the study are as follows
- Check for existence of unit
- Investigate the behavioral change in the data using ARMA and Exponential Smoothingprocesses
- Evaluate the accuracy of the models for forecasting
CHAPTER TWO
LITERATURE REVIEW
Introduction
This chapter gives literature reviews on the works of different scholars in Time series analysis. Several research works have now used the forecasting method with time series data such as the electric power consumption.
The Econometric Modeling Approach
The modeling of energy demand in econometric is plan with the aim to analyse relationship between variables. The past relationship can be analysed based on the observed relationships, estimating the effect of changes of the independent variable(s) on the dependent variable. The modeling technique of econometric has been frequently used for analysing energy demand because, of the availability of observations. It can be applied with sufficiently long historical observations on energy consumption, and explanatory variables such as population, income, and prices. For the end-use and input-output modeling approaches, the main strategy is the homogenous grouping of consumers in order to model common characteristics of the energy demand of these homogenous consumer groups (industrial, residential, etc.)
Although this strategy is utilized by the econometric modeling approach, the main difference between this and the two other approaches is that the econometric modeling approach statistically estimates energy demand relationships; the end-use and input-output approaches rely on energy surveys and technical studies which are not always available. One of the reasons that the econometric approach is arguably more attractive than the other approaches is that the econometric approach has a strong theoretical background consistent with economic theory (in particular consumer and production theory). In the econometrics literature there are several functional forms which have been developed for energy demand modeling such as the trans-log model (most often applied to a demand system).
CHAPTER THREE
METHODOLOGY
Time Series Model
The procedure employed in analyzing data is presented in this section. Time series linear models have been proposed and compared for forecasting electricity consumption (mwh). Moreover, the methods for model identification in forecasting ability have also been discussed. Time series can be defined as a chronological series of observations (Montgomery et al., 2008). The time can be a discrete, or a continuous function. Time series data can be deterministic, random, or a combination of the two (Gupta, 1989). Various statistical methods deal with models in which the observed values are assumed to be independent. However, a quite set of data such as in engineering, natural sciences etc occur in the form of time series where observed values are dependent. However, time series analysis is a systematic approach available to tackle the situation of dependency by observations. Serial correlation coefficient or the autocorrelation coefficient play important role in the analysis of time series which indicates the dependence in its successive values.
Correlogram is a kind of graph that displays the autocorrelation coefficient against the lag period. The process is purely random if the correlogram shows nearly or a zero values for the lag periods, while a value close to one indicates deterministic process (Gupta, 1989). The purpose of analysis of electric consumption is not only to analyze a time series but to generate the data based on the series. Statistical properties are very vital in order to reproduce series of similar statistical characteristics (Gupta, 1989).Electric consumption has a random component being denoted by the word “stochastic”. Stochastic modeling of electric consumption has been widely used for planning and management of power resources and forecasting the occurrence of future consumption. Stochastic models can be used for forecasting in days, weeks, months and years (Fortin et al., 2004). The previous information can be used as model inputs for forecasting the future event.
CHAPTER FOUR
RESULTS AND DISCUSSION
Data Analysis and Discussion of Findings
Introduction
In this chapter data is being analyzed, discussed and commented using ARMA and exponential smoothing modeling methods for forecasting of electricity consumption (mwh). The models will be checked to get a suitable model to fit the data. Data points from January 2011 to February 2015 are used as a training set while data points from March 2015 to December 2015 are used for validation.
CHAPTER FIVE
SUMMARY, CONCLUSION AND RECOMMENDATION
Summary
Augmented Dickey fuller test was performed to check for presence of unit root and was found that the series was stationary with an indication of trend and seasonal effect. The study compared the forecasting ability of Autoregressive and exponential smoothing techniques on electricity consumption (mwh). However, loss function was used as a basis for the comparison and hence Holt-winters multiplicative outperformed seasonal autoregressive moving average.
Conclusion
The objectives of the study is to propose a forecasting model for monthly electricity consumption using ARMA and Exponential Smoothing approaches in order to investigate their forecasting ability. Based on the findings, the tentative competing models that satisfied the criteria are Seasonal ARMA (1,1) (1,1)[12] and Holt-winters multiplicative. However, the results show that Holt-winters multiplicative emerged as the best forecasting method.
Recommendation
The result show an increase in the consumption which is probably as a result of rapid increase in population therefore, government should make proper arrangement in order to face the challenge thereby providing basic devices.
REFERENCES
- Akaike, H., (1969): Fitting Autoregressive Models for prediction. Annals of the institute of statistical mathematics, 21, 243-247
- Akaike, H., (1973): Information theory as an extension of the Maximum likelihood principle.
- Secondinternational Symposium on information theory, pp: 265-282
- Akaike, H.,(1974): A new look at the statistical model identification. IEEE Transactions on Automatic Control, 19(6):716-723
- Aras, H., and Aras, N., (2004): Forecasting Residential Natural Gas Demand. Energy Sources
- 26:463-472.
- Barron, D. N., (1992): The Analysis of Count Data: Over dispersionand Autocorrelation, Social Methodology, 22, pp 179-220.
- Bergstrom, A.R., (1990): Continuous-Time Econometric Modeling.Oxford University Press.
- Bowerman, B. L.,& Connell O., R.T., (1993): Forecasting and Time Series: An applied approach. 3rd Edition.
- Bollerslev, T.,(1986): Generalized Autoregressive Conditional Heteroskedasticity. Journal of Econometrics, 31:307-27
- Box, G.E.P., & Jenkins, G. M., (1976): Time Series Analysis, Forecasting and Control, Holden-Day, San Francisco.
- Box, G.E.P., and Jenkins, G.C., (1976): Time Series Analysis, Fore-casting and Control.
- Revised edition. Holden-Day.
- Box, G.E.P., and Jenkins, G.M., (1970):Time Series Analysis, Forecasting, and Control.
- Holden-Day.
- Box,G.E.P., Jenkins,G.M., and Reinsel, G.C.,(1994): Time Series Analysis. Forecasting and Control,3rd Ed. Prentice Hall
- Brockwell, P. J., and Davis, R. A., (2002):Introduction to Time series and forecasting, 2nd Ed. Springer-Verlag New York, Inc., 170-172
- Canova, F., and Hansen, B.E., (1995): “Are Seasonal Patterns Constant over Time? A Test for Seasonal Stability”, Journal of Business and Economic Statistics 13(3):236-250.
- Chatfield, C., (2004): The Analysis of Time Series: An Introduction. Chapman and Hall/CRC Conejo A. J., Miguel A., Plazas, Rosa E., and Ana B. M., (2005): “Day-ahead electricity
- price forecasting using the wavelet transform and ARIMA models,” IEEE Trans.
- Power Syst., vol. 20, no. 2, pp. 1034–1040.
