ChapterPDF Available

An Efficient CNN and KNN Data Analytics for Electricity Load Forecasting in the Smart Grid


Abstract and Figures

The significant part of smart grid is to make smart grid cost-efficient by predicting electricity price and load. To improve the prediction performance we proposed an Efficient Convolutional Neural Network (ECNN) and Efficient K-nearest Neighbour (EKNN) in which the parameters are tuned. It may be difficult to deal with huge amount of load data that is coming from the electricity market. To overcome this issue, we incorporated three modules in the proposed methodology. The proposed model consists of feature engineering and classification. Feature engineering is a two-step process (feature selection and feature extraction); for the purpose of feature selection Mutual Information (MI) is used which reduces the redundancy among features and for feature extraction Recursive Feature Elimination (RFE) is used to extract the principle features from the selected features and reduces the dimensionality of features. Finally, after training the data-set and the removal of the duplicate features load prediction is done by ECNN and EKNN. The ECNN and EKNN outperforms better then traditional Convolutional Neural Network (CNN) and K-nearest Neighbour (KNN). The forecast performance is evaluated by comparing the results with MAPE, RMSE, MAE and MSE. i.e. 10.8, 7.5, 7.15, and 10.4 respectively.
Content may be subject to copyright.
An Efficient CNN and KNN Data
Analytics for Electricity Load Forecasting
in the Smart Grid
Syeda Aimal1, Nadeem Javaid1(B
), Tahir Islam1, Wazir Zada Khan2,
Mohammed Y. Aalsalem2, and Hassan Sajjad1
1COMSATS University, Islamabad 44000, Pakistan
2Farasan Networking Research Laboratory,
Department of Computer Science and Information System, Jazan University,
Jazan 82822-6694, Saudi Arabia
Abstract. The significant part of smart grid is to make smart grid
cost-efficient by predicting electricity price and load. To improve the
prediction performance we proposed an Efficient Convolutional Neural
Network (ECNN) and Efficient K-nearest Neighbour (EKNN) in which
the parameters are tuned. It may be difficult to deal with huge amount
of load data that is coming from the electricity market. To overcome
this issue, we incorporated three modules in the proposed methodology.
The proposed model consists of feature engineering and classification.
Feature engineering is a two-step process (feature selection and feature
extraction); for the purpose of feature selection Mutual Information (MI)
is used which reduces the redundancy among features and for feature
extraction Recursive Feature Elimination (RFE) is used to extract the
principle features from the selected features and reduces the dimension-
ality of features. Finally, after training the data-set and the removal of
the duplicate features load prediction is done by ECNN and EKNN. The
ECNN and EKNN outperforms better then traditional Convolutional
Neural Network (CNN) and K-nearest Neighbour (KNN). The forecast
performance is evaluated by comparing the results with MAPE, RMSE,
MAE and MSE. i.e. 10.8, 7.5, 7.15, and 10.4 respectively.
1 Introduction
Nowadays data analytic plays significant role in every field. Smart grids intro-
duced an advance informational infrastructure that allows optimization in energy
production, transmission, distribution and storage. Smart grid collects data from
different resources and stores it using data analytic techniques. In Smart grid,
bi-directional communication generates data having huge volume, variety and
velocity. Advance data analytic techniques are needed to figure out the required
information from the available raw data. Machine learning techniques are one
Springer Nature Switzerland AG 2019
L. Barolli et al. (Eds.): WAINA 2019, AISC 927, pp. 592–603, 2019.
Electricity Load Forecasting 593
of the reliable methods for predicting load and price. Supervised learning algo-
rithms are the part of machine learning which are used for predictions. Cate-
gories of machine learning algorithms are classification and regression. KNN and
CNN are the part of supervised learning algorithm. Deep learning is a learning
method for multiple levels of abstraction and representation which contains an
input layer, hidden layer, and an output layer. Deep learning comes under an
umbrella of machine learning which is based on learning levels of representation,
corresponding to a hierarchy of features, factors or concepts, where the concept
of higher-level is derived from lower-level and vice versa.
The recent advancements in sensor based technologies hold the promise of
novelty in energy consumption models that can better characterize the underly-
ing patterns. Demand-side Management (DSM) is the main and essential com-
ponent towards energy efficiency and management. DSM in smart homes and
residential buildings is a key component towards sustainability and efficiency in
urban environment.
Smart grid is an electrical grid which consists of range of energy and opera-
tional measures. The main objective of grid is to scale back electricity peak load
and to attain the balance between power supply and demand. Customers are
encouraged to partake within the operations of smart grid, the cost of electric
load will be reduced by the preservation of electricity and by shifting the load
towards off-peak hours.
1.1 Motivation or Problem Statement
In paper [1], the authors forecasted electricity price with deep learning
approaches and empirical comparison of traditional algorithms. Similarly, in
[3], the authors tried to forecast price. However, the load is neglected. Authors
also worked on peak load reduction. However, the accuracy of load and price is
1.2 Contribution
In this paper, We have proposed new classification techniques (i.e. ECNN and
EKNN). The terms forecasting and prediction are used alternatively.
Load is predicted using the new classifiers which beats the traditional tech-
niques in terms of accuracy. Prediction is done on NYISO Data-set. We incorpo-
rated three modules in the proposed methodology. The proposed model consists
of feature engineering and classification. Feature engineering is a two-step pro-
cess (feature selection and feature extraction); for the purpose of feature selec-
tion Mutual Information (MI) is used which reduces the redundancy among fea-
tures and for feature extraction Recursive Feature Elimination (RFE) is used to
extract the principle features from the selected features and reduces the dimen-
sionality of features. Finally, after training the data-set and the removal of the
duplicate features load prediction is done by ECNN and EKNN. The ECNN and
EKNN outperforms better then traditional Convolutional Neural Network and
K-nearest Neighbour.
594 S. Aimal et al.
The main contributions are summarized as follow:
The accuracy must be improved.
Reduce the computational time.
Less resources are utilized.
Enhance the classifier to increase the prediction accuracy.
The rest of this paper is organized in the following way:
In Sect. 2, the related work in context of forecasting load and price is
reviewed. In Sect. 3, the system model is explained. In Sect. 4, the simulations
are performed in python environment on intel core i5, 8 GB RAM, and 1 TB
hard disk. In Sect. 5, the result of this paper is concluded.
2 Related Work
In this section, we have discussed different proposed schemes and techniques for
price and load forecasting [1]. The accuracy in forecasting helps the energy sup-
pliers to optimize their strategies for gaining profit. Likewise, to gain a successful
trade in electricity market different techniques have been proposed by different
authors. Short-term Price Forecasting (STPF) and Short-term Load Forecasting
(STLF) play principle role in the management of energy and resource trading
in the grids [2]. Similarly, Auto-regressive Integrated Moving Average (ARIMA)
performs good in stable electricity market.
In the electricity market, load and price forecasting are essential for optimal
operation planning; most of the load and price forecasting methodologies suffer
from lack of efficient feature selection techniques. To overcome this issue, the
authors in [3] proposed a new hybrid filter-wrapper approach that selects the
most relevant features according to the costumer input in a coordinated manner.
In [4], to construct a forecasting model authors presented a hybrid algorithm
which concatenate Empirical Mode Decomposition (EMD), Similar Days (SD)
Selection, and Long Short-term Memory (LSTM) neural network. Moreover,
to evaluate the accuracy of model between the forecasting and historical data
extreme gradient boosting-based weighted k-means algorithm is adopted.
The authors also deal with two types of forecasting online and day-ahead
hourly forecasting [5]. Stacked Denoising Auto-encoder (SDA) models are com-
pared with data-driven approaches in online forecasting whereas in day-ahead
forecasting, the accuracy and effectiveness is further validated. Mathematical
and statistical models like ARIMA, SARIMA etc., belongs to the classical type
of classifiers whereas artificially intelligent models are CNN, LSTM [6] (Table 1).
The authors in [7], proposed an Improved Elman Neural Network (IENN)
based forecast engine to forecast the load. The weights for this prediction model
carried out with an intelligent algorithm to find better forecasting results.
The efficiency of this model is brought-out to the real world engineering in
the comparison with other forecasting models [8]. In [9], the authors focused
on coordinated management of renewable and traditional energy in which the
Electricity Load Forecasting 595
Table 1. Summary of related work
Technique Objective Data Set Limitations
RBM, Relu Load PJM Overfitting problem
SNN, DNN Price AEMO Cannot deal with
RNN, LSTM Load IRISH Time complexity
CNN, LSTM Price International
Overfitting problem,
Computationally expensive
C4.5, ID3 Price prediction,
C4.5 beats ID3 in
decision tree
Redundant feature not
Price EPEX Cannot deal with extreme
non linearity
MLR, NN Load ISONE Not suitable for long term
MI, ANN Load PJM Time complexity
Load and price EPEX Over fitting
DRN Load and
ISONE Over fitting
Load and price NSW Hard to tune parameters
SVM, Sperm
Load and price NYISO Slow convergence
STLF, STPF Load and weather PJM, NYISO Computationally expensive
Load and price NYISO, PJM Time complexity
DNN Load and price EPEX Space complexity
DNN, LSTM Price and load
NYISO Redundancy features not
authors formulate the problem as a three-stage Stackel-berg game in which the
authors employ backward induction and nash equilibrium method.
The authors also proposed unsupervised data clustering and frequent pattern
mining analysis on energy time series [10], and Bayesian network [10] prediction
for energy usage. The authors also perform experiments using real-world smart
meter data-sets. Accuracy in prediction is a considerable issue with multiple
data [11] which are applicable for a specified time frame. For each day ahead
coming-up with, the operations and the facility from the wind resources can be
foreseen. Observations and a series of models that span the temporal and spacial
596 S. Aimal et al.
scales of the matter; correct prediction could be a massive knowledge downside
that needs disparate knowledge. In [12], multiple models that are applicable for
a particular time-frame, and application of artificial intelligence techniques [13]
are studied.
In paper [14], the authors propose a new prediction methodology that makes
an attempt to generalize the quality of ARMAX statistical model to pair of
metric space. The structure of this planned model is a statistical regression where
parameters operate variables. The variables are lagged values of the series. i.e.
past determined innovations (like moving average terms), or exogenous variables.
In paper [15], a prediction strategy is planned for electricity markets. The
strategy uses large amount of data with hourly information; which is further
used as an input for two separate prediction models. Moreover, a rolling horizon
framework is planned to supply correct updates for the predictions.
The model in paper [16] is intended to find the incidence of abnormal events
and conditions. A generative model for anomaly detection is taken into consider-
ation and the hierarchical data-set of the network is collected from smart meters.
We tend to additionally address three challenges existing in smart grid:
1. large-scale variable count measurements
2. missing points
3. variable choice
In [17], the value prediction is entirely viewed as a symbol process. Data
integrity in [18], the attacks are expected to damage the performance of predic-
tion systems, which can have a significant impact on power systems and therefore
the resilience of power grids.
It reveals the result of knowledge integrity attacks on the accuracy of four repre-
sentative load forecasting models (multiple regression towards the mean, support
vector regression, artificial neural networks, and fuzzy interaction regression).
In paper [19], the authors tend to propose a probabilistic data-driven model
for consumption prediction in residential buildings. The model relies on theorem
network framework that is ready to find dependency relations between conduc-
tive variables which are able to relax the assumptions that are usually created
in ancient prediction models.
In the proposed approach, we enquire the problem of electricity load prediction.
In our prediction model, we tunned the hyper-parameters of KNN. In addition
to this, we also enhanced CNN in which we used two MAX-pooling and three
convolutional layers. Dense and flatten are used as a fully connected layers.
Historical load data that is taken from grid (NYISO) is used as a training data-
set for EKNN and ECNN; we used 80% of data for training purpose and rest of
20% is used for testing (Fig. 1).
Electricity Load Forecasting 597
Fig. 1. Prediction model
The proposed approach depicts that model is made up of three parts:
1. Feature selection
2. Feature extraction
3. Classification
Feature engineering which is two-step process (selection and extraction), and
then classification. In our prediction model, we first took the real data-set of
one year (i.e. january 2016–2017) from NYISO. The data is hourly-base which
is further utilize for the purpose of training and testing.
Design Goal: The goal of this model is to predict the load accurately and
efficiently. For this purpose, the preprocessed data is used on which feature
engineering process and classification is done.
To overcome these constraints:
The data sampling must be done for the reduction of consumption time.
Data formatting must be done carefully.
Redundant data must be removed.
3.1 The Following Metrics Are the Most Essential Metrics
for Accessing the Performance of the Model
Proposed model that is applied for the purpose of predicting load must work
Accuracy of classification can be influenced directly by the performance of
feature engineering.
Forecasted accuracy is highly dependent on the goal of model.
598 S. Aimal et al.
Algorithm 1. Prediction Algorithm
Input : Load Data
/* Separating target from features */
X : features of data
Y : target data
xtrain, x test, y train, y test = split(x,y);
combine imp = MI-imp + RFE-imp;
MI imp = importance calculate by MI;
selected feature = RFE(5, x train, y train);
Combine imp of MI threshold And RFE is true, then
Reserve feature;
Drop feature;
Prediction using EKNN and ECNN with tuned parameters;
Compare predictions with y test
3.1.1 Feature Selector
In the feature selection, the importance of features can be generated. In our
model we first set a certain threshold value i.e. 0.5 for the selection of important
features. In our model, MI is used for feature selection.
Feature selection is used for the following top reasons:
It allows us to reduce over-fitting.
The accuracy of a model can be improved.
The complexity of a model can be improved.
It increases the efficiency of machine learning algorithm to train faster.
3.1.2 Feature Extractor
When it comes to feature extraction, it is different from feature selection. After
the process of feature selection, feature extraction is done which down-sample
the most principle features. In our case, we extracted five principle features by
using RFE.
It is used to further improve the quality, speed, and effectiveness of supervised
It is a reduction process.
it is used for the transformation of attributes.
3.1.3 Classifier
In Classification, we discover a new set of observations on the basis of a trained
set of data. For load prediction, we used ECNN and EKNN as a classifiers.
Electricity Load Forecasting 599
Algorithm 2. Algorithm of KNN
Training data is loaded
Choose from the feature F
for Each point in data: do
Find Euclidean distance
Euclidean distance is stored in sorted list
The first k point is choosen
Assign a class on the basis of majority of classes
end forR(f-i + 1) - f*
Algorithm 3. Hyperparameter Optimization Algorithm
Start procedure
Choose from the feature-set F
for Trai n N e two r k do
end forReturnBestHyperparameters( )
4 Simulation and Reasoning
In this section, we are going to explain the following:
Evaluation Performance.
Each of the points mentioned above are describe in detail below.
4.1 Simulation Setup
To simulate our proposed model, we performed the simulation in python envi-
ronment. The simulation is performed on system with specification intel core i5,
8 GB RAM, and 1 TB hard disk. One year hourly electricity load data of NYISO
is used for the evaluation of proposed scheme.
4.2 Simulation Results
Features are selected with MI.
Principle Components are extracted with RFE.
Actual load is compared with ECNN and EKNN.
The performance is compared with four performance evaluators i.e. Mean
Average Error (MAE), Mean Square Error (MSE), Root Mean Square Error
(RMSE), and Mean Absolute Percentage Error (MAPE).
600 S. Aimal et al.
Fig. 2. Importance feature selection with MI
Fig. 3. Feature extraction with RFE
Electricity Load Forecasting 601
Fig. 4. Load predicted by ECNN and EKNN
MI Technique is applied to select the importance of features with respect to
the target on the basis of a certain threshold value (i.e. 0.5) (Fig.2). Features
which have little impact on load are dropped from data as shown in Fig. 4.The
most five principle feature are extracted to reduce the features dimensionally
through RFE as shown in Fig. 3. ECNN and EKNN as a classifier outperforms
best at the forecasting accuracy of electric load as shown in Fig. 4. The accuracy
that is achieved by ECNN is 92% and the accuracy that is achieved by EKNN
is 93%. The ECNN and EKNN is close to the actual load which is good result.
4.3 Performance Evaluation
In order to investigate the robustness and accuracy of the proposed model; a
comparison is conducted as shown in Fig. 5. For performance evaluation, four
means are used. i.e. MAE, MSE, RMSE, and MAPE. MAPE whose error values
are i.e. 10.847888888888887, 7.566138821530217, 7.155807083333333, and 10.424
MAP E =1
|∗100 (1)
MSE =1
n=1 |(FvAv)|
602 S. Aimal et al.
Fig. 5. ECNN performance evaluation metrics
5 Conclusions
In this paper, the problem of electricity load forecast is taken into consideration.
The data is obtained from grid (NYISO data-set) for one year (i.e. january 2016
to 2017). We took required data from the grid which is hourly-base data that
is further utilize for the purpose of training and testing. The process of feature
engineering is done in which the important features are selected using MI. After
the selection of important features, RFE is used for extracting five principle
The trained data-set is further used for load prediction that is done by our
proposed classifiers (ECNN accuracy is 92% and for EKNN accuracy is 93%).
The predicted electricity load outperforms better then other classifiers in terms
of accuracy. The forecast performance is evaluated by comparing the results with
MAPE, RMSE, MAE and MSE. i.e. 10.8,7.5, 7.15, and 10.4 respectively.
1. Lago, J., De Ridder, F., De Schutter, B.: Forecasting spot electricity prices: deep
learning approaches and empirical comparison of traditional algorithms. Appl.
Energy 221, 386–405 (2018)
2. Liu, B., Nowotarski, J., Hong, T., Weron, R.: Probabilistic load forecasting via
quantile regression averaging on sister forecasts. IEEE Trans. Smart Grid 8(2),
730–737 (2017)
3. Wang, K., Xu, C., Zhang, Y., Guo, S., Zomaya, A.: Robust big data analytics for
electricity price forecasting in the smart grid. IEEE Trans. Big Data (2017)
4. Zheng, H., Yuan, J., Chen, L.: Short-term load forecasting using EMD-LSTM neu-
ral networks with a Xgboost algorithm for feature importance evaluation. Energies
10(8), 1168 (2017)
5. Ahmad, A., Javaid, N., Alrajeh, N., Khan, Z.A., Qasim, U., Khan, A.: A modified
feature selection and artificial neural network-based day-ahead load forecasting
model for a smart grid. Appl. Sci. 5(4), 1756–1772 (2015)
Electricity Load Forecasting 603
6. Ahmad, A., Javaid, N., Guizani, M., Alrajeh, N., Khan, Z.A.: An accurate and
fast converging short-term load forecasting model for industrial applications in a
smart grid. IEEE Trans. Ind. Inform. 13(5), 2587–2596 (2017)
7. Vrablecov´a, P., Ezzeddine, A.B., Rozinajov´a, V., ˇ
arik, S., Sangaiah, A.K.: Smart
grid load forecasting using online support vector regression. Comput. Electr. Eng.
65, 102–117 (2018)
8. Liu, Y., Wang, W., Ghadimi, N.: Electricity load forecasting by an improved fore-
cast engine for building level consumers. Energy 139, 18–30 (2017)
9. Javaid, N., Hafeez, G., Iqbal, S., Alrajeh, N., Alabed, M.S., Guizani, M.: Energy
efficient integration of renewable energy sources in the smart grid for demand side
management. IEEE Access 6, 77077 (2018)
10. Singh, S., Yassine, A.: Big data mining of energy time series for behavioral analytics
and energy consumption forecasting. Energies 11(2), 452 (2018)
11. Abedinia, O., Amjady, N., Zareipour, H.: A new feature selection technique for load
and price forecast of electrical power systems. IEEE Trans. Power Syst. 32(1), 62–
74 (2017)
12. Javaid, N., Javaid, S., Abdul, W., Ahmed, I., Almogren, A., Alamri, A., Niaz,
I.A.: A hybrid genetic wind driven heuristic optimization algorithm for demand
side management in smart grid. Energies 10(3), 319 (2017)
13. Yu, C.N., Mirowski, P., Ho, T.K.: A sparse coding approach to household electricity
demand forecasting in smart grids. IEEE Trans. Smart Grid 8(2), 738–748 (2017)
14. Gonz´alez, J.P., San Roque, A.M., P´erez, E.A.: Forecasting functional time series
with a new Hilbertian ARMAX model: application to electricity price forecasting.
IEEE Trans. Power Syst. 33(1), 545–556 (2018)
15. Chitsaz, H., Zamani-Dehkordi, P., Zareipour, H., Parikh, P.P.: Electricity price
forecasting for operational scheduling of behind-the-meter storage systems. IEEE
Trans. Smart Grid 9(6), 6612–6622 (2018)
16. Moghaddass, R., Wang, J.: A hierarchical framework for smart grid anomaly detec-
tion using large-scale smart meter data. IEEE Trans. Smart Grid 9(6), 5820–5830
17. Singh, N., Mohanty, S.R., Shukla, R.D.: Short term electricity price forecast based
on environmentally adapted generalized neuron. Energy 125, 127–139 (2017)
18. Luo, J., Hong, T., Fang, S.C.: Benchmarking robustness of load forecasting models
under data integrity attacks. Int. J. Forecast. 34(1), 89–104 (2018)
19. Bassamzadeh, N., Ghanem, R.: Multiscale stochastic prediction of electricity
demand in smart grids using Bayesian networks. Appl. Energy 193, 369–380 (2017)
... On the other hand, CNN (Convolutional Neural networks) outperforms SVM as will be shown in our proposed model. Syeda et al. [22] published their recent study about predicting power price and load using smart grid. Their research uses feature engineering to overcome the problem of load parameters and in order to use Efficient Convolutional Neural Network (ECNN) and Efficient K-nearest Neighbour (EKNN) for prediction. ...
Conference Paper
An assessment of wind speed and prediction of output power from a wind machine is important before implementing either commercial power plants or a standalone project. There are many studies targeting the improvement of wind power predictions and have achieved a desirable performance. This paper was conducted with the aim of obtaining accurate wind power production and leveraging the probability of reducing dependency on fossil fuels in the Kingdom of Saudi Arabia (KSA). The data of wind speed was collected from the Weather Online service (WO Service) to do the assessment for the city of Yanbu in KSA for two years. We have used four machine learning techniques in our proposed research to produce four models. Convolutional Neural Network (CNN), Multi-Layer Perceptron (MLP), Super Vector Machine (SVM), and Gaussian Naive Bayes (GNB) were used within these models to evaluate the hourly wind speed and the wind turbine output power. The results of the four trained models show the estimated output power of 2 Megawatt (MW) from the wind. The CNN model showed the best performance since the mean absolute error was 0:011, the test accuracy was 99:9886% and the predicted power was 7431:5 MW while the actual power is 7; 432:28 MW. Index Terms—Wind Energy, Power Prediction, Machine Learning, Convolutional Neural Network, Support Vector Machine, Multi-Layer Perceptron, Gaussian Na¨ıve Bayer
... Various mathematical method, considered to classify the data [19]. To identify matches data concerning objects in the clusters [20], [21]. In this method, the data function is shown by regular alteration between zero and one. ...
This chapter addresses the status of artificial intelligence (AI) as a central element in smart grid (SG) while focusing on the recent progress of research on machine learning techniques to pave the future work in the SG area. The SG framework provides a performance toolkit based on information and communication technologies. The SG paradigm supervises and promotes grid operations at a high level of expertise. AI systems have attributed to the realization of sustainable development goals including the vital integration of renewable energy sources (RES). The scarcity of conventional energy resources in the near future and their increasing threats to the environment extremely require the transition toward RES. The bulk penetration of RES into the electrical grid leads to unstable and volatile power generation. The chapter presents the commonly applied AI methods to the SG system and the key elements of the evaluation procedure.
Data mining can be used to predict the disease for the medical database to authenticate hidden strategies. It normally extracts the useful knowledge domain from medical datasets. Various classifi- cation and feature selection methods are available to predict diseases. This paper examines on five different diseases which include the BCD, PID, HS and CID databases. It normally considered the existing related work to search out important knowledge domains in this field and explained different methods such as ELM, CNN, SVM, PSOA and KNN used in disease prediction. The main discus- sion about the classification methods is used to improve the accuracy rate, recall, and precision rate. Data mining methods are very efficient for the job of classification. The current classification methods are used for classifying the diseases (Medical Databases). Various classification tech- niques have experimented on different databases of the UCI machine learning repository site for managing medical database classification. The simulation result depicts that various methods can attain better performance and compare to the results of other classifiers.
With the continuous progress of modern urbanization, urban noise pollution is also increasing. Noise pollution has been disturbing the normal life of people, and serious noise pollution may even affect people’s health. For this reason, it is necessary to make an effective noise prediction. Noise prediction uses historical data from noise monitoring points to predict future noise values, helping to provide effective noise regulation. In this chapter, the RF, BFGS, and GRU models are used to conduct feasibility studies for noise prediction. These three models are used to predict the public noise, traffic noise, and neighborhood noise. Through comprehensive comparative analysis of the experimental results, it can be concluded that the noise prediction performance of the BFGS model is the best in this experiment. Neighborhood noise is the most predictable among the three types of noise data.
Full-text available
In this paper, a novel modeling framework for forecasting electricity prices is proposed. While many predictive models have been already proposed to perform this task, the area of deep learning algorithms remains yet unexplored. To fill this scientific gap, we propose four different deep learning models for predicting electricity prices and we show how they lead to improvements in predictive accuracy. In addition, we also consider that, despite the large number of proposed methods for predicting electricity prices, an extensive benchmark is still missing. To tackle that, we compare and analyze the accuracy of 27 common approaches for electricity price forecasting. Based on the benchmark results, we show how the proposed deep learning models outperform the state-of-the-art methods and obtain results that are statistically significant. Finally, using the same results, we also show that: (i) machine learning methods yield, in general, a better accuracy than statistical models; (ii) moving average terms do not improve the predictive accuracy; (iii) hybrid models do not outperform their simpler counterparts.
Full-text available
Responsible, efficient and environmentally aware energy consumption behavior is becoming a necessity for the reliable modern electricity grid. In this paper, we present an intelligent data mining model to analyze, forecast and visualize energy time series to uncover various temporal energy consumption patterns. These patterns define the appliance usage in terms of association with time such as hour of the day, period of the day, weekday, week, month and season of the year as well as appliance-appliance associations in a household, which are key factors to infer and analyze the impact of consumers’ energy consumption behavior and energy forecasting trend. This is challenging since it is not trivial to determine the multiple relationships among different appliances usage from concurrent streams of data. Also, it is difficult to derive accurate relationships between interval-based events where multiple appliance usages persist for some duration. To overcome these challenges, we propose unsupervised data clustering and frequent pattern mining analysis on energy time series, and Bayesian network prediction for energy usage forecasting. We perform extensive experiments using real-world context-rich smart meter datasets. The accuracy results of identifying appliance usage patterns using the proposed model outperformed Support Vector Machine (SVM) and Multi-Layer Perceptron (MLP) at each stage while attaining a combined accuracy of 81.82%, 85.90%, 89.58% for 25%, 50% and 75% of the training data size respectively. Moreover, we achieved energy consumption forecast accuracies of 81.89% for short-term (hourly) and 75.88%, 79.23%, 74.74%, and 72.81% for the long-term; i.e., day, week, month, and season respectively.
Full-text available
Accurate load forecasting is an important issue for the reliable and efficient operation of a power system. This study presents a hybrid algorithm that combines similar days (SD) selection, empirical mode decomposition (EMD), and long short-term memory (LSTM) neural networks to construct a prediction model (i.e., SD-EMD-LSTM) for short-term load forecasting. The extreme gradient boosting-based weighted k-means algorithm is used to evaluate the similarity between the forecasting and historical days. The EMD method is employed to decompose the SD load to several intrinsic mode functions (IMFs) and residual. Separated LSTM neural networks were also employed to forecast each IMF and residual. Lastly, the forecasting values from each LSTM model were reconstructed. Numerical testing demonstrates that the SD-EMD-LSTM method can accurately forecast the electric load.
Full-text available
Electricity price forecast plays a key role in strategic behavior of participants in competitive electricity markets. With the growth of behind-the-meter energy storage, price forecasting becomes important in energy management and control of such small-scale storage systems. In this paper, a forecasting strategy is proposed for real-time electricity markets using publicly available market data. The proposed strategy uses high-resolution data along with hourly data as inputs of two separate forecasting models with different forecast horizons. Moreover, an intra-hour rolling horizon framework is proposed to provide accurate updates on price predictions. The proposed forecasting strategy has the capability to detect price spikes and capture severe price variations. The real data from Ontario’s electricity market is used to evaluate the performance of the proposed forecasting strategy from the statistical point of view. The generated price forecasts are also applied to an optimization platform for operation scheduling of a battery energy storage system within a grid-connected micro-grid in Ontario to show the value of the proposed strategy from an economic perspective.
With the emergence of smart grid (SG), the consumers have the opportunity to integrate renewable energy sources (RESs) and take part in demand side management (DSM). In this paper, we introduce generic home energy management control system (HEMCS) model having energy management control unit (EMCU) to efficiently schedule household load and integrate RESs. The EMCU is based on genetic algorithm (GA), binary particle swarm optimization (BPSO), wind driven optimization (WDO) and our proposed genetic WDO (GWDO) algorithm to schedule appliances of single and multiple homes. For energy pricing, real time pricing (RTP) and inclined block rate (IBR) are combined, because in case of only RTP there is a possibility of building peaks during off peak hours that may damage the entire power system. Moreover, to control demand under the capacity of electricity grid station feasible region is defined and problem is formulated using multiple knapsack (MK). Energy efficient integration of RESs in SG is a challenge due to time varying and intermittent nature of RESs. In this paper, two techniques are proposed: (i) energy storage system (ESS) that smooth out variations in renewable energy generation and (ii) trading of the surplus generation among neighboring consumers. The simulation results show that proposed scheme can avoid voltage rise problem in areas with high penetration of renewable energy and also reduce the electricity cost and peak to average ratio (PAR) of aggregated load.
As the internet's footprint continues to expand, cybersecurity is becoming a major concern for both governments and the private sector. One such cybersecurity issue relates to data integrity attacks. This paper focuses on the power industry, where the forecasting processes rely heavily on the quality of the data. Data integrity attacks are expected to harm the performances of forecasting systems, which will have a major impact on both the financial bottom line of power companies and the resilience of power grids. This paper reveals the effect of data integrity attacks on the accuracy of four representative load forecasting models (multiple linear regression, support vector regression, artificial neural networks, and fuzzy interaction regression). We begin by simulating some data integrity attacks through the random injection of some multipliers that follow a normal or uniform distribution into the load series. Then, the four aforementioned load forecasting models are used to generate one-year-ahead ex post point forecasts in order to provide a comparison of their forecast errors. The results show that the support vector regression model is most robust, followed closely by the multiple linear regression model, while the fuzzy interaction regression model is the least robust of the four. Nevertheless, all four models fail to provide satisfying forecasts when the scale of the data integrity attacks becomes large. This presents a serious challenge to both load forecasters and the broader forecasting community: the generation of accurate forecasts under data integrity attacks. We construct our case study using the publicly-available data from Global Energy Forecasting Competition 2012. At the end, we also offer an overview of potential research topics for future studies.
Smart grid, an integral part of a smart city, provides new opportunities for efficient energy management, possibly leading to big cost savings and a great contribution to the environment. Grid innovations and liberalization of the electricity market have significantly changed the character of data analysis in power engineering. Online processing of large amounts of data continuously generated by the smart grid can deliver timely and precise power load forecasts – an important input for interactions on the market where the energy can be contracted even minutes ahead of its consumption to minimize the grid imbalances. We demonstrate the suitability of online support vector regression (SVR) method to short term power load forecasting and thoroughly explore its pros and cons. We present a comparison of ten state-of-the-art forecasting methods in terms of accuracy on public Irish CER dataset. Online SVR achieved accuracy of complex tree-based ensemble methods and advanced online methods.
Electricity price forecasting is a significant part of smart grid because it makes smart grid cost efficient. Nevertheless, existing methods for price forecasting may be difficult to handle with huge price data in the grid, since the redundancy from feature selection cannot be averted and an integrated infrastructure is also lacked for coordinating the procedures in electricity price forecasting. To solve such a problem, a novel electricity price forecasting model is developed. Specifically, three modules are integrated in the proposed model. First, by merging of Random Forest (RF) and Relief-F algorithm, we propose a hybrid feature selector based on Grey Correlation Analysis (GCA) to eliminate the feature redundancy. Second, an integration of Kernel function and Principle Component Analysis (KPCA) is used in feature extraction process to realize the dimensionality reduction. Finally, to forecast price classification, we put forward a differential evolution (DE) based Support Vector Machine (SVM) classifier. Our proposed electricity price forecasting model is realized via these three parts. Numerical results show that our proposal has superior performance than other methods.
A functional time series is the realization of a stochastic process where each observation is a continuous function defined on a finite interval. These processes are commonly found in electricity markets and are gaining more importance as more market data become available and markets head toward continuous-time marginal pricing approaches. Forecasting these time series requires models that operate with continuous functions. This paper proposes a new functional forecasting method that attempts to generalize the standard seasonal ARMAX time series model to the L <sup xmlns:mml="" xmlns:xlink="">2</sup> Hilbert space. The structure of the proposed model is a linear regression where functional parameters operate on functional variables. The variables can be lagged values of the series (autoregressive terms), past observed innovations (moving average terms), or exogenous variables. In this approach, the functional parameters used are integral operators whose kernels are modeled as linear combinations of sigmoid functions. The parameters of each sigmoid are optimized using a Quasi-Newton algorithm that minimizes the sum of squared errors. This novel approach allows us to estimate the moving average terms in functional time series models. The new model is tested by forecasting the daily price profile of the Spanish and German electricity markets and it is compared to other functional reference models.