Conference PaperPDF Available

Abstract and Figures

A robust model for power system load forecasting covering different horizons of time from short-term to long-term is an indispensable tool to have a better management of the system. However, as the horizon of time in load forecasting increases, it will be more challenging to have an accurate forecast. Machine learning methods have got more attention as efficient methods in dealing with the stochastic load pattern and resulting in accurate forecasting. In this study, the problem of long-term load forecasting for the case study of New England Network is studied using several commonly used machine learning methods such as feedforward artificial neural network, support vector machine, recurrent neural network, generalized regression neural network, k-nearest neighbors, and Gaussian Process Regression. The results of these methods are compared with mean absolute percentage error (MAPE).
Content may be subject to copyright.
Long Term Forecasting using Machine Learning
Hossein Sangrody
, Student Member, IEEE, Ning Zhou
, Senior Member, IEEE, Salih Tutun
, Benyamin
, Student Member, IEEE, Mahdi Motalleb
Student Member, IEEE, Morteza Sarailoo
Electrical and Computer Engineering Department, State University of New York at Binghamton, NY, USA
{habdoll1, ningzhou, msarail1}}
Industrial Engineering Department, State University of New York at Binghamton, NY, USA,
Department of Electrical and Computer Engineering, University of Saskatchewan, Canada,
Department of Mechanical Engineering, Massachusetts Institute of Technology (MIT), MA, USA,
Abstract— A robust model for power system load forecasting
covering different horizons of time from short-term to long-term is
an indispensable tool to have a better management of the system.
However, as the horizon of time in load forecasting increases, it will
be more challenging to have an accurate forecast. Machine
learning methods have got more attention as efficient methods in
dealing with the stochastic load pattern and resulting in accurate
forecasting. In this study, the problem of long-term load
forecasting for the case study of New England Network is studied
using several commonly used machine learning methods such as
feedforward artificial neural network, support vector machine,
recurrent neural network, generalized regression neural network,
k-nearest neighbors, and Gaussian Process Regression. The results
of these methods are compared with mean absolute percentage
error (MAPE).
Distributed energy resources (DERs) are penetrating in
power system more and more and a lot of studies have focused
on implementation of them [1-3]. Although DERs have benefits
for a power system, their indispatchability, intermittency, and
uncertainty have presented unprecedented challenges to power
grid operation and planning [4]. At this condition, load
forecasting (LF) plays a critical role in the operation and
planning of a power system. Depending on the purposes of LF,
the lead times of LF can vary from seconds to years. Very short-
term load forecasting (VSTLF) [5] and short-term load
forecasting (STLF) [6] usually have lead times of seconds to
weeks and are often used for control and operation purposes. In
contrast, medium-term load forecasting (MTLF) [7] and long-
term load forecasting (LTLF) [8] have lead times of month(s),
years, even decades and are often used for scheduling and
planning purposes [9].
The driving inputs of a forecasting model are important factors
to yield an efficient forecast model. The inputs of forecasting
model depend on the purpose of forecasting and its term (very
short to long term). In [10] historical data of price are applied to
predict the hourly prices in the California Independent System
Operator (CAISO)’s day-ahead electricity market. Along with
historical data of load in load forecasting model, temperature is
also one of the most common input variables. However, since
load pattern is a nonlinear function of temperature, heating
degree days (HDD) and cooling degree days (CDD) are applied
as weather indicators in load forecasting modeling [11].
The methodologies applied in LF can be classified in three main
categories of statistical analysis, machine learning, and hybrid
methods. Among them, machine learning methods have got
more attentions in recent years [12]. Despite of the benefits of
hybrid methods, their parameters need to be adjusted well to
achieve accurate forecasting [7]. Among all machine learning
methods, feedforward artificial neural network (ANN), support
vector regression (SVR), recurrent neural network (RNN), k-
nearest neighbors (KNN), generalized regression neural
network, and Gaussian Process Regression (GPR) are the most
common methods in load forecasting. In this study, the LTLF
for monthly load forecasting in ISO New England Network is
applied using the aforementioned machine learning methods
and their results are quantified and compared by mean absolute
percentage error (MAPE).
The rest of the paper is organized as follows. Section II
elaborates the forecasting inputs, output, the methodology of
applying them, and forecasting models. Section III represents
the simulation results for the case study using different methods.
The conclusion is drawn in Section V.
Obtaining accurate forecast results depend on various factors.
Generally, the horizon of LF, certainty of the inputs, and
efficiency of forecasting models are the major influential factors
on the accuracy of forecasting results. As the horizon of load
forecast increases, having accurate prediction with highly time
resolution will be more challenging. The main reason is the high
uncertainty in the inputs of forecasting model in a long term
forecasting. Accordingly, in VSTL and STLF, since the horizon
of forecasting spans only in the time frame of seconds to weeks,
the weather indicators are more accurate inputs while for the
MTLF and LTLF which cover the lead time of months to years,
the prediction relying on inaccurate weather indicators.
A. Input and Output of the Models
As mentioned the most common variables in LF are weather
indicators. However, the relationship of weather indicators and
energy usage is not linear. Thus, two other weather indicators,
HDD and CDD are applied in forecasting model. HDD is a
criterion showing whether a unit requires to be heated and it is
obtained by the number of degrees which average temperature
of the day is below 65 F. On the other hand, CDD showing that
a unit needs to be colder is the number of degrees that the
average temperature of a day is above 65 F. Such variables
yield linear relationships between energy and weather
indicators. In this study, the forecasting model is applied to
predict monthly energy for the New England Network and the
total HDDs and CDDs for each month are applied as inputs of
the model.
To improve the accuracy of forecasting model, the historical
record of energy is also used as input using moving average
method. As shown in (1), to have prediction of energy in a target
month, an 11-month average energy corresponding to the target
month is also fed to the model along with weather indicators.
Here is historical data time which is 11 for this case and
represents target variable.
Fig. 1 illustrates the overall structure of inputs and outputs in
the forecasting model. In this figure, for each month of
forecasted energy as target, the model uses corresponding total
HDD, CDD, and average energy of 11 corresponding months of
the past years.
Fig. 1. Forecasting structure
B. Forecasting Models
In this sections, a brief review on the commonly used machine
learning methods in forecasting are discussed.
1) Feed forward Artificial Neural Network Model
The ANN method provides an efficient way to address
modelling of a complex nonlinear system. In other words, in
forecasting model using ANN, there is no need for a forecaster
to have a clear understanding of the complex relationship
between inputs and outputs.
Fig. 2 depicts a typical neural network normally consists of
three layers of input, hidden and output layers. Each layer
consists of several neurons which are connected to other layer’s
neuron(s) with weighted connections. As shown in this figure,
the arrowheads of the connections indicate that all data
propagate in the direction from inputs to the output. Such a
structure is entitled feedforward ANN model. The number of
neurons in the input and output layers are the number of inputs
and output, respectively. The hidden layer located between the
input layer and output layer has an arbitrary number of neurons
which are defined by the forecaster. For many problems in
forecasting, one or two hidden layers often give good results.
Fig. 2 Typical structure of feedforward ANN model
The training algorithm for this study is the Levenberg-
Marquardt algorithm which takes more training time but gives
better results.
2) Support Vector Machine
Support vector regression is the version of the support vector
machine method that are applying for forecasting model.
Assuming as the input variable vector and as the output
variables, the SVR solution can be obtained by minimizing the
sum of training error (
and regularization term
in (2) subjected to constraints (3).
0,=1,, (3)
Here, N is the total number of observation sets,
upper and lower training errors associated to (margin of
tolerance) and is the kernel function, which transforms
higher dimensional space.
3) Recurrent neural network
The structure of Recurrent neural network(RNN) includes an
input layer, hidden layer, a context layer, and output layer [13,
14]. (4) and (5) show the calculations in RNN model for
training data points
and a target values
) (4)
) (5)
denotes the hidden layer vector, is bias vector,
and are weigh matrices, and represents the activation
4) Generalized Regression Neural Network
Generalized regression neural network (GRNN) is
nonparametric model whose structure includes input layer,
output layer, radial basis layer, and a special linear layer. The
GRNN model derives prediction of a target value corresponding
to a given data point by calculating the weighted average of
target values in the training data points in the vicinity of the data
point [15]. As shown in (6), a target point () corresponding
to the data point is predicted by the average of the target points
and assigning weights using a kernel function considering the
distance of predictors in training set to the data point . In this
case, the kernel function () is a standard Gaussian kernel.
5) K nearest neighbors Regression
K nearest neighbors method (KNN) is a nonparametric
method applied for both regression and classification. In this
method, the prediction is yielded based on the target values of
the K nearest neighbors in the given point. In other words, given
a data point, the K nearest data points in the training data set are
selected and the average of their target values are considered as
predicted target value as follows.
6) Gaussian Process Regression
Gaussian Process Regression (GPR) is a nonparametric
method which is modeled based on considering a priori
distribution (multivariate joint Gaussian distribution) for any
subset of target values of different data points in training set
[16]. In GPR, if two input vectors are close, the correlation
between their function value is higher. The posterior
distribution for a predicted value is derived using the prior
distribution. The covariance two inputs
in training
inputs can be model as follows [15].
So, the vector function (consisting function values
training point ) follows multivariate Gaussian density function
as follows.
(0,(,)) (10)
denotes multivariate normal density function and
(,) denotes covariance matrix whose (,)
element is
. Considering the target values as (11), the predicted
posterior function for an input value
is derived as (12).
=+ (11)
Where denotes the vector of noise with standard deviation
C. Accuracy metric
To quantify the results of forecasting models, there are
various statics metrics. Some common metrics are mean
absolute percentage error (MAPE) [17] defined by (13), mean
absolute error (MAE) [18] defined by (14), mean squared error
(MSE) [19] defined by (15), and root-mean-square error
(RMSE) [20] defined by (16).
×100 (13)
Where, is the total number of time instants,
is the target
value at instance of and
is the corresponding forecasted
Between aforementioned error metrics, MAPE is the most
common metrics. As shown in (13) the MAPE gives relative
errors in percentage, which does not depend on the scale of
forecasted variables. Therefore, the MAPE has been widely
used to compare forecasting accuracy under different scenarios.
Accordingly, in this study, the MAPE is used as the criterion to
measure the proficiency of the models.
In this section, the long-term load forecasting models are
applied for the case study of the New England Network. The
load profile of the network during 2000 to March 2016 is shown
in Fig. 3.
Fig. 3. New England Network Energy during 2000 to March 2016
As mentioned before, the input variables of each load forecast
model depend on the case study and horizon of forecasting. In
long term forecasting, one of the likely inputs is population
growing rate which has a positive exponential rate. However,
Fig. 3 illustrates that the load did not grow exponentially during
long term of 17 years which dismisses the influence of growing
population as a input for this case.
In Fig. 4, the energy usage is represented in month and year
axes. As shown the load in July and August peaks dramatically.
Such a stochastic load behavior will make it more difficult for
the forecasting model to have accurate prediction if the model
is supposed to rely on only weather variables. Note that although
residential load pattern depends on whether indicators like
temperature, industrial and commercial loads do not correlate
strongly on weather temperature. However, in this case since the
dramatic changes of load in the aforementioned months as well
as January are repeated at the same time during all years, using
historical value of load data can be a good solution to deal with
such a problem. In addition, considering the number of month
as an input is one of the inputs which improves the results of
prediction. Accordingly, each target value in a load forecasting
model corresponds to weather indicators, historical average, and
dummy variable of the month number.
The monthly energy data are divided in three categories of
historical data (which is applied in forecasting model as one of
the inputs), training set, and cross validation data set. As shown
in Fig. 5, the green color are historical data from 2000 to 2011
and sixty percent of the rest of data which is the monthly energy
during January 2011 to March 2016 are used for training of
(blue color) the forecast model and 40 percent are applied for
validation (red color).
The data before 2011 are implemented as one of inputs
corresponding to each target value. In other words, each month
of target gets benefit from the corresponding monthly historical
data of during 11 years ago and the historical data moves ahead
as the target value of monthly energy moves ahead in training
and validation processes.
Fig. 4. 2D representation of energy usage for New England Network
As an example, to forecast energy in January 2011, as the first
target, the inputs are total HDD and CDD of January 2011,
dummy variable for the month number which is one for January
and zero for other months, and historical monthly energy of the
January during 2000 to 2010. For the next slide, for forecasting
the energy in February 2011, the energy usage in February 2000
to 2010, the month variable (which is 1 for February), as well as
HDD and CDD of February 2011 are applied in the model.
Fig. 5. Training and validation set of data
As mentioned before, the most common machine learning
methods in load forecasting are ANN, SVR, RNN, GRNN,
KNN, and GPR. All LF models are implemented using
and the results for both training and validation set
are compared with MAPE metric.
For the feedforward ANN, by trying with different hidden
layers, one hidden layer and 3 neurons results in low forecasting
errors for both training and validation data sets. The SVR
method is applied for LF using LIBSVM [21].
Table I represents the results of the 6 forecasting models. In
this table, the results are resented in MAPEs for both training
and validation data sets.
As seen in the table, although the results of LF for all methods
are close to each other, the feedforward ANN represents better
results than other methods for the validation set while it also has
decent result for the training set. Note that the MAPE for
training set in the KNN method is zero since the same data for
model training is applied in testing of training data set.
Method Training Validation
ANN 0.7 1.5
SVR 0.9 1.7
RNN 0.7 1.9
GRNN 0.7 2.3
KNN 0.0 2.3
GPR 0.6 2.0
The result of ANN is also represented in Fig. 6. In this figure,
the green color graph exhibits the actual energy and the blue and
red ones demonstrate the training and validation set,
respectively. Note that since the ANN method, random process
is applied, each running of the simulation may result in slightly
different values. In this regard, the average of several simulation
running is considered as ANN’s final result.
Fig. 6. LF using ANN model
In this study, the performance of the most commonly used
machine learning methods in load forecasting has been studied.
These methods are feedforward artificial neural network
(ANN), support vector regression (SVR), recurrent neural
network (RNN), k-nearest neighbors (KNN), generalized
regression neural network (GRNN), and Gaussian Process
Regression (GPR). The case study is New England Network and
its monthly energy usage during 2000 to May 2016 is
considered for training and validation of the load forecasting
models. The inputs of the load forecasting models are weather
indicators (HDD and CDD), dummy variable of month number,
and the moving average of the target variable before 2011. The
results of forecasting models which are represented by MAPE
indicate that although for both training and validation data set,
all LF methods depict proficient performance, the feedforward
ANN method shows better results than the other forecasting
[1] H. Sadeghian and Z. Wang, "Decentralized Demand Side
Management with Rooftop PV in Residential Distribution Network,"
in Innovative Smart Grid Technologies Conference (ISGT), 2018
IEEE Power & Energy Society, 2017, pp. 1-5.
[2] M. Ghorbaniparvar, X. Li, and N. Zhou, "Demand side management
with a human behavior model for energy cost optimization in smart
grids," in Signal and Information Processing (GlobalSIP), 2015
IEEE Global Conference on, 2015, pp. 503-507.
[3] M. Motalleb, A. Eshraghi, E. Reihani, H. Sangrody, and R.
Ghorbani, "A Game-Theoretic Demand Response Market with
Networked Competition Model," in 49th North American Power
Symposium (NAPS), 2017.
[4] M. H. Athari and Z. Wang, "Impacts of Wind Power Uncertainty on
Grid Vulnerability to Cascading Overload Failures," IEEE
Transactions on Sustainable Energy, vol. 9, pp. 128-137, 2018.
[5] J. W. Taylor, "An evaluation of methods for very short-term load
forecasting using minute-by-minute British data," International
Journal of Forecasting, vol. 24, pp. 645-658, 2008.
[6] T. Hong, Short term electric load forecasting: North Carolina State
University, 2010.
[7] N. Amjady and A. Daraeepour, "Midterm demand prediction of
electrical power systems using a new hybrid forecast technique,"
Power Systems, IEEE Transactions on, vol. 26, pp. 755-765, 2011.
[8] R. J. Hyndman and S. Fan, "Density forecasting for long-term peak
electricity demand," Power Systems, IEEE Transactions on, vol. 25,
pp. 1142-1153, 2010.
[9] H. Sangrody and N. Zhou, "An initial study on load forecasting
considering economic factors," in 2016 IEEE Power and Energy
Society General Meeting (PESGM), Boston, MA, 2016, pp. 1-5.
[10] A. Sadeghi-Mobarakeh, M. Kohansal, E. E. Papalexakis, and H.
Mohsenian-Rad, "Data mining based on random forest model to
predict the California ISO day-ahead market prices," in Power &
Energy Society Innovative Smart Grid Technologies Conference
(ISGT), 2017 IEEE, 2017, pp. 1-5.
[11] C.-L. Hor, S. J. Watson, and S. Majithia, "Analyzing the impact of
weather variables on monthly electricity demand," Power Systems,
IEEE Transactions on, vol. 20, pp. 2078-2085, 2005.
[12] E. Foruzan, S. D. Scott, and J. Lin, "A comparative study of different
machine learning methods for electricity prices forecasting of an
electricity market," in North American Power Symposium (NAPS),
2015, 2015, pp. 1-6.
[13] H. Liu, X.-w. Mi, and Y.-f. Li, "Wind speed forecasting method
based on deep learning strategy using empirical wavelet transform,
long short term memory neural network and Elman neural network,"
Energy Conversion and Management, vol. 156, pp. 498-514, 2018.
[14] X. Chen, X. Chen, J. She, and M. Wu, "A hybrid time series
prediction model based on recurrent neural network and double joint
linear–nonlinear extreme learning network for prediction of carbon
efficiency in iron ore sintering process," Neurocomputing, vol. 249,
pp. 128-139, 2017.
[15] N. K. Ahmed, A. F. Atiya, N. E. Gayar, and H. El-Shishin y, "An
empirical comparison of machine learning models for time seri es
forecasting," Econometric Reviews, vol. 29, pp. 594-621, 2010.
[16] C. E. Rasmussen and C. K. Williams, Gaussian processes for
machine learning vol. 1: MIT press Cambridge, 2006.
[17] J. W. Taylor and R. Buizza, "Neural network load forecasting with
weather ensemble predictions," Power Systems, IEEE Transactions
on, vol. 17, pp. 626-632, 2002.
[18] H. Shih-Che, L. Chan-Nan, and L. Yuan-Liang, "Evaluation of AMI
and SCADA Data Synergy for Distribution Feeder Modeling,"
Smart Grid, IEEE Transactions on, vol. 6, pp. 1639-1647, 2015.
[19] K. Siwek, S. Osowski, and R. Szupiluk, "Ensemble neural network
approach for accurate load forecasting in a power system,"
International Journal of Applied Mathematics and Computer
Science, vol. 19, pp. 303-315, 2009.
[20] T. Chen and Y.-C. Wang, "Long-term load forecasting by a
collaborative fuzzy-neural approach," International Journal of
Electrical Power & Energy Systems, vol. 43, pp. 454-464, 2012.
[21] C.-C. Chang and C.-J. Lin, "LIBSVM: A library for support vector
machines," ACM Transactions on Intelligent Systems and
Technology (TIST), vol. 2, p. 27, 2011.
2010 2011 2012 2013 2014 2015 2016 2017
1.3 104
Actual Data
Training Data
Validatio n Data
... He also proposed an FID with possibility functions. To calculate the accident risk of a diaphragm wall collapse, Lin et al. [57] used IDs and FST. FIDs are more able to handle ambiguity in risk analysis than IDs are. ...
... The model for short-term electricity demand in Albania, based on Fuzzy Logic, where the time, the previous day load, and the historical value of the temperature were observed as an independent parameter, gave very good prediction results [56]. The adaptive neuro-fuzzy inference system (ANFIS) model was used for short-term load forecasting in South Africa, but it was concluded that observing temperature as a separate input parameter did not improve the accuracy of the forecast [57]. Authors in [58] also deal with fuzzy logic for load forecasting. ...
... An LSTM-RNN model is also suggested for this purpose by Bouktif et al. [56]. Six frequently used machine learning (ML) methods were examined by Sangrody et al. [57]: ANN, SVM, RNN, KNN, GPR, and generalized regression neural network (GRNN). For LTLF, ANN performed better than the other five approaches. ...
An effective infrastructure for vehicle charging is necessary given the rising popularity of electrical cars. The standard method for determining the best location for a charging station involves minimizing travel expenses while keeping limits on the capacity of the public electrical network.
... Many studies are related to forecasting waste generation rates but vary widely regarding in-process models and data types. Regression analysis, material flow models, time series analysis, and machine learning (ML) are so far adopted methods for waste generation forecasting (Abdallah et al., 2020;Chhay et al., 2018;Sangrody et al., 2018). The most prominent ML methods for waste generation forecasting are support vector machine (SVM), artificial neural network (ANN), and recurrent neural network (RNN) (Karevan & Suykens, 2020). ...
... Abbasi and El Hanandeh (2016) compared the accuracy of time series and ANN approaches with additional ML procedures regarding their ability to forecast scheduled (for every 30 days) waste generation in Logan City, Australia. Sangrody et al. (2018) established an ANN method that integrates socioeconomic variables to forecast yearly MSW generation and rates at the municipal level. ...
Full-text available
Projecting municipal solid waste generation and identifying socioeconomic factors affecting waste generation is crucial for integrated waste management strategies. The present research work focuses on the projection of municipal solid waste (MSW) generation in Prayagraj, India, based on demographics and socioeconomic factors, using long short-term memory (LSTM), autoregressive integrated moving average (ARIMA), and incremental increase models (IIM). The model was integrated with nine socioeconomic variables to improve accuracy. The influence of socioeconomic variables on MSW generation was evaluated using correlation and fuzzy logic approaches. Waste generation data collected from the Central Pollution Control Board (CPCB) from 1997 to 2015 were used to train the models. The results of the correlation study indicate that population growth, employment, and households have a substantial impact on waste generation rates. Root mean squared error (RMSE), mean absolute percent error (MAPE), and coefficient of determination (R²) suggest that LSTM is the best model to forecast MSW generation in Prayagraj, India. The R² value indicates that the LSTM is more accurate (0.92) than ARIMA (0.72) and IIM (0.70). LSTM projection indicates that the city will have a population of 1.6 million by 2031, and waste generation will increase by 70.6% in 2031. Graphical Abstract
... However, managing the flexibility requirements of interconnected power systems by network reinforcement is often encountered with the problem of data privacy preservation between neighboring utilities. Thus, an equivalent economic dispatch model that considers hidden flexibility information alongside other power systems' constraints is proposed in [109]. The equivalent flexibility model for network enhancement involves a multiperiod economic dispatch problem involving interconnected power networks using mixed-integer linear programming (MILP) approach. ...
Due to the complicated load and supply balance dynamics, the massive amounts of renewable energy being introduced into the energy mix pose significant challenges for utilities and their customers. The renewable energy generators’ outputs are intermittent and thus create an imbalance between the instantaneous load demand and available supply at different instances of time. Besides, the inertia in power systems is becoming lesser due to the displacement of the rotating mass of conventional generators with inverter-based generators. Thus, the challenge of meeting the flexibility needs of modern power systems is becoming significantly high in recent times. Because of this, the traditional methods of meeting the flexibility needs of power systems are becoming insufficient; this calls for developing new intelligent approaches that can handle complex situations. Different concepts of artificial intelligence (AI) are deployed as a solution provider to numerous complex power systems operational problems, especially in resource forecasting, electricity market dynamics prediction, intelligent decision-making for generator scheduling, and more. Hence, this book chapter reviews existing flexibility management techniques and some crucial areas of AI deployment in energy management systems toward meeting the flexibility needs of modern energy supply systems.KeywordsFlexibility managementArtificial intelligence (AI)Demand-side management (DSM)Variable renewable energy resources (VREs)Energy storage systems (ESSs)Pumped hydro storage systems (PHESSs)Battery energy storage systems (BESSs)Grid-to-vehicle (G2V) and vehicle-to-grid (V2G) technologiesPeer-to-peer energy (P2P) tradingEnergy management systemDeep learningmachine learningArtificial neural networkIntelligent decision-makingResource forecastDemand responseDynamic electricity marketSystem planningGenerator scheduling
Full-text available
This comprehensive review paper aims to provide an in-depth analysis of the most recent developments in the applications of artificial intelligence (AI) techniques, with an emphasis on their critical role in the demand side of power distribution systems. This paper offers a meticulous examination of various AI models and a pragmatic guide to aid in selecting the suitable techniques for three areas: load forecasting, anomaly detection, and demand response in real-world applications. In the realm of load forecasting, the paper presents a thorough guide for choosing the most fitting machine learning and deep learning models, inclusive of reinforcement learning, in conjunction with the application of hybrid models and learning optimization strategies. This selection process is informed by the properties of load data and the specific scenarios that necessitate forecasting. Concerning anomaly detection, this paper provides an overview of the merits and limitations of disparate learning methods, fostering a discussion on the optimization strategies that can be harnessed to navigate the issue of imbalanced data, a prevalent concern in power system anomaly detection. As for demand response, we delve into the utilization of AI techniques, examining both incentive-based and price-based demand response schemes. We take into account various control targets, input sources, and applications that pertain to their use and effectiveness. In conclusion, this review paper is structured to offer useful insights into the selection and design of AI techniques focusing on the demand-side applications of future energy systems. It provides guidance and future directions for the development of sustainable energy systems, aiming to serve as a cornerstone for ongoing research within this swiftly evolving field.
Full-text available
The Long Short-Term Memory (LSTM) neural network model is an effective deep learning approach for predicting streamflow, and the investigation of the interpretability of deep learning models in streamflow prediction is of great significance for model transfer and improvement. In this study, four key hydrological stations in the Xijiang River Basin (XJB) in South China are taken as examples, and the performance of the LSTM model and its variant models in runoff prediction were evaluated under the same foresight period, and the impacts of different foresight periods on the prediction results were investigated based on the SHapley Additive exPlanations (SHAP) method to explore the interpretability of the LSTM model in runoff prediction. The results showed that (1) LSTM was the optimal model among the four models in the XJB; (2) the predicted results of the LSTM model decreased with the increase in foresight period, with the Nash–Sutcliffe efficiency coefficient (NSE) decreasing by 4.7% when the foresight period increased from one month to two months, and decreasing by 3.9% when the foresight period increased from two months to three months; (3) historical runoff had the greatest impact on streamflow prediction, followed by precipitation, evaporation, and the North Pacific Index (NPI); except evaporation, all the others were positively correlated. The results can provide a reference for monthly runoff prediction in the XJB.
Conference Paper
Short-term utility grid planning and operation are based on projected future energy consumption and transmission/generation capacity. The industrial sector being a major influence on the total electrical power consumption. Despite its importance, industrial load forecasting (LF) is not a widely discussed topic in literature, despite the fact that it is influenced by a number of factors, such as planned operations and work shifts, that are unusual or unnecessary in conventional load forecasting models. Unlike time-series analysis, machine learning (ML) techniques are well-suited to the nature of the loads since they can simulate complex nonlinear relationships through a process that uses previous data patterns. In this paper, four data-driven machine learning strategies including Support Vector Machine (SVM), Decision Tree (DT), Artificial Neural Network (ANN), and Gaussian Process (GP), are compared to forecast an industrial load. These strategies can help in selecting the optimal model for load prediction. The results show that GP outperformed other ML techniques, thus yielding more accurate and reliable readings with lower estimation errors for electrical LF, but at the expense of higher computational time.
Machine learning is becoming a fundamental tool in current energy systems. It helps to obtain accurate predictions of the variable renewable energy (VRE) generation, energy demand, or possible network outages, conferring to power system operators the possibility to make the needed actions to balance load and generation in intraday and day-ahead scheduling with benefits for operational costs, environmental impact, and system reliability. If the short-term forecasting is fundamental for operational aspects, the long-term forecasting is crucial for system planning. Moreover, an improvement in the granularity of the forecasting can help to go towards real-time solutions. The machine learning can be employed effectively in all these contexts. This chapter deals with a critical analysis of machine learning methods and techniques for renewable-based energy systems showing the advantages of introducing these novel approaches in future energetic scenarios through discussing some relevant case studies.KeywordsAncillary servicesAnomaly detectionArtificial intelligenceDeep learningEnergy demandEnergy sourcesFault predictionForecastingLoad forecastingMachine learningNeural networksPower outage forecastingPhotovoltaic forecastingPredictionRenewable energy generation forecastingRenewable energy sourcesRenewable energy systemWind forecasting
Full-text available
Developing an accurate and robust prediction of long-term average global solar irradiation plays a crucial role in industries such as renewable energy, agribusiness, and hydrology. However, forecasting solar radiation with a high level of precision is historically challenging due to the nature of this source of energy. Challenges may be due to location constraints, stochastic atmospheric parameters, and discrete sequential data. This paper reports on a new hybrid deep residual learning and gated long short-term memory recurrent network boosted by a differential covariance matrix adaptation evolution strategy (ADCMA) to forecast solar radiation one hour-ahead. The efficiency of the proposed hybrid model was enriched using an adaptive multivariate empirical mode decomposition (MEMD) algorithm and 1+1EA-Nelder–Mead simplex search algorithm. To compare the performance of the hybrid model to previous models, a comprehensive comparative deep learning framework was developed consisting of five modern machine learning algorithms, three stacked recurrent neural networks, 13 hybrid convolutional (CNN) recurrent deep learning models, and five evolutionary CNN recurrent models. The developed forecasting model was trained and validated using real meteorological and Shortwave Radiation (SRAD1) data from an installed offshore buoy station located in Lake Michigan, Chicago, United States, supported by the National Data Buoy Centre (NDBC). As a part of pre-processing, we applied an autoencoder to detect the outliers in improving the accuracy of solar radiation prediction. The experimental results demonstrate that, firstly, the hybrid deep residual learning model performed best compared with other machine learning and hybrid deep learning methods. Secondly, a cooperative architecture of gated recurrent units (GRU) and long short-term memory (LSTM) recurrent models can enhance the performance of Xception and ResNet. Finally, using an effective evolutionary hyper-parameters tuner (ADCMA) reinforces the prediction accuracy of solar radiation.
Conference Paper
This study examines the application of machine learning models to week-ahead load forecasting for power system short-term planning applications. Electrical energy consumption has increased significantly over the last two decades because of rising living standards, economic development, and population growth. For power system planners and operators, short-term load forecasting is necessary to make economic, safe, reliable, and operational planning decisions and to account for variations in energy demand based on the seasonal period. These models are essential for the operation of utilities to meet system short-term adequacy assessment studies. With uncertainties in national, regional, and local economic growth, along with seasonal variability in energy consumption patterns and conservation habits of electricity consumers, utilities may face new challenges in forecasting short-term load demand. This work considers five machine learning models for week-ahead hourly load forecasting based on historical actual load data and expected meteorological features such as humidity, pressure, cloud cover, temperature, and wind speed. The performances of the models are evaluated using real-world data provided on the IEEE DataPort for Electricity Demand Forecasting. The accuracies are assessed using the root mean square error (RMSE) and coefficient of determination (R 2 ) metrics. The results show that the Nonlinear autoregressive neural network with exogenous input (NARX) model performs well in predicting week-ahead load demand with an R 2 value of 0.99 on the test set and can accommodate seasonal trends in energy consumption.
Full-text available
In the past extensive researches have been conducted on demand side management (DSM) program which aims at reducing peak loads and saving electricity cost. In this paper, we propose a framework to study decentralized household demand side management in a residential distribution network which consists of multiple smart homes with schedulable electrical appliances and some rooftop photovoltaic generation units. Each smart home makes individual appliance scheduling to optimize the electric energy cost according to the day-ahead forecast of electricity prices and its willingness for convenience sacrifice. Using the developed simulation model, we examine the performance of decentralized household DSM and study their impacts on the distribution network operation and renewable integration, in terms of utilization efficiency of rooftop PV generation, overall voltage deviation, real power loss, and possible reverse power flows.
Conference Paper
Full-text available
In classical competition models, each firm decides on the amount of production to maximize its own profit and only one market is shared among all firms. In the case with more than one market which has been proposed in recent years, Networked Cournot Competition (NCC) models the relation between firms and markets. This paper describes a model of competition between demand response aggregators (as firms) which sell aggregated energy stored in residential storage devices (as a homogeneous good) in a networked environment with a market maker. This game is called Networked Stackelberg Competition (NSC). In this paper, for each firm, the optimal bidding plan and Nash equilibrium are obtained in an incomplete information game. Demand response aggregators submit their bids and the market maker (system operator) controls the transaction powers and price subject to grid's constraints and policies. Also, the effect of price-based demand scheduling has been studied on the firms' payoffs in a real world case study.
Full-text available
Higher penetration of renewables like wind power generation will introduce an unprecedented amount of uncertainty into the grid that might severely affect the grid vulnerability to cascading failures. In this study, we propose a mixed OPF-stochastic approach to analyze and simulate cascading failures in power grid and to evaluate the impact of wind generation in terms of its penetration and uncertainty level. The proposed approach incorporates both thermal stability model for transmission line outage and automatic power balance algorithms. Numerical simulation results on the IEEE 300 bus system indicate that uncertainty coming from wind energy has severe impacts on grid vulnerability to cascading overload failures under different contingency scenarios. Results also suggest that higher penetration levels of wind energy, if not managed appropriately, will add to this severity because higher uncertainties may be injected into weaker lines in a grid.
Conference Paper
Full-text available
Generally, it is difficult to accurately forecast electricity prices because they are unpredictable. Yet, accurate price forecasting is expected to provide crucial information, needed by power producers and consumers to bid strategically, thereby decreasing their risks and increasing their profits in the electricity market. In this paper, two models using artificial neural networks (ANN) and support vector machines (SVM) were developed for electricity price forecasting. In addition, ant colony optimization (ACO) was used to reduce the feature space and give the best attribute subset for ANN model. Using ACO for feature selection significantly reduced the training time for ANN-based electricity price forecasting model while the results were almost as accurate as those from ANN model.
The wind speed forecasting plays an important role in the planning, controlling and monitoring of the intelligent wind power systems. Since the wind speed signal is stochastic and intermittent, it is difficult to achieve their satisfactory prediction. In the study, a novel hybrid deep-learning wind speed prediction model, which combines the empirical wavelet transformation and two kinds of recurrent neural network, is proposed. In the proposed new model, the empirical wavelet transformation is adopted to decompose the raw wind speed data into several sub-layers. The long short term memory neural network, a deep learning algorithm based method, is utilized to predict the low-frequency wind speed sub-layers. The Elman neural network, a mainstream recurrent neural network, is built to predict the high-frequency sub-layers. In the executed forecasting experiments, eleven different forecasting models are included to validate the real prediction performance of the proposed model. The experimental results indicate that the proposed model has satisfactory performance in the high-precision wind speed prediction.
Conference Paper
In this paper, an ensemble learning model, namely the random forest (RF) model, is used to predict both the exact values as well as the class labels of 24 hourly prices in the California Independent System Operator (CAISO)'s day-ahead electricity market. The focus is on predicting the prices for the Pacific Gas and Company (PG&E) default load aggregation point (DLAP). Several effective features, such as the historical hourly prices at different locations, calender data, and new ancillary service requirements are engineered and the model is trained in order to capture the best relations between the features and the target electricity price variables. Insightful case studies are implemented on the CAISO market data from January 1, 2014 to February 28, 2016. It is observed that the proposed data mining approach provides promising results in both predicting the exact value and in classifying the prices as low, medium and high.
Iron ore sintering process is the second-most energy-consuming procedure in the iron making industry. The main energy for it is the combustion of coke, which consists primary of carbon. In order to improve the carbon efficiency, it is necessary to predict it. A comprehensive carbon ratio (CCR) was used to be the metric for estimating the carbon efficiency. An iron ore sintering process has the characteristics of autocorrelation of time series of CCR, multiple variables, linearity and nonlinearity, and time delay. In this study, a hybrid time series prediction model was built to predict the CCR based on these characteristics. It consists of two parts: time series prediction based on Elman recurrent neural network (RNN) and Elman-residuals prediction based on double joint linear-nonlinear extreme learning network (JLNELN). The Elman RNN with a context layer has the ability to model the dynamical and nonlinear components in the time series, and the double JLNELN with the input neurons not only connected to the hidden neurons but also to the output neurons has the ability to model both the nonlinear and linear components in the prediction residuals. Actual run data was collected to verify the validity of the devised hybrid model. Experiment results have shown that the hybrid model achieved much higher regression precision than a single Elman RNN, which shows the necessity and validity of the double JLNELN model in the prediction of the Elman residuals. The experiment results of the double JLNELN method also show higher regression precision than both a double extreme learning machine method and a single JLNELN method, which verified the validity of the JLNELN method and the double structure of the prediction model.
Conference Paper
This paper proposes a new objective function and quantile regression (QR) algorithm for load forecasting (LF). In LF, the positive forecasting errors often have different economic impact from the negative forecasting errors. Considering this difference, a new objective function is proposed to put different prices on the positive and negative forecasting errors. QR is used to find the optimal solution of the proposed objective function. Using normalized net energy load of New England network, the proposed method is compared with a time series method, the artificial neural network method, and the support vector machine method. The simulation results show that the proposed method is more effective in reducing the economic cost of the LF errors than the other three methods. Index Terms—Economic objective function, load forecast, power system planning, quantile regression, weighted objective function.
Conference Paper
Demand side management (DSM) is an essential function to schedule and optimize the energy consumption in smart grids. One of the challenges of DSM is how to take the complex human behavior into consideration. In this paper, we apply a special population dynamic model to investigate the performance of DSM algorithms under human behavior effects. Specifically, we adapt the SISa (susceptible-infected-susceptible with autonomous infection) model to describe the population behavior of smart grid customers. The SISa model and the DSM algorithms interact with each other via the recovery probability and the DSM population size. The convergence and the equilibrium of the composite model are studied both analytically and numerically.