Improved estimation of electricity demand function by using of artificial neural network, principal component analysis and data envelopment analysis

https://doi.org/10.1016/j.cie.2012.09.017Get rights and content

Abstract

Due to various seasonal and monthly changes in electricity consumption and difficulties in modeling it with the conventional methods, a novel algorithm is proposed in this paper. This study presents an approach that uses Artificial Neural Network (ANN), Principal Component Analysis (PCA), Data Envelopment Analysis (DEA) and ANOVA methods to estimate and predict electricity demand for seasonal and monthly changes in electricity consumption. Pre-processing and post-processing techniques in the data mining field are used in the present study. We analyze the impact of the data pre-processing and post-processing on the ANN performance and a 680 ANN-MLP is constructed for this purpose. DEA is used to compare the constructed ANN models as well as ANN learning algorithm performance. The average, minimum, maximum and standard deviation of mean absolute percentage error (MAPE) of each constructed ANN are used as the DEA inputs. The DEA helps the user to use an appropriate ANN model as an acceptable forecasting tool. In the other words, various error calculation methods are used to find a robust ANN learning algorithm. Moreover, PCA is used as an input selection method, and a preferred time series model is chosen from the linear (ARIMA) and nonlinear models. After selecting the preferred ARIMA model, the Mcleod–Li test is applied to determine the nonlinearity condition. Once the nonlinearity condition is satisfied, the preferred nonlinear model is selected and compared with the preferred ARIMA model, and the best time series model is selected. Then, a new algorithm is developed for the time series estimation; in each case an ANN or conventional time series model is selected for the estimation and prediction. To show the applicability and superiority of the proposed ANN-PCA-DEA-ANOVA algorithm, the data regarding the Iranian electricity consumption from April 1992 to February 2004 are used. The results show that the proposed algorithm provides an accurate solution for the problem of estimating electricity consumption.

Highlights

► We develop an integrated method to estimate electricity demand for seasonal and monthly changes in electricity consumption. ► We study the impact of data preprocessing and postprocessing on artificial neural network performance. ► Data envelopment analysis is utilized to compare constructed artificial neural network models. ► A new algorithm is developed for time series estimation.

Section snippets

Significance

The significance of the proposed algorithm is fourfold. First, it is flexible and identifies the preferred estimation model based on the results of MAPE (Minimum Absolute Percentage Error), ANOVA and DEA, whereas previous studies consider the best fitted fuzzy regression model based on MAPE or other relative error results. Second, the proposed model may identify a linear (ARIMA) or nonlinear time series model. ARIMA is the best model to be used for the prediction of electricity consumption

Literature review

Artificial Neural Networks, Genetic Algorithms, Neuro Fuzzy and Fuzzy Inference Systems are often used in the energy sector as significant tools in Artificial Intelligence (AI) science. High flexibility, good estimation, forecasting capability and their ability to deal with noisy data are the main reasons that these methods are used in energy estimation and prediction. Some of the AI applications in the estimation of energy demand in various fields are discussed below.

Peng et al. (1992)

Explanation of estimation tools

Artificial Neural Networks (ANNs), conventional time series models and data pre-processing methods are described in the following sub-sections.

Proposed algorithm

We now develop an algorithm to model the time series process, which is shown in Fig. 3. Moreover, the conceptual framework of the proposed algorithm is depicted in Fig. 4. Fig. 3, Fig. 4 assist readers to follow the proposed algorithm. Two ANN models (ANNW and ANNWO) are considered to determine the impact of pre-processing on the ANN models for estimation. Another conventional time series estimation method is also considered to study the efficiency of ANN compared with conventional models.

This

Case study

The proposed algorithm is applied to 130 data, which are the monthly consumption values from April 1992 to February 2004 in Iran. The data are derived from the Energy Balances of Islamic Republic of Iran book (2005 version) that was prepared by the Ministry of Energy, Energy Planning Department. The raw data is shown in Table 2. The flowchart of the algorithm in the pre-processing case and for ANN is depicted in Fig. 6. The flowchart shows only the flow of the algorithm at the end of DEA and

Comparison with other intelligent methods

The results of four intelligent methods have been used in the present study and are compared with the proposed algorithm in terms of MAPE value.

First: Genetic Algorithm (GA) is similar to the natural evolution process where a population of specific species adapts to the natural environment under consideration. A population of designs is also created and then allowed to evolve in order to adapt to the design environment under consideration. Here, several GA studies are cited to show the

Conclusion

In this paper, an algorithm was developed to improve the electricity consumption estimation. The intelligent ANN–PCA–DEA algorithm was developed by different data pre-processing methods and its efficiency was examined for Iranian electricity consumption. DEA and Granger–Newbold test were used to show the efficiency of the ANN. The algorithm for calculating the ANN performance is based on its closed and open simulation abilities. DEA was used to find a suitable ANN learning algorithm, which is

Acknowledgments

This study was supported by a grant from University of Tehran (Grant No. 8106013/1/09), and also another grant from Iran National Science Foundation (INSF), Grant No. 91000659.

References (53)

  • W.C. Chiang et al.

    A neural network approach to mutual fund net asset value forecasting

    Omega

    (1996)
  • T.S. Dillon et al.

    Short term load forecasting using an adaptive neural network

    International Journal of Electrical Power & Energy Systems

    (1991)
  • R. Gareta et al.

    Forecasting of electricity prices with neural networks

    Energy Conversion and Management

    (2006)
  • M. Ghiassi et al.

    A dynamic artificial neural network model for forecasting time series events

    International Journal of Forecasting

    (2005)
  • C.W.J. Granger et al.

    Spurious regressions in econometrics

    Journal of Econometrics

    (1974)
  • C.C. Hsu et al.

    Regional load forecasting in Taiwan—Applications of artificial neural networks

    Energy Conversion and Management

    (2003)
  • D.C. Indro et al.

    Predicting mutual fund performance using artificial neural networks

    Omega

    (1999)
  • S.A. Kalogirou

    Applications of artificial neural networks for energy systems

    Applied Energy

    (2000)
  • S.A. Kalogirou et al.

    Artificial neural networks for the prediction of the energy consumption of a passive solar building

    Energy

    (2000)
  • D.S.K. Karunasinghe et al.

    Chaotic time series prediction with a global model artificial neural network

    Journal of Hydrology

    (2006)
  • K. Kavaklioglu

    Modeling and prediction of Turkey’s electricity consumption using support vector regression

    Applied Energy

    (2011)
  • T.Y. Kim et al.

    Artificial neural networks for non-stationary time series

    Neurocomputing

    (2004)
  • N. Kohzadi et al.

    A comparison of artificial neural network and time series models for forecasting commodity prices

    Neurocomputing

    (1996)
  • K. Metaxiotis et al.

    Artificial intelligence in short term electric load forecasting: a state-of-the-art survey for the researcher

    Energy Conversion and Management

    (2003)
  • P.C. Nayak et al.

    A neuro-fuzzy computing technique for modeling hydrological time series

    Journal of Hydrology

    (2004)
  • A.L.I. Oliveira et al.

    Detecting novelties in time series through neural networks forecasting with robust confidence intervals

    Neurocomputing

    (2006)
  • Cited by (65)

    • Supporting better practice benchmarking: A DEA-ANN approach to bank branch performance assessment

      2020, Expert Systems with Applications
      Citation Excerpt :

      BP networks are a class of feed-forward NNs (i.e., the information flows from the input to the output layer) and the network’s forecasts are compared with the known target or desired output and the weights are adjusted based on the forecasting error to minimize an error function (Shokrollahpour, Hosseinzadeh Lotfi, & Zandieh, 2016). For different algorithms, see Gallant and Stephen (1993) and Kheirkhah, Azadeh, Saberi, Azaron, and Shakouri (2013). In this paper, we focus on proposing a practical methodology to support benchmarking analysis in bank branch networks by employing DEA and ANN in tandem.

    View all citing articles on Scopus
    View full text