Elsevier

Pattern Recognition Letters

Volume 109, 15 July 2018, Pages 72-80
Pattern Recognition Letters

Distributed electricity load forecasting model mining based on hybrid gene expression programming and cloud computing

https://doi.org/10.1016/j.patrec.2017.10.004Get rights and content

Highlights

  • An load forecasting model based on artificial fish swarm and GEP is proposed.

  • Distributed load forecast model mining relies on hybrid GEP and cloud computing.

  • A global forecast model generation is achieved by error minimization crossover.

Abstract

Load forecasting is an important part of power grid management. Accurate and timely load forecasting is of great significance to formulate economical and reasonable power allocation plan, improve safety and economy of power grid operation and improve power quality. In this paper, in order to find electricity load forecasting model, we propose an electricity load forecasting function mining algorithm based on artificial fish swarm and gene expression programming (ELFFM-AFSGEP). On the basis, distributed load forecast model mining based on hybrid gene expression programming and cloud computing (DLFMM-HGEPCloud) is proposed to solve the problem of massive electricity load forecasting. In order to better solve global electricity load forecasting model, error minimization crossover is introduced into DLFMM-HGEPCloud. The performance of the proposed algorithm in this paper is evaluated with a real-world dataset, and compared with GEP and some published algorithms by using the same dataset. Experimental results show that our proposed algorithm has an advantage in average time-consumption, average number of convergence, forecasted accuracy and excellent parallel performance in speedup and scaleup.

Introduction

Electricity load forecasting plays an important role for power plant operators, power suppliers and traders, and energy managers in decision making [8]. From the point of view of the forecasting, the power load forecasting includes the prediction of future power demand and the prediction of future electricity consumption and the forecasting of the load curve. The main task is to predict the future distribution of power load and spatial distribution, power system planning and operation to provide a reliable basis for decision-making. The load forecasting is highly related to power system operations such as dispatch scheduling, preventive maintenance plan for the generators, and the reliability evaluation of the power systems. In addition, accurate power load data is crucial to power price forecasting in the electricity market. Accurate load forecasting is conducive to improve the safe operation of the power grid, the economy and power quality. Load forecasting is an important part of power grid management. Accurate and timely load forecasting is of great significance to formulate economical and reasonable power allocation plan, improve safety and economy of power grid operation and improve power quality.

However, with the rapid development of informationization and intelligent degree of power grid and the increasing demand of power demand factors, the increasing size of power load data, the first-tier cities in the peak period, faced with millions of records of the size of the power data collection, One year of data storage scale will grow from the current GB level to the terabyte level, and even the PB level [23], [36], [37], [38], [39]. Power load forecasting will face challenges and difficulties. First, the increasing mass of load data makes it possible for existing centralized load forecasting algorithms to become more difficult in terms of timely analysis and processing. Secondly, the existing power load forecasting algorithms mainly include statistical and artificial intelligence methods [1], [2], [7], [13], [25], [32], [34]. The load forecasting algorithm based on the statistical method is better in predicting the linear sequence of the power load, but the shortcomings of the inflexibility of the structure lead to the weak ability of the algorithm to predict the nonlinear load sequence, and subjective of the statistical method is too strong. The load forecasting algorithm based on artificial intelligence can deal with the nonlinear relationship between load and related factors, but this kind of algorithm is easy to lead to over-fitting and fall into local optimum, and the training time is longer and the efficiency is lower. In this paper, we mainly introduce the gene expression programming (GEP)[9], [10] to automatically mine the function form of load forecasting model, and ultimately achieve the purpose of accurate and timely prediction. The major contributions of our work are listed as follows:

  • (1)

    The power load forecasting function model mining algorithm based on a single GEP, is easy to produce premature phenomena in the population. In order to solve this problem, this paper combines the artificial fish swarm algorithm with strong local and global optimization ability, and proposes an electricity load forecasting function mining algorithm based on artificial fish swarm and gene expression programming (ELFFM-AFSGEP).

  • (2)

    With the rapid development of informationization of distribution network, the power load data show the characteristics of distribution, mass and high dimension. The centralized mining model of power load forecasting function will undoubtedly increase the safety, privacy and data storage problem, but also reduce the efficiency of power load forecasting. In order to solve prediction of massive power load data, based on the ELFFM-AFSGEP algorithm, this paper puts forward distributed load forecast model mining based on hybrid gene expression programming and cloud computing (DLFMM-HGEPCloud).

  • (3)

    Meanwhile, In order to solve the global electricity load forecasting model in DLFMM-HGEPCloud algorithm, it is not possible to solve the effective merging of nonlinear function model based on statistical method. Therefore, this paper proposes a global model generation algorithm for electricity load forecasting based on error minimization crossover (GMGELF-EMC).

The remainder of this paper is organized as follows. In Section 2, we briefly describe the related works. In Section 3, we introduce electricity load forecasting function mining algorithm based on artificial fish swarm and gene expression programming. In Section 4, we analyze distributed electricity load forecasting model mining algorithm gene expression programming and cloud computing. Experimental results are provided in Section 5 and we conclude this paper in Section 6.

Section snippets

Load forecasting

In the past decades, many methods have been proposed to improve load forecasting accuracy by many researchers. Taylor used exponentially weighted method to forecast short-term load [34]. Infield et al. proposed optimal smoothing for trend removal in short term electricity demand forecasting by using Kalman filters [15]. Huang et al. used ARMA model identification to forecast short-term load including non-Gaussian process considerations [13]. Experimental results show the effectiveness of the

Local optimization for GEP population based on artificial fish swarm algorithm

The existing load forecasting function model mining algorithm based on GEP is easy to produce individual precocious phenomenon in the population. In order to solve this problem, this paper proposes a local optimization for GEP population based on artificial fish swarm algorithm (LOGEP-AFSA).

Let S be the GEP population space, which contains N individuals. Each individual is regarded as an artificial fish. For any Xi, i ∈ [1, N], f(Xi) is the fitness value of the ith individual in population

Idea of algorithm

With the continuous development of distribution information, the electricity load data show the characteristics of distribution, mass and high dimension. The centralized electricity load forecasting function model mining algorithms require these distributed load data to be transmitted to the back-end servers, then analyzed uniformly. A large number of data transmission will undoubtedly increase the security, privacy and other issues, but also reduce the efficiency of electricity load

Experimental environment and data resource

To test the performance and effectiveness of the proposed algorithm in this paper, a cloud computing platform based on Hadoop 0.20.2 is built in the Lab. The parameters of computing platform are shown in Table 1. Fig. 2 shows experimental platform architecture based on Hadoop.

To evaluate the performance of the proposed algorithm in this paper, the dataset are provided by the EUNITE network during the daily peak load competition. The dataset is downloaded from http://neuron.tuke.sk/competition/.

Conclusions

In this paper, first, a new approach which called ELFFM-AFSGEP can be used to solve electricity load forecasting. The approach introduces artificial fish swarm into gene expression programming to improve the global optimization ability of gene expression programming. Secondly, on the basis of ELFFM-AFSGEP, for massive electricity load data, distributed electricity load forecasting based on hybrid gene expression programming and cloud computing (DELF-HGEPCloud) is proposed. Meanwhile, in

Acknowledgments

The authors would like to thank the anonymous reviewers for their comments and constructive suggestions that have improved the paper. The subject is sponsored by the National Natural Science Foundation of P. R. China (No. 51507084) and NUPTSF (No. NY214203).

References (44)

  • S. Li et al.

    Short-term load forecasting by wavelet transform and evolutionary extreme learning machine

    Electr. Power Syst. Res.

    (2015)
  • X. Lu et al.

    Distributed hs-artmap and its forecasting model for electricity load

    Appl. Soft Comput.

    (2015)
  • J.D. Melo et al.

    Determining spatial resolution in spatial load forecasting using a grid-based model

    Electr. Power Syst. Res.

    (2014)
  • Y.Z. Peng et al.

    An improved gene expression programming approach for symbolic regression problems

    Neurocomputing

    (2014)
  • A. Selakov et al.

    Hybrid PSO–SVM method for short-term load forecasting during periods with significant temperature variations in city of burbank

    Appl. Soft Comput.

    (2014)
  • K. Siah Yap et al.

    Daily maximum load forecasting of consecutive national holidays using oselm-based multi-agents system with weighted average strategy

    Neurocomputing

    (2012)
  • X. Zhu et al.

    Block-row sparse multiview multilabel learning for image classification

    IEEE Trans. Cybern.

    (2016)
  • X. Zhu et al.

    Robust joint graph sparse coding for unsupervised spectral feature selection

    IEEE Trans. Neural Netw. Learn. Syst.

    (2017)
  • S. Deng et al.

    Distributed mining for content filtering function based on simulated annealing and gene expression programming in active distribution network

    IEEE Access

    (2017)
  • E.E. Elattar et al.

    Electric load forecasting based on locally weighted support vector regression

    IEEE Trans. Syst. Man Cybern. Part C

    (2010)
  • G.F. Fan et al.

    Support vector regression model based on empirical mode decomposition and auto regression for electric load forecasting

    Energies

    (2013)
  • E.A. Feinberg et al.

    Load Forecasting

    (2005)
  • Cited by (0)

    View full text