entropy

Review
A Systematic Review of Statistical and Machine
Learning Methods for Electrical Power Forecasting
with Reported MAPE Score
Eliana Vivas 1, * , Héctor Allende-Cid 1 and Rodrigo Salas 2
1 Escuela de Ingeniería Informática, Pontificia Universidad Católica de Valparaíso, Brasil 2950, Valparaíso,
Chile; [email protected]
2 Escuela de Ingeniería C. Biomédica, Universidad de Valparaíso, Chacabuco 2092-2220, Valparaíso, Chile;
[email protected]
* Correspondence: [email protected]; Tel.: +56-9-870-25259




Received: 29 October 2020; Accepted: 10 December 2020; Published: 15 December 2020


Abstract: Electric power forecasting plays a substantial role in the administration and balance of
current power systems. For this reason, accurate predictions of service demands are needed to
develop better programming for the generation and distribution of power and to reduce the risk
of vulnerabilities in the integration of an electric power system. For the purposes of the current
study, a systematic literature review was applied to identify the type of model that has the highest
propensity to show precision in the context of electric power forecasting. The state-of-the-art model
in accurate electric power forecasting was determined from the results reported in 257 accuracy tests
from five geographic regions. Two classes of forecasting models were compared: classical statistical
or mathematical (MSC) and machine learning (ML) models. Furthermore, the use of hybrid models
that have made significant contributions to electric power forecasting is identified, and a case of study
is applied to demonstrate its good performance when compared with traditional models. Among our
main findings, we conclude that forecasting errors are minimized by reducing the time horizon,
that ML models that consider various sources of exogenous variability tend to have better forecast
accuracy, and finally, that the accuracy of the forecasting models has significantly increased over the
last five years.

Keywords: electric power; forecasting accuracy; machine learning

1. Introduction
Electric power forecasting plays a substantial role in the administration and balance of
current power systems. The load forecasts help to identify strategies to optimize the operating
mechanisms in a determined period and thus ensure the demand even in situations adverse to the
system [1]. Accompanying the rapid advances in forecasting theory [2,3] and machine learning [4–6],
the technology in the energy forecasting research area has also developed rapidly [7]. Additionally,
the popular prediction methods for the generation and demand of energy can be divided into two
categories. The first category is statistical or mathematical methods, and the second category is modern
statistical-learning-based methods (also known as machine learning). In addition, hybrid methods can
be found that apply not only statistical tools but also other elements, such as mathematical optimization
or signal processing [8,9]. Additionally, other authors [10] consider hybrid approaches that focus
on a series of individual methods, such as noise reduction, seasonal adjustment and clustering,
to process the data in advance, whereas combined methods use weight coefficients. With respect to
the techniques implemented to forecast energy in recent years, in the international context, we can

Entropy 2020, 22, 1412; doi:10.3390/e22121412 www.mdpi.com/journal/entropy

Entropy 2020, 22, 1412 2 of 24

find a wide diversity; e.g., the application of kernel-based multitask learning methodologies [11],
energy load forecasting methodologies based on deep neural networks as in [12,13], methodologies
based on the classic time series approach as in [14–16], and mathematical representations as in [17–19].
Developing a model that achieves the highest forecasting precision in the context of electric power
has been the object of study in recent years. Additionally, the determination of the appropriate
input variables in load forecasting constitutes an important part of the forecasting procedure [20].
Due to the importance of the area, several review papers have appeared that present insights into
current applications and future challenges and opportunities [21,22]. However, existing review papers
examine the applications of a single model, e.g., an ANN [23], or cover only one energy domain,
e.g., solar radiation prediction [24,25], and do not perform comparisons among specific metrics, such as
MAPE, for multiple applications. Therefore, a systematic review to identify the type of model that has
the highest propensity to show precision in the forecast is the main objective of this paper.

Motivation and Scope of the Review

The number of papers published on the topic of electric power forecasting has been growing at an
exponential rate throughout the last decade, as Figure 1 shows. The order of magnitude of the increase
in the number of scientific publications on the subject revolves around 61.59%, between 2016–2020,
with respect to 2011–2015. Generally, the studies are site-specific, and the results strongly depend
on the nature of the model and the time horizon of the forecast, along with a large number of
other characteristics pertaining to the data and models. This is a major limitation, which makes a
generalization of the results difficult. A test of a given model over all different mentioned factors is
needed to measure the average effect of the model [25]. Consequently, the contribution of this paper is
to present the state-of-the-art of models in electric power systems and discuss their likely future trends,
considering:

• (I) The models that tend to provide precision in electric power forecasts according to the literature.
• (II) Exogenous sources that tend to lead to accurate forecasting of electrical energy according to
the literature.
• (III) Relationships between the times of forecasting and the accuracy of existing models.

Figure 1. Number of articles published per year. The line represents an exponential fit, highlighting
the yearly growth trend. The publications from 2020 are excluded since only partial data are available.

The rest of this paper is organized as follows. In Section 2, the methodology of the research is
presented. In Section 3, a description of the data set is presented. In Section 4, a performance analysis of
the forecasting models is presented, and finally, the overall discussion and conclusions are presented.
Entropy 2020, 22, 1412 3 of 24

2. Theoretical and Referential Framework

This chapter presents an analysis of the documents found in the literature during the last 15 years
on the subject of electric power forecasting.

2.1. Selection Criteria

The number of documents published on the topic of electric power forecasting has been growing
at an exponential rate throughout the last 15 years, as Figure 1 shows. We analyze in the review
the documents published for electric power forecasting contained in SCOPUS, Web of Sciences,
Science Direct and IEEE (Figure 2), according to the criteria shown in Figure 3 and following the steps
of the PRISMA (Preferred Reporting items for Systematic Review and Meta-Analyses) methodology.
A large number of papers published between January 2005 and March 2020 were analyzed.
The qualitative and quantitative synthesis of the analysis was collected from 164 documents selected
based on the criteria shown in Figure 3; the documents that only forecast electric power in buildings,
universities, homes, and rooms were excluded; likewise, if the time horizons are not mentioned in the
Abstract, the article was also skipped in our research. Similarly, if in an article, MAPE was not used as
a criterion for accuracy, it was not considered in our review. When considering the accuracy of the
results reported by the selected papers in terms of the MAPE Equation (1), we can compare samples of
different magnitudes, thus ensuring a common basis for intercomparison analyses.
It is important to highlight that under our filtering criteria a significant volume of valuable
references may have been excluded; in this sense, our searches may not be specific. If the readers are
interested exclusively in consulting documents related to forecasting under machine learning methods,
then they could consult [21], or if they are interested in specific documents on solar energy, they could
consult [25]; in the case of documents associated with the forecast under classical statistical techniques,
there are specific documents that can be consulted, such as [26].

Figure 2. PRISMA Flow Diagram.

Entropy 2020, 22, 1412 4 of 24

Figure 3. Search methodology for finding relevant literature.

2.2. Statistical Indicators of Accuracy in Electric Power Forecasting

The number of papers from the publications we studied that are eligible according to the criteria
was 164. We collected the mean absolute percentage error (MAPE), a statistical indicator of accuracy.
This index indicates an average of the absolute percentage errors (Equation (1)); the lower the MAPE, the higher
is the accuracy [27].
1 m tk − yk
mk k∑
MAPE = ∗ 100 (1)
=1
tk

where tk is the actual value of electric power, yk is the forecasting value produced by the model, and m
is the total number of observations. The final quality-controlled database from the 164 documents
contained 4883 entries (MAPE, type of MAPE, country, date, input variables, model, type of model,
latitude, longitude, and size of sample), and we saved 257 entries associated with a MAPE value
linked to the best model proposed in the cases of study with data from 33 countries. The locations
are represented on the world map in Figure 4. We can see that the studied publications cover all
continents. Occupying the first positions in the list of the countries with the most electrical energy
forecast documents, under the criteria used, are Australia, China, Iran, and Turkey.

Figure 4. Number of forecasts by country considered in the review.

3. Description of the Dataset

The analysis was performed from five perspectives: the class of a forecasting model (MSC or ML),
the type of model (hybrid or not), the time horizon, and the input variables and performance trend
Entropy 2020, 22, 1412 5 of 24

over time (MAPE). The dataset analyzed in this paper contains 257 entries associated with a MAPE
value linked to the best model proposed in the document.
The MAPE value was classified according to the criteria drawn up by [28], which contain typical
MAPE values for business and industrial data and their interpretation in four evaluation criteria (in our
case, four prediction capabilities); this table was used in [29–32], and can be seen in Table 1.

Table 1. MAPE qualitative criteria.

MAPE (%) Prediction Capability

<10 Highly accurate prediction (HAP)
10–20 Good prediction (GPR)
20–50 Reasonable prediction (RP)
>50 Inaccurate prediction (IPR)

Table 2 shows that of the 164 documents processed in the systematic review, 99 contain a highly
accurate prediction (HAP). Additionally, more ML documents with an HAP were found than MSC
documents. Regarding the sources of variability considered by the documents that contained HAPs,
it can be seen that multivariate models have a higher recurrence than univariate models. As explained
in [10], despite the introduction of artificial intelligence, each of the individual methods are still not
able to produce the desired outcomes because of their disadvantages. For instance, neural networks
attain local optimal results instead of global optimal results. Expert systems excessively rely on
knowledge and cannot always obtain optimal results, whereas grey prediction systems are suitable for
exponential growth models. Thus, by considering every method’s merits and taking full advantage of
them, the concept of hybrid and combination methods developed rapidly.

Table 2. Systematic review documents. Techniques type used in electric power forecast and qualitative
values of the average MAPE.

Multivariate Model Univariate Model Total

MAPE Type
Not Hybrid Hybrid Not Hybrid Hybrid
HAP ML [33–72] [6,27,73–91] [1,14,17,18,92–107] [29,32,85,108–122] 99
HAP MSC [15,123–147] [148,149] [1,14,17,18,92–107] [19,150–152] 52
GPR ML [153–157] [158] [159–162] − 10
GPR MSC [16,163] − − − 2
RP ML [164] − − − 1
Total 74 24 44 22 164

3.1. Forecasting Horizon

Figure 5 shows that the minimum MAPE values (<2) were reached more frequently when the
forecast time horizon is 5 min.
Entropy 2020, 22, 1412 6 of 24

Figure 5. MAPE value interval and the percent of forecasts extracted from the 164 documents
considered by time horizon.

3.2. Exogenous Influence

Because forecasting electric power demand is typically based on historical electricity consumption
and its relationship with exogenous influences, such as gross domestic product (GDP), population,
urbanization, income and exports, research on forecasting electric power demand has evolved using
both univariate and multivariate time-series models [15].
Similarly, weather associated variables such as humidity, temperature, and dew point are pertinent
for electric power forecasting for extensive time scales. For short-term forecasting such as minutes
ahead, the climate changes are already captured in the electric power series [165]. Forecasting models
using only previous electricity data (univariate) have been shown to provide HAP and to perform
better than models that also use weather variables as exogenous influences (multivariate) [166].
Nevertheless, the use of weather influences was found to be beneficial for electric power forecasting
horizons beyond several hours [166,167]. In Figure 6, it can be seen that the precision of the electric
energy forecast on average tends to improve when various sources of variability are considered.
In our sample of filtered documents, the average MAPE is lower for the forecasts whose models
consider sources of variability from the calendar information, weather information, and economic or
sociodemographic information, as in [60,125,130,146].

Figure 6. Average MAPE according to the considered source of variability.

4. Classes of Forecasting Models

From the multitude of methods that have been tested and evaluated, the ML and MSC classes
seem to be the main competitors.
Entropy 2020, 22, 1412 7 of 24

4.1. Classical Statistical Models

A popular technique such as time series forecasting is applied in several areas [25]. The most
widely used statistical method is the ARIMA of Box and Jenkins, which was applied with more force
during the eighties, when intelligent systems began to appear [168]. Several time series models make
use of the high autocorrelation for small lags in the time series of electric power, and supply electric
power forecasts using only previously measured values of electric power as input.
From the multitude of methods that have been tested and evaluated in this review, in this class,
regression analysis and ARIMA modeling seem to be the main competitors (Figure 7).

4.2. Classical Regression in the Time Series Context

To explain linear regression in the the context of time series, we assume some output or dependent
time series. Assume xt for t = 1, · · · , n, is being influenced by a collection of possible inputs or
independent series, such as zt1 , zt2 , · · · , ztq , where we first regard the inputs as fixed and known [169].
We express this relation through the linear regression model:

xt = β 0 + β 1 zt1 + β 2 zt2 + · · · + β q ztq + wt (2)

where β 0 , β 1 , · · · , β q are unknown fixed regression coefficients, and wt is a random error or noise
process consisting of independent and identically distributed (iid) normal variables with a zero mean
and variance σw2 . For time series regression, it is rarely the case that the noise is white, and we will
need to eventually relax that assumption.
Classical regression models have been used in several academic papers for electric power
forecasting [97,98,102,124,130,134,138,140], reaching an accuracy in the forecast with an average MAPE
value of 1.569%. Classical regression is often insufficient for explaining all of the interesting dynamics
of a time series; instead, the introduction of correlations that may be generated through lagged linear
relations led to the autoregressive (AR) and autoregressive moving average (ARMA) models that were
presented in [169,170]. Adding nonstationary models to the mix led to the autoregressive integrated
moving average (ARIMA) model popularized in the landmark work by Box and Jenkins [169,171].

4.3. Autoregressive Integrated Moving Average

Autoregressive models are based on the idea that the current value of the series, xt , can be
explained as a function of p past values, xt−1 , xt−2 , · · · , xt− p , where p determines the number of
previous steps required to forecast the current value [169].
The acronym ARIMA refers to an autoregressive integrated moving average model.
ARIMA models can be applied to non-stationary data, and when the data are seasonal, the SARIMA
model can be implemented. The ARIMA and SARIMA models have been used in many studies for
forecasting [14–16,99,100,103,127,136], reaching forecast accuracies with an average MAPE value of
3.214%. A typical ARIMA ( p, d, q) model can be expressed by Equation (3), where the variable ut is
replaced by a new variable wt obtained by differencing ut d times [25]:

w t = (1 − B ) d u t . (3)

4.4. Machine Learning (ML) Models

ML methods have been suggested in the academic literature as an alternative to MSC methods
for time series prediction, with the same objective. They attempt to improve the forecasting accuracy
precision by minimizing some loss functions, as for example the sum of squared errors. The distinction
between ML and MSC is in how the minimization is performed: the ML methods use nonlinear
algorithms while the MSC method use linear processes. The ML methods require a greater dependence
on computer science to be implemented and are more demanding than MSC methods, as they are
positioned at the intersection of MSC and computer science [172]. There are several approaches
Entropy 2020, 22, 1412 8 of 24

developed under ML theory. In this review, artificial neural networks (ANNs), support vector machines
(SVMs), decision trees (DTs), adaptive neuro fuzzy inference systems (ANFISs), and recurrent neural
networks (RNNs) were found to support the bases of the models that were implemented more
frequently in electric power forecasting (Figure 7).

Figure 7. Graph with nodes weighted by sample size for the case of the techniques (light-gray node)
and probability of occurrence for the MAPE intervals (dark-gray node).

4.4.1. Artificial Neural Networks (ANN)

Neural networks have been the subject of great interest for many decades due to the desire
to understand the brain and to build learning machines [173]. A neural network is an interconnected
assembly of simple processing elements, units or nodes whose functionalities are loosely based on animal neurons.
The processing ability of a network is stored in the inter-unit connection strengths, or weights, obtained by a
process of adaptation to, or learning from, a set of training patterns [174].
The ANN models have been used in many studies for electric power forecasting [6,39,44,46,48,53–
55,57–62,65,70,75,77,81,82,86,108,153,155,165,175,176] and have reached a forcasting accuracy with an
average MAPE value of 3.781%.

4.4.2. Recurrent Neural Networks (RNN)

Models known as a recurrent neural networks allow feedback connections; these models define
nonlinear dynamical systems but do not have simple probabilistic interpretations [173]. RNN models
have been used in many studies for electric power forecasting [64,69,71,73,88,90,157,177,178] and have
reached a forecasting accuracy with an average MAPE value of 3.610%.

4.4.3. Fuzzy Neural Network-Based Forecasting Methods

Fuzzy logic systems (or, simply, fuzzy systems (FSs)) and neural networks are universal
approximators; that is, they can approximate any nonlinear function (mapping) with any desired
accuracy and have found wide application in the identification, planning, and model-free control of
complex nonlinear systems, such as robotic systems and industrial processes. Fuzzy logic offers a
linguistic (approximate) approach to drawing conclusions from uncertain data, and neural networks
offer the capability of learning and training with or without a teacher (supervisor) [179].
Fuzzy logic algorithms have been used in many studies for electric power forecasting [10,27,33,
45,50,52,63,79,80,83,85,91,113,118,122,154,160,180,181] and have reached a forecasting accuracy with
an average MAPE value of 4.013%.
Entropy 2020, 22, 1412 9 of 24

4.4.4. Support Vector Machines (SVMs)

Support vector machines are supervised learning algorithms used for solving binary classification
and regression problems. The main idea of support vector machines is to construct a hyperplane such
that the margin of separation between the two classes is maximized. In this algorithm, each of the data
points is plotted as a data point in n-dimensional hyperspace. Then, a hyperplane that maximizes the
separation between the two classes is constructed [182]. This technique was originally designed for binary
classification but can be extended to regression and multiclass classification [173]. Support vector regression
algorithms have been used in many studies for electric power forecasting [41,47,49,76,109,110,112,114,
119,164,183–185] and have reached a forecasting accuracy with an average MAPE value of 4.326%.

5. Evaluation of Model Accuracy

As can be seen in Table 3, as mentioned in [77], there are many factors, such as economic
development, regional industrial production, holiday periods, weather conditions, social change,
electricity price, and population, that are unavoidable, affect electric power randomly, and allow the
data to demonstrate different features.
Short-term load forecast models that rely on weather information require the prediction of weather
parameters for the next few hours or at most the next few days [75]. Similarly, economic indicators and
electrical infrastructure measures are usually useful in forecasting electric power with a long forecast
horizon, e.g., a prediction of the annual peak load at least one year in advanced [39]. However, in the
daily peak load forecasting for the following month, these indicators are not effective, since the forecast
step and horizon are too short to observe their effect [75]; this behavior is shown in Table 3.
Similarly, from Figure 8, it can be seen that among the records of documents that reach HAP,
the average MAPE value is lower in the frameworks that implement hybrid models of ML and
multivariate dependency, such as those developed in [6,27,73–91]. To verify the hypotheses of the
differences in the means and variances in the MAPE, three hypothesis tests are carried out. Table 4
shows that for small and medium effects, the alternative hypothesis on the minor indicates that the
MAPE is accepted for the ML model, based on a hybrid method and with multivariate dependencies.

Figure 8. Boxplot of the papers included in the systematic review, with HAP-MAPE.
Entropy 2020, 22, 1412 10 of 24

Table 3. Summary of the papers included in the systematic review, with a HAP-MAPE, hybrid ML forecasting approach and multivariate model. The abbreviations
are displayed in Abbreviations Section. 1/ Average MAPE; 2/ Approximate sample size. The Classification of the forecasting models can be seen in Figure A1.

Ref Year Country Energy Type Technique Forecast Other Input MAPE 1/ N 2/ Scale Date Sample
[73] 2006 Australia No Specific ERNN; WT Electricity Load TM, HM, WS 0.794 26,297 Hours 1999 2002
[74] 2013 Iran Wind PSO; ACO Wind Power TM, WS 3.513 8736 Hours 2010 2011
[75] 2008 Iran No Specific ANN; EA Peak Load CI 1.760 26,280 Hours 1997 1999
[76] 2008 EEUU No Specific SVR; BT Electricity Load CI, TM, HM 1.960 30,144 Hours 2001 2004
[77] 2015 Australia No specific ANN Electrical power CI 3.710 70,080 Half-Hour 2006 2009
[78] 2017 UK No Specific PSO; ANN Load Demand CI, TM 1.723 8760 Hours 2008 2008
[79] 2016 Algeria No Specific HW-ES; KNN; WD; Peak Electricity TM 2.796 1064 Days 2012 2014
Fuzzy-CM; ANFIS
[80] 2010 Iran No Specific ANFIS Electricity GDP, POP, EXP, CPI 2.789 37 Years 1971 2007
[81] 2017 Poland No Specific ANN; PCA Power Load CI, TM 1.235 26,280 Hours 2009 2012
[82] 2018 India No Specific ANN; PSO; GA Electricity CPI, GDP 0.220 25 Years 1991 2015
Demand
[83] 2017 UK No Specific ELM; Fuzzy Electricity Load CI, TM, DP 1.435 43,852 Hours 2004 2008
[84] 2019 EEUU Wind NWP; WD; CNN Wind Power CI, TM, WS, DP 2.550 26,280 Hours 2015 2017
[85] 2008 Iran No Specific BNN; MCM; Fuzzy Load CI, TM 2.421 1460 Days 2004 2007
[27] 2018 Vietnam No Specific WT; ANFIS; COA Electricity CI, TM, HM, PRS, 4.330 132 Months 2003 2013
RFL, RT, WS
[86] 2017 UK No Specific ANN; JOA Electricity Load CI, TM, DP 5.710 52,560 Hours 2004 2009
[87] 2015 India No Specific ANN; BBO Electrical Energy GDP, POP 2.510 33 Years 1980 2012
[88] 2019 China Wind GM; ERNN; BP Power Generation TM, HM, WS, WDD, 3.730 1441 15 min 2016 2016
PRS
[6] 2019 Australia No Specific ANN; BOOT Electricity 57 Index 5.290 4300 6h 2014 2017
[89] 2019 Uganda No Specific PSO; ABC Electricity POP, GDP, EP, NS 1.306 17 Years 1990 2016
[90] 2020 Australia Photovoltaic WD; LSTM Power TM, HM, WS, HR 1.868 213,984 5 min 2014 2016
[91] 2018 Turkey No Specific ANFIS Electrical Load CI, TM 8.869 8760 Hours 2017 2017
Entropy 2020, 22, 1412 11 of 24

Table 4. Hypothesis Test for Difference in Means. 1/ Levene Test (p-value); 2/ test (p-value).

Effect Size
Variable Hypothesis ( H0 ) Homogeneity of Variance 1/ Difference in Means 2/
(Cohen’s)
Model µ ML ≥ µ MSC 0.00386 0.07252 Small
Hybrid µYes ≥ µ No 0.09063 0.04321 Small
Dependency µ Multi ≥ µUni 0.00125 0.00059 Medium

In this sense, a summary of the documents found in the review with MAPEs and HAPs that base
their models on ML with a hybrid approach and multivariate dependence is presented in the Table 3.
When we analyzed these documents, we observed that there are common elements; for example,
when building a word cloud from the abstracts, keywords and titles of these documents, we can identify
that in 24% of the cases multiple scale decomposition and wavelet theory (WT) were mentioned.
The wavelet transform, including filtering and forecasting, has been suggest for detailed
examination of the elements or structure of time series in several academic papers in recent years [73].
WT has been extensively implemented in electric power forecasting for decomposing electricity series
into series with particular characteristics that can be predicted more accurately than the original time
series [186–188].

6. Case Study
In this section, we propose a hybrid model to forecast the electric power by using a type of
recurrent artificial neural network known as long short-term memory (LSTM), developed by [189];
we also implemented wavelet decomposition for the data preprocessing (WD-LSTM), as was used
in [90]. We use the acronym WD-LSTM for the proposed hybrid model. The results were compared
with those of traditional neural network models (LSTM), as was applied in [71,177] and with results of
the lagged regression analysis as in [96].
The performance of this methods is demonstrated with a case study using an actual dataset
collected from Chile (Table 5). The objective is to illustrated the approach that allows the electric power
demand forecasting, in terms of its lagged values, identifying the type of model that tends to show
better forecast accuracy.
Maximum daily and hourly electric power demand data over a diverse period were used (Table 5).
Figure 9 shows that there is a regularity in electric power demand data. We observe a clear pattern
based on the year and day of the week. The electric power demand also follows a group of patterns
within any day and depending on the time of the day.

Table 5. Electric power demand object of forecasting.

Set Size
Type Variable Date
Training Validation Test
Local Dmax Maximum daily electricity demand ( MW ). 2006–2019 2475 1516 1062
Energy Hed Hourly electricity demand ( MW ). 2016–2020 865 371 530

The values of four performance evaluation indicators—RMSE: root mean square error,
MAE: mean absolute error, R2 : coefficient of determination, and MAPE: mean absolute percentage
error—showed that the hybrid deep learning model (WD-LSTM) exhibits superior performance in
both forecasting accuracy and stability.
Entropy 2020, 22, 1412 12 of 24

Figure 9. Electric power demand series (MW) in Chile.

Figures 10 and 11 provide the comparative hourly and daily-ahead performance results for
three types of days (weekday, weekend, and all days). They likewise provide the performance
evaluation results of regression, LSTM, and WD-LSTM applied to each dataset (training, validation,
and test). The hybrid deep learning model (WD-LSTM) had the best performance of all forecasting
models. The WD-LSTM method generated forecasting results with the lowest MAE, MAPE, and RMSE
and with the higher R2 in most cases. The results further reveal the robustness of the hybrid deep
learning model. The superior accuracy of the hybrid model is primarily due to the deep learning
framework comprising between two and four independent LSTM networks, which provide an
effective means to approximate inherent invariant features and structures. In addition, the low-
and high-frequency components exhibited in the electric power datasets can be better extracted by
wavelet decomposition. Likewise, each LSTM network managed to focus more on capturing the linear
and nonlinear relationships in the energy series, which could not be done with the lagged regression,
at least in non-linear cases.

Figure 10. Performance evaluations of different methods for each type of day.
Entropy 2020, 22, 1412 13 of 24

Figure 11. Performance evaluations of different methods for hour.

7. Discussion and Conclusions

This paper presented a systematic review of the forecasting models for electric power from
the last 15 years based on ML and MSC techniques. We presented an in-depth analysis of the
performance of electric power forecasting models and compared different forecasting models based on
their MAPE values. A rigorous framework for comparing different classes of models was introduced,
thus generating a reliable picture of the state-of-the-art models’ accuracy of electric power forecasting.
We were able to identify that a large number of techniques are being used and are aimed at forecasting
electrical energy; the techniques with the greatest use are in the fields of ML and ANNs, followed by
Entropy 2020, 22, 1412 14 of 24

those that implement algorithms with fuzzy logic and RNNs, while in the MSC area, the use of ARIMA
models and regression analysis predominates.
The results can be stratified from three perspectives. The forecasting models (I) from the hybrid
class, (II) of multivariate dependency, and (III) based on the machine learning approach demonstrate
the best performance for electric power forecasting. Regarding the hybrid models, it is highlighted that
24% of the adjustments with the greatest forecasting precision merged wavelet theory into their models.
With regard to multivariate models, we were able to identify that those models that incorporate various
sources of variability in their adjustment tend to have, on average, greater precision in their forecasts.
A case of study was presented, in which the implementation of MSC and ML models was
compared; we found that the linear models, such as lagged regression, are relatively simple and cannot
capture with precision the inherent nonlinear structure of the electric power time series, whereas the
deep learning models implemented have a better performance.
Likewise, it was observed that when decomposing the series according to the type of day of
electricity consumption (workday or weekend), the models tend to have better forecast accuracy and,
in the same way, forecasting errors are minimized by reducing the time horizon (hourly).
Due to electric power systems’ participation in the growing trend of environmental optimization
around the world, a substantial increase in the contribution of diverse sources to the energy generation
is observed. This trend brings about challenges in terms of electric power generation and distribution
system operation, because the dimension and complexity of such advances, among other aspects,
require the use of a computational intelligence systems that act as sources of data and deal with the
control, management, and trading needs at the distribution level in an efficient and robust manner.
In this sense, further research could deepen the understanding of the relationship between the type of
energy, climate, preprocessing techniques, and performance of machine learning models under various
normalized metrics of residuals.

Author Contributions: Conceptualization, E.V., H.A.-C. and R.S.; methodology, E.V., H.A-.C. and R.S.; software,
E.V. and H.A.-C.; validation, all authors; writing—original draft preparation, E.V.; writing—review and editing,
all authors; supervision, H.A.-C. All authors have read and agreed to the published version of the manuscript.
Funding: This research received internal funding form the Pontificia Universidad Católica de Valparaíso (PUCV).
Acknowledgments: We are grateful to the School of Informatic Engineering at Pontificia Universidad Católica
de Valparaíso (PUCV) of Chile for the Scholarship Programme under which the first author is funded for her
Doctoral study. Héctor Allende-Cid’s work was funded by project 039.457/2020 from VRIEA PUCV.
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design of the
study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to
publish the results.

Abbreviations
The following abbreviations are used in this manuscript:

Nomenclature
Artificial bee colony ABC Gray model GM
Ant Colony Optimization ACO Gross domestic product GDP
Adaptive Neuro Fuzzy Inference System ANFIS Hybrid Monte Carlo HCM
Artificial Neural Network ANN Humidity HM
Autoregressive Integrated Moving Aevrage ARIMA Horizontal Radiation HR
Bayesian Clustering by Dynamics BCD Holt Winters HW
Bayesian neural network BNN Jaya optimization algorithm JOA
Biogeography based optimization BOA K-nearest neighbors KNN
Back propagation BP Long short term memory LSTM
Calendar information CI Number of subscribers NS
Convolution Neural Network CNN Numerical Weather Prediction NWP
Entropy 2020, 22, 1412 15 of 24

Cuckoo Search Algorithm COA Principal component analysis PCA

Consumer Price Index CPI Population POP
Deep Belief Network DBN Air pressure PRS
Dew point DP Particle Swarm Optimization PSO
Evolutionary Algorithm EA Radial basis function network RBF
Extreme learning machine ELM Rainfall RFL
Electricity price EP Rainy time RT
Elman Recurrent Neural Network ERNN Recurrent Neural Network RNN
Exponential smoothing ES Regression Analysis RA
Exports EXP Support Vector Regression SVR
Fuzzy Neural Network FNN Temperature TM
Gaussian Process GP Wind direction WDD
Genetic algorithm GA Wind speed WS
Generalized Additive Model GAM Wavelet theory WT

Appendix A

Figure A1. Classification of the forecasting models.

References
1. Taylor, J.W.; De Menezes, L.M.; McSharry, P.E. A comparison of univariate methods for forecasting electricity
demand up to a day ahead. Int. J. Forecast. 2006, 22, 1–16. [CrossRef]
2. Banos, R.; Manzano-Agugliaro, F.; Montoya, F.; Gil, C.; Alcayde, A.; Gómez, J. Optimization methods
applied to renewable and sustainable energy: A review. Renew. Sustain. Energy Rev. 2011, 15, 1753–1766.
[CrossRef]
3. Sharma, A.; Kakkar, A. Forecasting daily global solar irradiance generation using machine learning.
Renew. Sustain. Energy Rev. 2018, 82, 2254–2269. [CrossRef]
4. Li, J.; Ward, J.K.; Tong, J.; Collins, L.; Platt, G. Machine learning for solar irradiance forecasting of photovoltaic
system. Renew. Energy 2016, 90, 542–553. [CrossRef]
5. Rahman, M.N.; Esmailpour, A.; Zhao, J. Machine learning with big data an efficient electricity generation
forecasting system. Big Data Res. 2016, 5, 9–15. [CrossRef]
6. Al-Musaylh, M.S.; Deo, R.C.; Adamowski, J.F.; Li, Y. Short-term electricity demand forecasting using machine
learning methods enriched with ground-based climate and ECMWF Reanalysis atmospheric predictors in
southeast Queensland, Australia. Renew. Sustain. Energy Rev. 2019, 113, 109293. [CrossRef]
7. Kim, S.G.; Jung, J.Y.; Sim, M.K. A Two-Step Approach to Solar Power Generation Prediction Based on
Weather Data Using Machine Learning. Sustainability 2019, 11, 1501. [CrossRef]
Entropy 2020, 22, 1412 16 of 24

8. Abedinia, O.; Raisz, D.; Amjady, N. Effective prediction model for Hungarian small-scale solar power output.
IET Renew. Power Gener. 2017, 11, 1648–1658. [CrossRef]
9. Abuella, M.; Chowdhury, B. Improving combined solar power forecasts using estimated ramp rates:
Data-driven post-processing approach. IET Renew. Power Gener. 2018, 12, 1127–1135. [CrossRef]
10. Yang, Y.; Chen, Y.; Wang, Y.; Li, C.; Li, L. Modelling a combined method based on ANFIS and neural network
improved by DE algorithm: A case study for short-term electricity demand forecasting. Appl. Soft Comput.
2016, 49, 663–675. [CrossRef]
11. Fiot, J.B.; Dinuzzo, F. Electricity demand forecasting by multi-task learning. IEEE Trans. Smart Grid 2016,
9, 544–551. [CrossRef]
12. Wang, F.; Yu, Y.; Zhang, Z.; Li, J.; Zhen, Z.; Li, K. Wavelet decomposition and convolutional LSTM networks
based improved deep learning model for solar irradiance forecasting. Appl. Sci. 2018, 8, 1286. [CrossRef]
13. Liu, Y.; Guan, L.; Hou, C.; Han, H.; Liu, Z.; Sun, Y.; Zheng, M. Wind Power Short-Term Prediction Based on
LSTM and Discrete Wavelet Transform. Appl. Sci. 2019, 9, 1108. [CrossRef]
14. Jin, X.; Dong, Y.; Wu, J.; Wang, J. An improved combined forecasting method for electric power load based
on autoregressive integrated moving average model. In Proceedings of the 2010 International Conference of
Information Science and Management Engineering, Xi’an, China, 7–8 August 2010; Volume 2, pp. 476–480.
15. Sarkodie, S.A. Estimating Ghana’s electricity consumption by 2030: An ARIMA forecast. Energy Sources Part
B Econ. Plan. Policy 2017, 12, 936–944. [CrossRef]
16. Barzola-Monteses, J.; Mite-León, M.; Espinoza-Andaluz, M.; Gómez-Romero, J.; Fajardo, W. Time Series
Analysis for Predicting Hydroelectric Power Production: The Ecuador Case. Sustainability 2019, 11, 6539.
[CrossRef]
17. Filik, U.B.; Gerek, O.N.; Kurban, M. Hourly forecasting of long term electric energy demand using a novel
modeling approach. In Proceedings of the 2009 Fourth International Conference on Innovative Computing,
Information and Control (ICICIC), Kaohsiung, Taiwan, 7–9 December 2009; pp. 115–118.
18. Filik, Ü.B.; Gerek, Ö.N.; Kurban, M. A novel modeling approach for hourly forecasting of long-term electric
energy demand. Energy Convers. Manag. 2011, 52, 199–211. [CrossRef]
19. Liang, J.; Liang, Y. Analysis and modeling for China’s electricity demand forecasting based on a new
mathematical hybrid method. Information 2017, 8, 33. [CrossRef]
20. Khan, A.R.; Mahmood, A.; Safdar, A.; Khan, Z.A.; Khan, N.A. Load forecasting, dynamic pricing and DSM
in smart grid: A review. Renew. Sustain. Energy Rev. 2016, 54, 1311–1322. [CrossRef]
21. Mosavi, A.; Salimi, M.; Faizollahzadeh Ardabili, S.; Rabczuk, T.; Shamshirband, S.; Varkonyi-Koczy, A.R.
State of the art of machine learning models in energy systems, a systematic review. Energies 2019, 12, 1301.
[CrossRef]
22. Qasem, S.N.; Samadianfard, S.; Sadri Nahand, H.; Mosavi, A.; Shamshirband, S.; Chau, K.w. Estimating
daily dew point temperature using machine learning algorithms. Water 2019, 11, 582. [CrossRef]
23. Kalogirou, S.A. Applications of artificial neural-networks for energy systems. Appl. Energy 2000, 67, 17–35.
[CrossRef]
24. Voyant, C.; Notton, G.; Kalogirou, S.; Nivet, M.L.; Paoli, C.; Motte, F.; Fouilloy, A. Machine learning methods
for solar radiation forecasting: A review. Renew. Energy 2017, 105, 569–582. [CrossRef]
25. Blaga, R.; Sabadus, A.; Stefu, N.; Dughir, C.; Paulescu, M.; Badescu, V. A current perspective on the accuracy
of incoming solar energy forecasting. Prog. Energy Combust. Sci. 2019, 70, 119–144. [CrossRef]
26. Verdejo, H.; Awerkin, A.; Becker, C.; Olguin, G. Statistic linear parametric techniques for residential electric
energy demand forecasting. A review and an implementation to Chile. Renew. Sustain. Energy Rev. 2017,
74, 512–521. [CrossRef]
27. Chen, J.F.; Do, Q.H.; Nguyen, T.V.A.; Doan, T.T.H. Forecasting monthly electricity demands by wavelet
neuro-fuzzy system optimized by heuristic algorithms. Information 2018, 9, 51. [CrossRef]
28. Lewis, C.D. Industrial and Business Forecasting Methods: A Practical Guide to Exponential Smoothing and Curve
Fitting; Butterworth-Heinemann: Oxford, UK, 1982.
29. Zhao, H.; Guo, S. An optimized grey model for annual power load forecasting. Energy 2016, 107, 272–286.
[CrossRef]
30. Ma, X.; Liu, Z. Application of a novel time-delayed polynomial grey model to predict the natural gas
consumption in China. J. Comput. Appl. Math. 2017, 324, 17–24. [CrossRef]
Entropy 2020, 22, 1412 17 of 24

31. Lee, Y.S.; Tong, L.I. Forecasting energy consumption using a grey model improved by incorporating genetic
programming. Energy Convers. Manag. 2011, 52, 147–152. [CrossRef]
32. Wang, J.; Du, P.; Lu, H.; Yang, W.; Niu, T. An improved grey model optimized by multi-objective ant lion
optimization algorithm for annual electricity consumption forecasting. Appl. Soft Comput. 2018, 72, 321–337.
[CrossRef]
33. Chang, P.C.; Fan, C.Y.; Lin, J.J. Monthly electricity demand forecasting based on a weighted evolving fuzzy
neural network approach. Int. J. Electr. Power Energy Syst. 2011, 33, 17–27. [CrossRef]
34. Dedinec, A.; Filiposka, S.; Dedinec, A.; Kocarev, L. Deep belief network based electricity load forecasting:
An analysis of Macedonian case. Energy 2016, 115, 1688–1700. [CrossRef]
35. Chang, G.; Lu, H.; Chang, Y.; Lee, Y. An improved neural network-based approach for short-term wind
speed and power forecast. Renew. Energy 2017, 105, 301–311. [CrossRef]
36. Kaboli, S.H.A.; Selvaraj, J.; Rahim, N. Long-term electric energy consumption forecasting via artificial
cooperative search algorithm. Energy 2016, 115, 857–871. [CrossRef]
37. El-Telbany, M.; El-Karmi, F. Short-term forecasting of Jordanian electricity demand using particle swarm
optimization. Electr. Power Syst. Res. 2008, 78, 425–433. [CrossRef]
38. Ardakani, F.; Ardehali, M. Long-term electrical energy consumption forecasting for developing and developed
economies based on different optimized models and historical data types. Energy 2014, 65, 452–461. [CrossRef]
39. Tsekouras, G.J.; Hatziargyriou, N.D.; Dialynas, E.N. An optimized adaptive neural network for annual
midterm energy forecasting. IEEE Trans. Power Syst. 2006, 21, 385–391. [CrossRef]
40. López, M.; Valero, S.; Senabre, C.; Aparicio, J.; Gabaldon, A. Application of SOM neural networks to
short-term load forecasting: The Spanish electricity market case study. Electr. Power Syst. Res. 2012, 91, 18–27.
[CrossRef]
41. Setiawan, A.; Koprinska, I.; Agelidis, V.G. Very short-term electricity load demand forecasting using support
vector regression. In Proceedings of the 2009 International Joint Conference on Neural Networks, Atlanta,
GA, USA, 14–19 June 2009; pp. 2888–2894.
42. Al-Qahtani, F.H.; Crone, S.F. Multivariate k-nearest neighbour regression for time series data—A novel
algorithm for forecasting UK electricity demand. In Proceedings of the 2013 International Joint Conference
on Neural Networks (IJCNN), Dallas, TX, USA, 4–9 August 2013; pp. 1–8.
43. Guo, Z.; Zhou, K.; Zhang, X.; Yang, S. A deep learning model for short-term power load and probability
density forecasting. Energy 2018, 160, 1186–1200. [CrossRef]
44. Ghanbari, A.; Naghavi, A.; Ghaderi, S.; Sabaghian, M. Artificial Neural Networks and regression approaches
comparison for forecasting Iran’s annual electricity load. In Proceedings of the 2009 International Conference
on Power Engineering, Energy and Electrical Drives, Lisbon, Portugal, 18–20 March 2009; pp. 675–679.
45. Elias, C.N.; Hatziargyriou, N.D. An annual midterm energy forecasting model using fuzzy logic. IEEE Trans.
Power Syst. 2009, 24, 469–478. [CrossRef]
46. Sahay, K.B.; Tripathi, M. Day ahead hourly load forecast of PJM electricity market and ISO New
England market by using artificial neural network. In Proceedings of the ISGT 2014, Istanbul, Turkey,
15–18 October 2014; pp. 1–5.
47. Wang, J.; Zhou, Y.; Chen, X. Electricity load forecasting based on support vector machines and simulated
annealing particle swarm optimization algorithm. In Proceedings of the 2007 IEEE International Conference
on Automation and Logistics, Jinan, China, 18–21 August 2007; pp. 2836–2841.
48. Ehsan, R.M.; Simon, S.P.; Venkateswaran, P. Day-ahead forecasting of solar photovoltaic output power using
multilayer perceptron. Neural Comput. Appl. 2017, 28, 3981–3992. [CrossRef]
49. De Leone, R.; Pietrini, M.; Giovannelli, A. Photovoltaic energy production forecast using support vector
regression. Neural Comput. Appl. 2015, 26, 1955–1962. [CrossRef]
50. Vahabie, A.; Yousefi, M.R.; Araabi, B.; Lucas, C.; Barghinia, S.; Ansarimehr, P. Mutual information based
input selection in neuro-fuzzy modeling for short term load forecasting of iran national power system.
In Proceedings of the 2007 IEEE International Conference on Control and Automation, Guangzhou, China,
30 May–1 June 2007; pp. 2710–2715.
51. Ouyang, T.; He, Y.; Li, H.; Sun, Z.; Baek, S. Modeling and forecasting short-term power load with copula
model and deep belief network. IEEE Trans. Emerg. Top. Comput. Intell. 2019, 3, 127–136. [CrossRef]
52. Barzamini, R.; Menhaj, M.; Khosravi, A.; Kamalvand, S. Short term load forecasting for iran national power
system and its regions using multi layer perceptron and fuzzy inference systems. In Proceedings of the 2005
Entropy 2020, 22, 1412 18 of 24

IEEE International Joint Conference on Neural Networks, Montreal, QC, Canada, 31 July–4 August 2005;
Volume 4, pp. 2619–2624.
53. Dilhani, M.S.; Jeenanunta, C. Effect of Neural Network structure for daily electricity load forecasting.
In Proceedings of the 2017 Moratuwa Engineering Research Conference (MERCon), Moratuwa, Sri Lanka,
29–31 May 2017; pp. 419–424.
54. Chen, J.F.; Lo, S.K.; Do, Q.H. Forecasting monthly electricity demands: An application of neural networks
trained by heuristic algorithms. Information 2017, 8, 31. [CrossRef]
55. Azadeh, A.; Ghaderi, S.F.; Sheikhalishahi, M.; Nokhandan, B.P. Optimization of short load forecasting in
electricity market of Iran using artificial neural networks. Optim. Eng. 2014, 15, 485–508. [CrossRef]
56. Raza, M.Q.; Baharudin, Z.; Nallagownden, P. A comparative analysis of PSO and LM based NN short
term load forecast with exogenous variables for smart power generation. In Proceedings of the 2014
5th International Conference on Intelligent and Advanced Systems (ICIAS), Kuala Lumpur, Malaysia,
3–5 June 2014; pp. 1–6.
57. Ishik, M.Y.; Göze, T.; Özcan, İ.; Güngör, V.Ç.; Aydın, Z. Short term electricity load forecasting: A case study of
electric utility market in Turkey. In Proceedings of the 2015 3rd International Istanbul Smart Grid Congress
and Fair (ICSG), Istanbul, Turkey, 29–30 April 2015; pp. 1–5.
58. Sahay, K.B.; Kumar, N.; Tripathi, M. Short-term load forecasting of Ontario Electricity Market by considering
the effect of temperature. In Proceedings of the 2014 6th IEEE Power India International Conference
(PIICON), Delhi, India, 5–7 December 2014; pp. 1–6.
59. Azadeh, A.; Ghadrei, S.; Nokhandan, B.P. One day ahead load forecasting for electricity market of Iran by
ANN. In Proceedings of the 2009 International Conference on Power Engineering, Energy and Electrical
Drives, Lisbon, Portugal, 18–20 March 2009; pp. 670–674.
60. Akole, M.; Tyagi, B. Artificial neural network based short term load forecasting for restructured power
system. In Proceedings of the 2009 International Conference on Power Systems, Kharagpur, India,
27–29 December 2009; pp. 1–7.
61. Chuang, F.K.; Hung, C.Y.; Chang, C.Y.; Kuo, K.C. Deploying arima and artificial neural networks models to
predict energy consumption in Taiwan. Sens. Lett. 2013, 11, 2333–2340. [CrossRef]
62. Kavanagh, K.; Barrett, M.; Conlon, M. Short-term electricity load forecasting for the integrated single
electricity market (I-SEM). In Proceedings of the 2017 52nd International Universities Power Engineering
Conference (UPEC), Heraklion, Greece, 28–31 August 2017; pp. 1–7.
63. Sarkar, M.R.; Rabbani, M.G.; Khan, A.R.; Hossain, M.M. Electricity demand forecasting of Rajshahi City in
Bangladesh using fuzzy linear regression model. In Proceedings of the 2015 International Conference on
Electrical Engineering and Information Communication Technology (ICEEICT), Savar, Dhaka, Bangladesh,
21–23 May 2015; pp. 1–3.
64. Kelo, S.M.; Dudul, S.V. Development of an intelligent system for short-term electrical power load forecasting
in Maharashtra state. In Proceedings of the 2008 Joint International Conference on Power System Technology
and IEEE Power India Conference, New Delhi, India, 12–15 October 2008; pp. 1–8.
65. Gireeshma, K.; Atla, C.R.; Rao, K.L. New Correlation Technique for RE Power Forecasting using Neural
Networks. In Proceedings of the 2019 Fifth International Conference on Electrical Energy Systems (ICEES),
Chennai, India, 21–22 February 2019; pp. 1–6.
66. Barzamini, R.; Menhaj, M.; Kamalvand, S.; Tajbakhsh, A. Short term load forecasting of Iran national power
system using artificial neural network generation two. In Proceedings of the 2005 IEEE Russia Power Tech,
St. Petersburg, Russia, 27–30 June 2005; pp. 1–5.
67. Viet, D.T.; Van Phuong, V.; Duong, M.Q.; Kies, A.; Schyska, B.U.; Wu, Y.K. A Short-Term Wind Power
Forecasting Tool for Vietnamese Wind Farms and Electricity Market. In Proceedings of the 2018 4th
International Conference on Green Technology and Sustainable Development (GTSD), Ho Chi Minh City,
Vietnam, 23–24 November 2018; pp. 130–135.
68. Sakunthala, K.; Iniyan, S.; Mahalingam, S. Forecasting energy consumption in Tamil Nadu using hybrid
heuristic based regression model. Therm. Sci. 2019, 23, 2885–2894. [CrossRef]
69. Siridhipakul, C.; Vateekul, P. Multi-step Power Consumption Forecasting in Thailand Using Dual-Stage
Attentional LSTM. In Proceedings of the 2019 11th International Conference on Information Technology and
Electrical Engineering (ICITEE), Pattaya, Thailand, 10–11 October 2019; pp. 1–6.
Entropy 2020, 22, 1412 19 of 24

70. Tanidir, Ö.; Tör, O. Accuracy of ANN based day-ahead load forecasting in Turkish power system: Degrading
and improving factors. Neural Netw. World 2015, 25, 443. [CrossRef]
71. Liu, J.; Wang, X.; Zhao, Y.; Dong, B.; Lu, K.; Wang, R. Heating Load Forecasting for Combined Heat and
Power Plants Via Strand-Based LSTM. IEEE Access 2020, 8, 33360–33369. [CrossRef]
72. Chen, S.; Huang, J. Forecasting China’s primary energy demand based on an improved AI model. Chin. J.
Popul. Resour. Environ. 2018, 16, 36–48. [CrossRef]
73. Benaouda, D.; Murtagh, F.; Starck, J.L.; Renaud, O. Wavelet-based nonlinear multiscale decomposition
model for electricity load forecasting. Neurocomputing 2006, 70, 139–154. [CrossRef]
74. Rahmani, R.; Yusof, R.; Seyedmahmoudian, M.; Mekhilef, S. Hybrid technique of ant colony and particle
swarm optimization for short term wind energy forecasting. J. Wind. Eng. Ind. Aerodyn. 2013, 123, 163–170.
[CrossRef]
75. Amjady, N.; Keynia, F. Mid-term load forecasting of power systems by a new prediction method.
Energy Convers. Manag. 2008, 49, 2678–2687. [CrossRef]
76. Fan, S.; Chen, L.; Lee, W.J. Machine learning based switching model for electricity load forecasting.
Energy Convers. Manag. 2008, 49, 1331–1344. [CrossRef]
77. Xiao, L.; Wang, J.; Yang, X.; Xiao, L. A hybrid model based on data preprocessing for electrical power
forecasting. Int. J. Electr. Power Energy Syst. 2015, 64, 311–327. [CrossRef]
78. Raza, M.Q.; Nadarajah, M.; Hung, D.Q.; Baharudin, Z. An intelligent hybrid short-term load forecasting
model for smart power grids. Sustain. Cities Soc. 2017, 31, 264–275. [CrossRef]
79. Laouafi, A.; Mordjaoui, M.; Laouafi, F.; Boukelia, T.E. Daily peak electricity demand forecasting based on an
adaptive hybrid two-stage methodology. Int. J. Electr. Power Energy Syst. 2016, 77, 136–144. [CrossRef]
80. Ghanbari, A.; Ghaderi, S.F.; Azadeh, M.A. Adaptive Neuro-Fuzzy Inference System vs. Regression based
approaches for annual electricity load forecasting. In Proceedings of the 2010 2nd International Conference
on Computer and Automation Engineering (ICCAE), Singapore, 26–28 February 2010; Volume 5, pp. 26–30.
81. Brodowski, S.; Bielecki, A.; Filocha, M. A hybrid system for forecasting 24-h power load profile for Polish
electric grid. Appl. Soft Comput. 2017, 58, 527–539. [CrossRef]
82. Anand, A.; Suganthi, L. Hybrid GA-PSO optimization of artificial neural network for forecasting electricity
demand. Energies 2018, 11, 728. [CrossRef]
83. Yeom, C.U.; Kwak, K.C. Short-term electricity-load forecasting using a TSK-based extreme learning machine
with knowledge representation. Energies 2017, 10, 1613. [CrossRef]
84. Mujeeb, S.; Alghamdi, T.A.; Ullah, S.; Fatima, A.; Javaid, N.; Saba, T. Exploiting Deep Learning for Wind
Power Forecasting Based on Big Data Analytics. Appl. Sci. 2019, 9, 4417. [CrossRef]
85. Barghinia, S.; Kamankesh, S.; Mahdavi, N.; Vahabie, A.; Gorji, A. A combination method for short term
load forecasting used in Iran electricity market by NeuroFuzzy, Bayesian and finding similar days methods.
In Proceedings of the 2008 5th International Conference on the European Electricity Market, Lisbon, Portugal,
28–30 May 2008; pp. 1–6.
86. Singh, P.; Mishra, K.; Dwivedi, P. Enhanced hybrid model for electricity load forecast through artificial
neural network and Jaya algorithm. In Proceedings of the 2017 International Conference on Intelligent
Computing and Control Systems (ICICCS), Madurai, India, 15–16 June 2017; pp. 115–120.
87. Kumaran, J.; Ravi, G. Long-term sector-wise electrical energy forecasting using artificial neural network and
biogeography-based optimization. Electr. Power Compon. Syst. 2015, 43, 1225–1235. [CrossRef]
88. Zhou, J.; Xu, X.; Huo, X.; Li, Y. Forecasting models for wind power using extreme-point symmetric mode
decomposition and artificial neural networks. Sustainability 2019, 11, 650. [CrossRef]
89. Kasule, A.; Ayan, K. Forecasting Uganda’s net electricity consumption using a hybrid PSO-ABC Algorithm.
Arab. J. Sci. Eng. 2019, 44, 3021–3031. [CrossRef]
90. Li, P.; Zhou, K.; Lu, X.; Yang, S. A hybrid deep learning model for short-term PV power forecasting.
Appl. Energy 2020, 259, 114216. [CrossRef]
91. Kaysal, A.; Köroglu, S.; Oguz, Y.; Kaysal, K. Artificial Neural Networks and Adaptive Neuro-Fuzzy Inference
Systems Approaches to Forecast the Electricity Data for Load Demand, an Analysis of Dinar District Case.
In Proceedings of the 2018 2nd International Symposium on Multidisciplinary Studies and Innovative
Technologies (ISMSIT), Ankara, Turkey, 19–21 October 2018; pp. 1–6.
92. Soares, L.J.; Medeiros, M.C. Modeling and forecasting short-term electricity load: A comparison of methods
with an application to Brazilian data. Int. J. Forecast. 2008, 24, 630–644. [CrossRef]
Entropy 2020, 22, 1412 20 of 24

93. Almeshaiei, E.; Soltan, H. A methodology for electric power load forecasting. Alex. Eng. J. 2011, 50, 137–144.
[CrossRef]
94. Hamzacebi, C.; Es, H.A. Forecasting the annual electricity consumption of Turkey using an optimized grey
model. Energy 2014, 70, 165–171. [CrossRef]
95. Zhao, W.; Wang, J.; Lu, H. Combining forecasts of electricity consumption in China with time-varying
weights updated by a high-order Markov chain model. Omega 2014, 45, 80–91. [CrossRef]
96. Wang, C.h.; Grozev, G.; Seo, S. Decomposition and statistical analysis for regional electricity demand
forecasting. Energy 2012, 41, 313–325. [CrossRef]
97. Yukseltan, E.; Yucekaya, A.; Bilge, A.H. Forecasting electricity demand for Turkey: Modeling periodic
variations and demand segregation. Appl. Energy 2017, 193, 287–296. [CrossRef]
98. Koprinska, I.; Rana, M.; Agelidis, V.G. Yearly and seasonal models for electricity load forecasting.
In Proceedings of the 2011 International Joint Conference on Neural Networks, San Jose, CA, USA,
31 July–5 August 2011; pp. 1474–1481.
99. Rallapalli, S.R.; Ghosh, S. Forecasting monthly peak demand of electricity in India—A critique. Energy Policy
2012, 45, 516–520. [CrossRef]
100. Damrongkulkamjorn, P.; Churueang, P. Monthly energy forecasting using decomposition method with
application of seasonal ARIMA. In Proceedings of the 2005 International Power Engineering Conference,
Singapore, 29 November–2 December 2005; pp. 1–229.
101. Alamaniotis, M.; Bargiotas, D.; Tsoukalas, L.H. Towards smart energy systems: Application of kernel machine
regression for medium term electricity load forecasting. SpringerPlus 2016, 5, 58. [CrossRef] [PubMed]
102. Koprinska, I.; Sood, R.; Agelidis, V. Variable selection for five-minute ahead electricity load forecasting.
In Proceedings of the 2010 20th International Conference on Pattern Recognition, Istanbul, Turkey, 23–26
August 2010; pp. 2901–2904.
103. He, H.; Liu, T.; Chen, R.; Xiao, Y.; Yang, J. High frequency short-term demand forecasting model for
distribution power grid based on ARIMA. In Proceedings of the 2012 IEEE International Conference on
Computer Science and Automation Engineering (CSAE), Zhangjiajie, China, 25–27 May 2012; Volume 3,
pp. 293–297.
104. An, Y.; Zhou, Y.; Li, R. Forecasting India’s Electricity Demand Using a Range of Probabilistic Methods.
Energies 2019, 12, 2574. [CrossRef]
105. Kantanantha, N.; Runsewa, S. Forecasting of electricity demand to reduce the inventory cost of imported
coal. In Proceedings of the 2017 4th International Conference on Industrial Engineering and Applications
(ICIEA), Nagoya, Japan, 21–23 April 2017; pp. 336–340.
106. Chen, H.; Li, F.; Wang, Y. Component GARCH-M type models for wind power forecasting. In Proceedings
of the 2015 IEEE Power & Energy Society General Meeting, Denver, CO, USA, 26–30 July 2015; pp. 1–5.
107. Kusakci, A.O.; Ayvaz, B. Electrical energy consumption forecasting for Turkey using grey forecasting technics
with rolling mechanism. In Proceedings of the 2015 2nd International Conference on Knowledge-Based
Engineering and Innovation (KBEI), Tehran, Iran, 5–6 November 2015; pp. 8–13.
108. González-Romera, E.; Jaramillo-Morán, M.; Carmona-Fernández, D. Monthly electric energy demand
forecasting with neural networks and Fourier series. Energy Convers. Manag. 2008, 49, 3135–3142. [CrossRef]
109. Cao, G.; Wu, L. Support vector regression with fruit fly optimization algorithm for seasonal electricity
consumption forecasting. Energy 2016, 115, 734–745. [CrossRef]
110. Li, C.; Li, S.; Liu, Y. A least squares support vector machine model optimized by moth-flame optimization
algorithm for annual power load forecasting. Appl. Intell. 2016, 45, 1166–1178. [CrossRef]
111. Koprinska, I.; Rana, M.; Troncoso, A.; Martínez-Álvarez, F. Combining pattern sequence similarity with
neural networks for forecasting electricity demand time series. In Proceedings of the 2013 International Joint
Conference on Neural Networks (IJCNN), Dallas, TX, USA, 4–9 August 2013; pp. 1–8.
112. Al-Musaylh, M.S.; Deo, R.C.; Li, Y.; Adamowski, J.F. Two-phase particle swarm optimized-support vector
regression hybrid model integrated with improved empirical mode decomposition with adaptive noise for
multiple-horizon electricity demand forecasting. Appl. Energy 2018, 217, 422–439. [CrossRef]
113. Puspitasari, I.; Akbar, M.S.; Lee, M.H. Two-level seasonal model based on hybrid ARIMA-ANFIS for
forecasting short-term electricity load in Indonesia. In Proceedings of the 2012 International Conference on
Statistics in Science, Business and Engineering (ICSSBE), Langkawi, Kedah, Malaysia, 10–12 September 2012;
pp. 1–5.
Entropy 2020, 22, 1412 21 of 24

114. Li, W.; Yang, X.; Li, H.; Su, L. Hybrid forecasting approach based on GRNN neural network and SVR
machine for electricity demand forecasting. Energies 2017, 10, 44. [CrossRef]
115. Benaouda, D.; Murtagh, F. Electricity load forecast using neural network trained from wavelet-transformed
data. In Proceedings of the 2006 IEEE International Conference on Engineering of Intelligent Systems,
Islamabad, Pakistan, 22–23 April 2006; pp. 1–6.
116. Saroha, S.; Aggarwal, S. Wind power forecasting using wavelet transforms and neural networks with tapped
delay. CSEE J. Power Energy Syst. 2018, 4, 197–209. [CrossRef]
117. Ni, K.; Wang, J.; Tang, G.; Wei, D. Research and Application of a Novel Hybrid Model Based on a Deep Neural
Network for Electricity Load Forecasting: A Case Study in Australia. Energies 2019, 12, 2467. [CrossRef]
118. Jadidi, A.; Menezes, R.; de Souza, N.; de Castro Lima, A.C. Short-Term Electric Power Demand Forecasting
Using NSGA II-ANFIS Model. Energies 2019, 12, 1891. [CrossRef]
119. Panapakidis, I.P.; Christoforidis, G.C.; Asimopoulos, N.; Dagoumas, A.S. Combining wavelet transform
and support vector regression model for day-ahead peak load forecasting in the Greek power system.
In Proceedings of the 2017 IEEE International Conference on Environment and Electrical Engineering
and 2017 IEEE Industrial and Commercial Power Systems Europe (EEEIC/I&CPS Europe), Milan, Italy,
6–9 June 2017; pp. 1–6.
120. Jiang, W.; Wu, X.; Gong, Y.; Yu, W.; Zhong, X. Holt–Winters smoothing enhanced by fruit fly optimization
algorithm to forecast monthly electricity consumption. Energy 2020, 193, 116779. [CrossRef]
121. Sulandari, W.; Lee, M.H.; Rodrigues, P.C. Indonesian electricity load forecasting using singular spectrum
analysis, fuzzy systems and neural networks. Energy 2020, 190, 116408. [CrossRef]
122. Laouafi, A.; Mordjaoui, M.; Medoued, A.; Boukelia, T.E.; Ganouche, A. Wind power forecasting approach
using neuro-fuzzy system combined with wavelet packet decomposition, data preprocessing, and forecast
combination framework. Wind Eng. 2017, 41, 235–244. [CrossRef]
123. Pao, H.T. Forecast of electricity consumption and economic growth in Taiwan by state space modeling.
Energy 2009, 34, 1779–1791. [CrossRef]
124. Vu, D.H.; Muttaqi, K.M.; Agalgaonkar, A. A variance inflation factor and backward elimination based robust
regression model for forecasting monthly electricity demand using climatic variables. Appl. Energy 2015,
140, 385–394. [CrossRef]
125. Tsekouras, G.; Dialynas, E.; Hatziargyriou, N.; Kavatza, S. A non-linear multivariable regression model for
midterm energy forecasting of power systems. Electr. Power Syst. Res. 2007, 77, 1560–1568. [CrossRef]
126. Clements, A.E.; Hurn, A.; Li, Z. Forecasting day-ahead electricity load using a multiple equation time series
approach. Eur. J. Oper. Res. 2016, 251, 522–530. [CrossRef]
127. Kim, M.S. Modeling special-day effects for forecasting intraday electricity demand. Eur. J. Oper. Res. 2013,
230, 170–180. [CrossRef]
128. Shang, H.L. Functional time series approach for forecasting very short-term electricity demand. J. Appl. Stat.
2013, 40, 152–168. [CrossRef]
129. Bessec, M.; Fouquau, J. Short-run electricity load forecasting with combinations of stationary wavelet
transforms. Eur. J. Oper. Res. 2018, 264, 149–164. [CrossRef]
130. Chui, F.; Elkamel, A.; Surit, R.; Croiset, E.; Douglas, P. Long-term electricity demand forecasting for power
system planning using economic, demographic and climatic variables. Eur. J. Ind. Eng. 2009, 3, 277–304.
[CrossRef]
131. Truong, N.V.; Wang, L.; Wong, P.K. Modelling and short-term forecasting of daily peak power demand in
Victoria using two-dimensional wavelet based SDP models. Int. J. Electr. Power Energy Syst. 2008, 30, 511–518.
[CrossRef]
132. Mestekemper, T.; Kauermann, G.; Smith, M.S. A comparison of periodic autoregressive and dynamic factor
models in intraday energy demand forecasting. Int. J. Forecast. 2013, 29, 1–12. [CrossRef]
133. Massidda, L.; Marrocu, M. Use of Multilinear Adaptive Regression Splines and numerical weather prediction
to forecast the power output of a PV plant in Borkum, Germany. Sol. Energy 2017, 146, 141–149. [CrossRef]
134. Duan, L.; Niu, D.; Gu, Z. Long and medium term power load forecasting with multi-level recursive
regression analysis. In Proceedings of the 2008 Second International Symposium on Intelligent Information
Technology Application, Washington, DC, USA, 20–22 December 2008; Volume 1, pp. 514–518.
Entropy 2020, 22, 1412 22 of 24

135. Guo, H.; Chen, Q.; Xia, Q.; Kang, C.; Zhang, X. A monthly electricity consumption forecasting method based
on vector error correction model and self-adaptive screening method. Int. J. Electr. Power Energy Syst. 2018,
95, 427–439. [CrossRef]
136. Elamin, N.; Fukushige, M. Modeling and forecasting hourly electricity demand by SARIMAX with
interactions. Energy 2018, 165, 257–268. [CrossRef]
137. Sigauke, C. Forecasting medium-term electricity demand in a South African electric power supply system.
J. Energy S. Afr. 2017, 28, 54–67. [CrossRef]
138. Tanrisever, F.; Derinkuyu, K.; Heeren, M. Forecasting electricity infeed for distribution system networks:
An analysis of the Dutch case. Energy 2013, 58, 247–257. [CrossRef]
139. Bernardi, M.; Petrella, L. Multiple seasonal cycles forecasting model: The Italian electricity demand.
Stat. Methods Appl. 2015, 24, 671–695. [CrossRef]
140. Angelopoulos, D.; Psarras, J.; Siskos, Y. Long-term electricity demand forecasting via ordinal regression
analysis: The case of Greece. In Proceedings of the 2017 IEEE Manchester PowerTech, Manchester, UK,
18–22 June 2017; pp. 1–6.
141. Bermúdez, J.D. Exponential smoothing with covariates applied to electricity demand forecast. Eur. J.
Ind. Eng. 2013, 7, 333–349. [CrossRef]
142. Sigauke, C.; Chikobvu, D. Peak electricity demand forecasting using time series regression models:
An application to South African data. J. Stat. Manag. Syst. 2016, 19, 567–586. [CrossRef]
143. Angelopoulos, D.; Siskos, Y.; Psarras, J. Disaggregating time series on multiple criteria for robust forecasting:
The case of long-term electricity demand in Greece. Eur. J. Oper. Res. 2019, 275, 252–265. [CrossRef]
144. Kumru, M.; Kumru, P.Y. Calendar-based short-term forecasting of daily average electricity demand.
In Proceedings of the 2015 International Conference on Industrial Engineering and Operations Management
(IEOM), Hyatt Regency, Dubai, 3–5 March 2015; pp. 1–5.
145. He, Y.; Zheng, Y.; Xu, Q. Forecasting energy consumption in Anhui province of China through two Box-Cox
transformation quantile regression probability density methods. Measurement 2019, 136, 579–593. [CrossRef]
146. İlseven, E.; Göl, M. Medium-term electricity demand forecasting based on MARS. In Proceedings of
the 2017 IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT-Europe), Torino, Italy,
26–29 September 2017; pp. 1–6.
147. Lebotsa, M.E.; Sigauke, C.; Bere, A.; Fildes, R.; Boylan, J.E. Short term electricity demand forecasting
using partially linear additive quantile regression with an application to the unit commitment problem.
Appl. Energy 2018, 222, 104–118. [CrossRef]
148. Cho, H.; Goude, Y.; Brossat, X.; Yao, Q. Modeling and forecasting daily electricity load curves: A hybrid
approach. J. Am. Stat. Assoc. 2013, 108, 7–21. [CrossRef]
149. Sigauke, C.; Chikobvu, D. Prediction of daily peak electricity demand in South Africa using volatility
forecasting models. Energy Econ. 2011, 33, 882–888. [CrossRef]
150. Li, S.; Yang, X.; Li, R. Forecasting China’s coal power installed capacity: A comparison of MGM, ARIMA,
GM-ARIMA, and NMGM models. Sustainability 2018, 10, 506. [CrossRef]
151. He, Y.; Zheng, Y. Short-term power load probability density forecasting based on Yeo-Johnson transformation
quantile regression and Gaussian kernel function. Energy 2018, 154, 143–156. [CrossRef]
152. Cheng, C.T.; Miao, S.M.; Luo, B.; Sun, Y.J. Forecasting monthly energy production of small hydropower
plants in ungauged basins using grey model and improved seasonal index. J. Hydroinform. 2017, 19, 993–1008.
[CrossRef]
153. Monteiro, C.; Ramirez-Rosado, I.J.; Fernandez-Jimenez, L.A. Short-term forecasting model for electric power
production of small-hydro power plants. Renew. Energy 2013, 50, 387–394. [CrossRef]
154. Moreno, J. Hydraulic plant generation forecasting in Colombian power market using ANFIS. Energy Econ.
2009, 31, 450–455. [CrossRef]
155. Perez-Mora, N.; Canals, V.; Martinez-Moll, V. Short-term Spanish aggregated solar energy forecast.
In International Work-Conference on Artificial Neural Networks; Springer: Berlin/Heidelberg, Germany, 2015;
pp. 307–319.
156. Lahouar, A.; Mejri, A.; Slama, J.B.H. Importance based selection method for day-ahead photovoltaic power
forecast using random forests. In Proceedings of the 2017 International Conference on Green Energy
Conversion Systems (GECS), Hammamet, Tunisia, 23–25 March 2017; pp. 1–7.
Entropy 2020, 22, 1412 23 of 24

157. Jung, Y.; Jung, J.; Kim, B.; Han, S. Long short-term memory recurrent neural network for modeling temporal
patterns in long-term power forecasting for solar PV facilities: Case study of South Korea. J. Clean. Prod.
2020, 250, 119476. [CrossRef]
158. Lahouar, A.; Slama, J.B.H. Hour-ahead wind power forecast based on random forests. Renew. Energy 2017,
109, 529–541. [CrossRef]
159. Rana, M.; Koprinska, I.; Troncoso, A. Forecasting hourly electricity load profile using neural networks.
In Proceedings of the 2014 International Joint Conference on Neural Networks (IJCNN), Beijing, China,
6–11 July 2014; pp. 824–831.
160. Lu, S.L. Integrating heuristic time series with modified grey forecasting for renewable energy in Taiwan.
Renew. Energy 2019, 133, 1436–1444. [CrossRef]
161. Colombo, T.; Koprinska, I.; Panella, M. Maximum Length Weighted Nearest Neighbor approach for electricity
load forecasting. In Proceedings of the 2015 International Joint Conference on Neural Networks (IJCNN),
Killarney, Ireland, 12–17 July 2015; pp. 1–8.
162. Wang, C.H.; Lin, K.P.; Lu, Y.M.; Wu, C.F. Deep Belief Network with Seasonal Decomposition for Solar Power
Output Forecasting. Int. J. Reliab. Qual. Saf. Eng. 2019, 26, 1950029. [CrossRef]
163. Lahouar, A. Gaussian Process Based Method for Point and Probabilistic Short-Term Wind Power Forecast.
In International Conference on the Sciences of Electronics, Technologies of Information and Telecommunications;
Springer: Berlin/Heidelberg, Germany, 2018; pp. 134–147.
164. da Silva Fonseca, J.G., Jr.; Oozeki, T.; Takashima, T.; Koshimizu, G.; Uchida, Y.; Ogimoto, K. Use of support
vector regression and numerically predicted cloudiness to forecast power output of a photovoltaic power
plant in Kitakyushu, Japan. Prog. Photovoltaics Res. Appl. 2012, 20, 874–882. [CrossRef]
165. Koprinska, I.; Rana, M.; Agelidis, V.G. Correlation and instance based feature selection for electricity load
forecasting. Knowl. Based Syst. 2015, 82, 29–40. [CrossRef]
166. Taylor, J.W. An evaluation of methods for very short-term load forecasting using minute-by-minute British
data. Int. J. Forecast. 2008, 24, 645–658. [CrossRef]
167. Taylor, J.W. Short-term load forecasting with exponentially weighted methods. IEEE Trans. Power Syst. 2012,
27, 458–464. [CrossRef]
168. Lotufo, A.; Minussi, C. Electric power systems load forecasting: A survey. In Proceedings of the PowerTech
Budapest 99. Abstract Records., (Cat. No. 99EX376), Budapest, Hungary, 29 August–2 September 1999; p. 36.
169. Shumway, R.H.; Stoffer, D.S. Time Series Analysis and Its Applications: With R Examples; Springer:
Berlin/Heidelberg, Germany, 2017.
170. Whittle, P. Hypothesis Testing in Time Series Analysis; Almqvist & Wiksells boktr.: Stockholm, Sweden, 1951;
Volume 4.
171. Box, G.E.; Jenkins, G.M.; Reinsel, G.C. Time Series Analysis: Forecasting and Control; John Wiley & Sons:
Hoboken, NJ, USA, 2011; Volume 734.
172. Makridakis, S.; Spiliotis, E.; Assimakopoulos, V. Statistical and Machine Learning forecasting methods:
Concerns and ways forward. PLoS ONE 2018, 13, e0194889. [CrossRef] [PubMed]
173. Murphy, K.P. Machine Learning: A Probabilistic Perspective; MIT press: Cambridge, MA, USA, 2012.
174. Tosh, C.R.; Ruxton, G.D. Modelling Perception with Artificial Neural Networks; Cambridge University Press:
Cambridge, UK, 2010.
175. Azadeh, A.; Ghaderi, S.; Sohrabkhani, S. A simulated-based neural network algorithm for forecasting
electrical energy consumption in Iran. Energy Policy 2008, 36, 2637–2644. [CrossRef]
176. Hamzaçebi, C.; Es, H.A.; Çakmak, R. Forecasting of Turkey’s monthly electricity demand by seasonal
artificial neural network. Neural Comput. Appl. 2017, 31, 2217–2231. [CrossRef]
177. Ma, J.; Oppong, A.; Acheampong, K.N.; Abruquah, L.A. Forecasting renewable energy consumption under
zero assumptions. Sustainability 2018, 10, 576. [CrossRef]
178. Li, G.; Wang, H.; Zhang, S.; Xin, J.; Liu, H. Recurrent neural networks based photovoltaic power forecasting
approach. Energies 2019, 12, 2538. [CrossRef]
179. Tzafestas, S.G. Introduction to Mobile Robot Control; Elsevier: Amsterdam, The Netherlands, 2013.
180. Efendi, R.; Ismail, Z.; Deris, M.M. A new linguistic out-sample approach of fuzzy time series for daily
forecasting of Malaysian electricity load demand. Appl. Soft Comput. 2015, 28, 422–430. [CrossRef]
Entropy 2020, 22, 1412 24 of 24

181. Marwala, L.; Twala, B. Forecasting electricity consumption in South Africa: ARMA, neural networks and
neuro-fuzzy systems. In Proceedings of the 2014 International Joint Conference on Neural Networks
(IJCNN), Beijing, China, 6–11 July 2014; pp. 3049–3055.
182. Rebala, G.; Ravi, A.; Churiwala, S. An Introduction to Machine Learning; Springer: Berlin/Heidelberg,
Germany, 2019.
183. Al-Musaylh, M.S.; Deo, R.C.; Adamowski, J.F.; Li, Y. Short-term electricity demand forecasting with MARS,
SVR and ARIMA models using aggregated demand data in Queensland, Australia. Adv. Eng. Inform. 2018,
35, 1–16. [CrossRef]
184. Lin, K.P.; Pai, P.F. Solar power output forecasting using evolutionary seasonal decomposition least-square
support vector regression. J. Clean. Prod. 2016, 134, 456–462. [CrossRef]
185. Sood, R.; Koprinska, I.; Agelidis, V.G. Electricity load forecasting based on autocorrelation analysis.
In Proceedings of the 2010 International Joint Conference on Neural Networks (IJCNN), Barcelona, Spain,
18–23 July 2010; pp. 1–8.
186. Mourad, M.; Bouzid, B.; Mohamed, B. A hybrid wavelet transform and ANFIS model for short term electric
load prediction. In Proceedings of the 2012 2nd International Conference on Advances in Computational
Tools for Engineering Applications (ACTEA), Beirut, Lebanon, 12–15 December 2012; pp. 292–295.
187. Sudheer, G.; Suseelatha, A. Pronóstico de carga a corto plazo usando la transformación wavelet combinada
con Holt—Winters y modelos vecinos más próximos ponderados. Int. J. Electr. Power Energy Syst. 2015.
188. Guan, C.; Luh, P.B.; Michel, L.D.; Wang, Y.; Friedland, P.B. Very short-term load forecasting: Wavelet neural
networks with data pre-filtering. IEEE Trans. Power Syst. 2012, 28, 30–41. [CrossRef]
189. Hochreiter, S.; Schmidhuber, J. Long short-term memory. Neural Comput. 1997, 9, 1735–1780. [CrossRef]
[PubMed]

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional
affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).