An evolutionary-based predictive soft computing model for the prediction of electricity consumption using multi expression programming

doi:10.1016/j.jclepro.2020.125287

Journal of Cleaner Production

Volume 283, 10 February 2021, 125287

https://doi.org/10.1016/j.jclepro.2020.125287 Get rights and content

Abstract

Proper estimation of electricity consumption is one of the influential factors for sustainability and cleaner production in both developed and developing countries. Many studies have been conducted to present accurate prediction models for forecasting electricity demand. However, researchers are still working to develop models with higher accuracy. This study applies a newer branch of Genetic Programming (GP) as a soft computing technique, known as Multi Expression Programming (MEP) to predict the electricity consumption of China for the first time based on the data collected from 1991 to 2019. Specifically, a robust mathematical model was developed using MEP for this purpose. Different predictive techniques known as Gene Expression Programming (GEP) and Adaptive Neuro Fuzzy Inference System (ANFIS) were used to compare the accuracy of the model. Based on the results, the proposed MEP model is more powerful and accurate than both GEP and ANFIS. In addition, a sensitivity analysis was conducted to present the impact of each factor on the electricity consumption of China. It was shown that among the four independent factors (Population, Gross Domestic Product (GDP), Import, and Export), Population has the highest impact, followed by Export, Import and GDP, respectively.

Introduction

Electricity demand forecasting is one of the critical issues in sustainability and cleaner production (Wang et al., 2019).de Azevedo et al. (2018) stated that sustainability comprises three elements which are economy, environment and society. In the recent decades, the demand for electricity and energy resources has increased tremendously and therefore, its linkage with sustainability has received much attention.

Long-term planning of energy supply-demand must satisfy the requirements for sustainable development of a country. Accurate forecasts can help decision makers to know the volume and trend of future energy consumption to better schedule and plan the operations of the supply system. Since most countries perform financial budget planning and allocation on a yearly basis, an annual forecast of energy demand is important. A precise estimation of energy consumption is necessary to manage social and economic growth of developing countries such as China. Previous reports showed that over the past two decades, the demand (consumption) of energy has gone up almost 200% in China (Petroleum, 2009). In addition, over these years, China has faced energy shortages many times (Wang, Q. et al., 2018). In order to avoid such a problem which strongly affects economic growth, developing an accurate predictive model for energy consumption is very significant for China’s policy makers.

The related literature on energy consumption prediction reveals that many scholars have applied different approaches to estimate energy consumption. Overall, these methods are classified into three categories: 1) Time series methods; 2) Regression-based methods; 3) Soft computing methods.

Over the recent years, soft computing techniques have been widely used for simulating and estimating energy consumption. Generally, soft computing techniques can be categorized into three groups: 1) Optimization algorithms such as Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO) and Genetic Algorithm (GA); 2) Neural-based Predictive Soft Computing (NPSC) methods such as Adaptive Neuro Fuzzy Inference System (ANFIS), Support Vector Machine (SVM) and Artificial Neural Network (ANN) (Zhou et al., 2020); 3) Evolutionary-based Predictive Soft Computing (EPSC) techniques such as Genetic Programming (GP) and Gene Expression Programming (GEP).

Many researchers have applied NPSC methods in their studies and indicated that these methods are powerful in their estimation, but their serious drawback is that they act as a black box system (Luo et al., 2019; Pattanaik et al., 2019). This denotes that they are not able to provide a clear mathematical equation connecting the output (energy consumption) and the input variables (such as Gross Domestic Product (GDP), Population, Import, and Export). Therefore, there are still some basic questions to be considered:

i)
How useful is an intangible neural-based mechanism which only predicts energy demand without any mathematical model?
ii)
How can the developed NPSC mechanism ease the forecasting procedure for decision makers if this type of technique strongly needs special knowledge?

In order to answer the aforementioned questions, EPSC techniques such as GP, GEP and Optimized GEP have been utilized by different researchers (Kaboli et al., 2017; Lu et al., 2017; Samadianfard et al., 2018). In fact, using these methods, an explicit mathematical equation that links the output and inputs can be generated. The relevant literature shows that there is a newer version of GP known as Multi Expression Programming (MEP) which is more accurate than other GP-based techniques (Alavi et al., 2010). Despite the many advantages of this technique over the other variants of GP, MEP is still very new in energy demand prediction.

The most significant aim of this paper is to generate a mathematical equation for forecasting the electricity consumption of China using MEP based on the data sets collected from 1991 to 2019. To date, the application of MEP has not been considered in the context of energy demand estimation. This is the first study that utilizes MEP which is not only applicable for China, but it can be used internationally for any country. It is a generic and universal approach. This technique can help decision makers of any country to estimate electricity consumption accurately. In order to show the accuracy of MEP, it is compared with other methods including GEP, ANFIS, and regression.

The rest of this paper is structured as follows: Section 2 reviews the relevant literature. Then, explanations of GP, GEP, MEP and the proposed steps are provided in Section 3. The developed mathematical model is presented in Section 4. Section 5 discusses the assessment results of the model. In Section 6, a sensitivity analysis is conducted. Managerial implications are described in Section 7. Finally, conclusions are given in Section 8.

Section snippets

Literature review

As mentioned earlier, there are three main categories of methods for energy demand forecasting. In this section, some of the studies are presented.

Methodology

This section comprises four sub-sections including an overview of GP, GEP, MEP and the proposed steps.

Formulation of electricity consumption of China

MEP was utilized to develop a mathematical formulation for the electricity consumption of China with respect to the data collected from 1991 to 2019. As mentioned before, the process was started by defining the most important inputs. In this study, Population (P), Gross Domestic Product (GDP), Import (Im), and Export (Ex) of China were determined as the independent variables, and Electricity Consumption (EC) was the dependent variable as illustrated in Fig. 4. The recorded data of China in the

Performance validation of the generated MEP mathematical model

This section includes four different sub-sections. Firstly, the MEP model was evaluated by applying different statistical tests. Secondly, a comparison between the accuracy of the MEP model and GEP (as one of the well-known GP variants) was done. Thirdly, a comparison between the MEP model (as an EPSC technique) and ANFIS (as one of the powerful NPSC methods) was carried out. In addition, the accuracy of the MEP model was compared with regression (as a traditional prediction technique).

Sensitivity analysis

Sensitivity analysis of the inputs or independent variables that influence the electricity consumption of China was performed to demonstrate the contributions of the utilized predictive factors in prediction. To this end, after finding the optimized parameters of MEP, the MEP structure was run thirty times to obtain the frequency of each independent variable in contributing to the fitness of the MEP programs over the thirty runs (Alavi et al., 2010). The sensitivity analysis showed that

Managerial implications

One of the important features of any prediction model is its applicability for decision makers or managers in real world. As mentioned before, ANFIS and other NPSC techniques are considered as a black box which means they can only forecast electricity consumption without providing any mathematical model. As there is no mathematical equation that links the dependent variable and independent variables, managers need to understand the neural-based structure of ANFIS. However, the structure of

Conclusions

In the past, many computer aided techniques such as ANN, ANFIS, etc., have been developed to predict energy consumption. These techniques only present a list of values for an output using a neural-based structure which consists of intangible information. Arguably, these techniques are less useful for policy makers and may make them more confused (due to the need for deep knowledge to understand a neural-based structure and its elements). In real world, a more tangible way to predict and

CRediT authorship contribution statement

Alireza Fallahpour: Writing - original draft, Writing - review & editing. Kuan Yew Wong: Writing - original draft, Writing - review & editing. Srithar Rajoo: Writing - original draft, Writing - review & editing. Guangdong Tian: Writing - original draft, Writing - review & editing.

Declaration of competing interest

The authors declare that they have no conflict of interest.

Acknowledgement

The authors are grateful to Universiti Teknologi Malaysia (UTM) (Vote Number: 20H69). In addition, this work is supported in part by Natural Science Foundation of China Under Grant No. 51775238 and 52075303, Science and technology development project of Jilin Province under Grant Nos 20180101060JC and 20180101058JC.

References (46)

E. Assareh et al.
Application of PSO (particle swarm optimization) and GA (genetic algorithm) techniques on demand estimation of oil in Iran
Energy
(2010)
A. Azadeh et al.
Integration of artificial neural networks and genetic algorithm to predict electrical energy consumption
Appl. Math. Comput.
(2007)
V. Bianco et al.
Electricity consumption forecasting in Italy using linear regression models
Energy
(2009)
J. Catalão et al.
Short-term electricity prices forecasting in a competitive market: a neural network approach
Elec. Power Syst. Res.
(2007)
R. Chen et al.
The long-term forecast of energy demand and uncertainty evaluation with limited data for energy-imported cities in China: a case study in Hunan
Energy Procedia
(2019)
W. Chen et al.
Neuro-genetic, neuro-imperialism and genetic programing models in predicting ultimate bearing capacity of pile
Eng. Comput.
(2019)
A.R. de Azevedo et al.
Recycling paper industry effluent sludge for use in mortars: a sustainability perspective
J. Clean. Prod.
(2018)
E.M. de Oliveira et al.
Forecasting mid-long term electric energy consumption through bagging ARIMA and exponential smoothing methods
Energy
(2018)
K. Ermis et al.
Artificial neural network analysis of world green energy use
Energy Pol.
(2007)
N. Fumo et al.
Regression analysis for prediction of residential energy consumption
Renew. Sustain. Energy Rev.
(2015)

A. Golbraikh et al.

Beware ofq²!

J. Mol. Graph. Model.

(2002)

S.H.A. Kaboli et al.

Long-term electric energy consumption forecasting via artificial cooperative search algorithm

Energy

(2016)

S.H.A. Kaboli et al.

Long-term electrical energy consumption formulating and forecasting via optimized gene expression programming

Energy

(2017)

M.-R. Kazemzadeh et al.

A hybrid data mining driven algorithm for long term electric peak load and energy demand forecasting

Energy

(2020)

M. Khashei et al.

A novel hybridization of artificial neural networks and ARIMA models for time series forecasting

Appl. Soft Comput.

(2011)

E.S. Mostafavi et al.

A novel machine learning approach for estimation of electricity demand: an empirical evidence from Thailand

Energy Convers. Manag.

(2013)

S.M. Mousavi et al.

Gene expression programming as a basis for new generation of electricity demand prediction models

Comput. Ind. Eng.

(2014)

K. Muralitharan et al.

Neural network based optimization approach for energy demand prediction in smart grid

Neurocomputing

(2018)

S.S. Pappas et al.

Electricity demand loads modeling using AutoRegressive Moving Average (ARMA) models

Energy

(2008)

S. Samadianfard et al.

Wavelet neural networks and gene expression programming models to predict short-term soil temperature at different depths

Soil Tillage Res.

(2018)

D.H. Vu et al.

A variance inflation factor and backward elimination based robust regression model for forecasting monthly electricity demand using climatic variables

Appl. Energy

(2015)

Q. Wang et al.

Forecasting energy demand in China and India: using single-linear, hybrid-linear, and non-linear time series forecast techniques

Energy

(2018)

X. Wang et al.

Estimates of energy consumption in China using a self-adaptive multi-verse optimizer-based support vector machine with rolling cross-validation

Energy

(2018)

Cited by (24)

Forecasting the strength characteristics of concrete incorporating waste foundry sand using advance machine algorithms including deep learning
2023, Case Studies in Construction Materials
The incorporation of waste foundry sand (WFS) into concrete has been recognized as a sustainable approach to improve the strength properties of waste foundry sand concrete (WFSC). However, machine learning (ML) techniques are still necessary to forecast the characteristics of WFSC and evaluate the dominant input features for the suitable mix design. For this purpose, the present work selected five ML-based techniques based on gene expression programming (GEP), deep neural network (DNN), and optimizable Gaussian process regressor (OGPR) to predict the mechanical characteristics of WFSC. To build up the predictive models, a database containing 397 values of compressive strength (CS) and 169 values of flexural strength (FS) is collected from published literature. The models' performance was evaluated via various statistical metrics and additionally, external validation criteria were employed to validate the developed models. Furthermore, the Shapley additive explanation (SHAP) was carried out to interpret the model's prediction. The DNN2 model exhibited superior performance, with R-values of 0.996 (training), 0.999 (testing), and 0.997 (validation) for the compressive strength estimation. In contrast, the GEP2 model showed poor accuracy in estimating the CS compared to other developed models, with R-values of 0.851, 0.901, and 0.844 for the training, testing, and validation sets, respectively. Similarly, for the flexural strength estimation, the DNN2 model provided R-values of 0.999 for training, 0.996 for testing, and 0.999 for validation sets, indicating its robust performance. The SHAP analysis revealed that the age, water-cement ratio, and coarse aggregate-to-cement ratio have the prime influence in determining flexural and compressive strength, respectively. The comparison of the models provided that the DNN2 model accurately estimated the output with high accuracy and lower error values and might be utilized in practical fields to reduce labor and cost by optimizing the mix combinations. Finally, for future studies, it is recommended to utilize ensemble and hybrid algorithms, as well as post-hoc explanatory techniques, to forecast the characteristics of WFSC accurately.
Energy consumption predictions by genetic programming methods for PCM integrated building in the tropical savanna climate zone
2023, Journal of Building Engineering
The development of energy-efficient buildings by considering early-stage design parameters can help reduce buildings' energy consumption. Machine learning tools are getting popular for forecasting the energy demand of buildings, which play a vital role in improving building energy efficiency. In this research, multi-expression and genetic expression programming were utilized to anticipate the energy consumption of PCM-integrated buildings by taking early-stage design parameters into consideration. The prediction models were developed using the data generated by energy simulations for the PCM-integrated building in eight cities within a tropical savanna climate. The statical parameters were used to evaluate and externally validate the proposed prediction model. The statistical evaluation reveals that the genetic expression programming-based predictive model gave more accurate energy consumption predictions for PCM-integrated buildings than multi-expression programming. The performance indices of the statistically analyzed gene expression programming-based prediction model (GEP7) showed excellent values: correlation coefficient (R) = 0.961, performance index (ρ) = 0.169, and Nash-Sutcliffe efficiency (NSE) = 0.108. Thereafter, the sensitivity and parametric analyses were performed. It was unearthed that the roof solar absorptance, window visible transmittance, wall solar absorptance, and the melting temperature of PCM were the influential early-stage design parameters for PCM-integrated buildings. In conclusion, the gene-expression programming-based predictive model can be utilized to predict the influence of early-stage design parameters on the energy consumption of PCM-integrated buildings.
Crack width prediction of self-healing engineered cementitious composite using multi-expression programming
2023, Journal of Materials Research and Technology
Concrete structures are prone to cracking due to environmental damage and/or tensile strain. Self-healing concrete has garnered scientific interest in recent years as it addresses the cracking problem and extends the service life of concrete structures. Engineered cementitious composite (ECC), a high-performance fiber-reinforced cement composite, is known to have some self-healing characteristics. However, the self-healing capability of ECC is characterized by the type and content of input parameters and is difficult to predict. This investigation utilised a multi-expression programming algorithm for the prediction of self-healing characteristics of ECC containing various admixtures. The database, containing 619 points, was extracted from literature containing crack width before self-healing (CB), limestone powder (LP), silica fume (SF), and fly ash (FA) as input parameters and crack width after self-healing as an output parameter. The performance of the developed model was evaluated based on various statistical indices and parametric analysis. The results showed that the developed model is robust and efficient in predicting the self-healing behavior of ECC containing various admixtures. Parametric and sensitivity analysis revealed that among the admixtures used, fly ash had the least impact and limestone powder had the highest impact on the self-healing capacity of ECC.
Low-dimensional representation of monthly electricity demand profiles
2023, Engineering Applications of Artificial Intelligence
Citation Excerpt :
Specifically, in the retailer arena, data-based decision making has gained increasing interest in methods for long-term load forecasting, energy procurement strategies, pricing schemes, and risk management in the retail market (Liu et al., 2020; Yang et al., 2018). Some authors have done some related works on the long-term evolution of electric demand (Fallahpour et al., 2021; Sánchez-Durán et al., 2019b) and the solar production (Sánchez-Durán et al., 2019a), and also in featuring electricity demand using spectral analysis (Luque et al., 2020). The characterization of load curves has highly exploited the developments in the data mining of time series, with techniques generally categorized into representation and indexing, similarity measure, segmentation, visualization, and mining (Fu, 2011).
This paper addresses the problem of reducing the number of values required to characterize an electricity demand profile, which is usually known as its dimensionality. This reduction may have a significant impact on the computational efforts and storage capacities required to analyze and process high volumes of electricity load curves. Also, the reduction to 2 or even 1 component enables its graphic representation. Specifically, this work is mainly focused on profiles defined by their monthly demand values, and where the clients are aggregated by locations and/or economic activities. This approach is of great interest for marketing analysis and decision-making of electricity retailers. In this sense, the use of dimensionality reduction techniques based on knowledge (calendar and temperature) along with the application of data-driven procedures (Principal Component Analysis and autoencoders), are explored in the paper. The results of this research show that autoencoders clearly outperform the other techniques, yielding errors in the reduction process between 15% to 40% lower and preserving distances between profiles in the low-dimensional spaces, with a correlation of 0.93 with the distances in high dimensional space. Additionally, the bidimensional graphical representation of a profile can easily be interpreted in a polar way, where the angle denotes the shape of the profile, and the radius reveals its scale. To reach these results, a very large dataset has been employed, with about half a million aggregated profiles corresponding to the electricity consumption during 3 years of more than 27 million clients in Spain.
A novel multivariate grey model for forecasting periodic oscillation time series
2023, Expert Systems with Applications
To solve the problem that the grey multivariate prediction model cannot well simulate systems with periodic oscillations, a novel multivariate grey model called the GM(1,N|sin) power model is proposed. The power exponential term and dynamic sinusoidal function are developed to represent the nonlinear relationship and periodic oscillations of the independent and dependent variables in the proposed model, respectively. First, the discrete time response formula for generating the simulated values at each time point is given. Second, the nonlinear programming model based on the particle swarm optimization algorithm is established to solve the power exponential and periodic coefficients. In addition, to enhance the generalization ability of the parameters, an improved nonlinear programming model considering the accuracy of the validation set is constructed. Finally, in the case studies, the quarterly electricity consumption of Jiangsu Province and PM_2.5 concentrations in Nanjing are adopted to test the effectiveness of this model, and the results are obtained respectively through the GM(1,N|sin) power model and alternative models. The results indicate that the accuracy of GM(1,N|sin) power model in the time series with periodic oscillations outperforms the other six models in this paper.
Simulation study on carbon emission of China's electricity supply and demand under the dual-carbon target
2022, Journal of Cleaner Production
As the world's largest developing country, China has set the target of “reaching carbon peak by 2030 and carbon neutrality by 2060”. China's power sector faces many challenges in reducing carbon emissions. This article forecasts electricity consumption from 2021 to 2060 by selecting five influential elements on power demand. Furthermore, the system dynamics model simulates the carbon emission system on the supply and demand side of the power industry. This paper innovatively incorporates the technology factor of carbon capture, utilization, and storage (CCUS) into the supply-side model. In this study, the process of carbon neutrality is separated into three stages, including carbon peak stage (2020–2030), deep low-carbon stage (2030–2050) and zero-carbon stage (2050–2060). The factors of energy storage technology and CCUS technology make significant contributions in the latter two stages. Based on the different development levels of the variables, three scenarios are set as business-as-usual scenario (BAU), low carbon scenario(LC) and ultra-low carbon scenario (ULC). Then scenario analysis is employed to simulate the carbon emission in the future. The scheme under the ULC scenario is the optimal solution to achieve carbon neutrality. Improving flexible transition of thermal power units and promoting the growth of renewable energy are China's primary strategies of achieving carbon peak and carbon neutrality.

View all citing articles on Scopus

View full text

An evolutionary-based predictive soft computing model for the prediction of electricity consumption using multi expression programming

Abstract

Introduction

Section snippets

Literature review

Methodology

Formulation of electricity consumption of China

Performance validation of the generated MEP mathematical model

Sensitivity analysis

Managerial implications

Conclusions

CRediT authorship contribution statement

Declaration of competing interest

Acknowledgement

Energy

Appl. Math. Comput.

Energy

Elec. Power Syst. Res.

Energy Procedia

Eng. Comput.

J. Clean. Prod.

Energy

Energy Pol.

Renew. Sustain. Energy Rev.

J. Mol. Graph. Model.

Energy

Energy

Energy

Appl. Soft Comput.

Energy Convers. Manag.

Comput. Ind. Eng.

Neurocomputing

Energy

Soil Tillage Res.

Appl. Energy

Energy

Energy