ChapterPDF Available

Prediction of Building Energy Consumption Using Enhance Convolutional Neural Network

Authors:
Chapter

Prediction of Building Energy Consumption Using Enhance Convolutional Neural Network

Abstract and Figures

Electricity load forecasting plays a vital role in improving the usage of energy through customers to make decisions efficiently. The accuracy of load prediction is a challenging task because of randomness and noise disturbance. An extreme deep learning model is applied in proposed system model to achieve better load prediction accuracy. The proposed model used to extract features by combining the mutual information (RF) and recursive feature elimination (RFE). Furthermore, extreme learning machine (ELM) and enhance CNN are used for load forecasting based on extracted features from MI and RFE. Additionally, to check the performance of our proposed scheme, we compared it with some benchmark schemes e.g. CNN, SVR and MLR. Simulation results reveal that our proposed approach outperformed in prediction performance.
Content may be subject to copyright.
Prediction of Building Energy
Consumption Using Enhance
Convolutional Neural Network
Hafiz Muhammad Faisal1, Nadeem Javaid1(B
), Bakhtawar Sarfraz2,
Abdul Baqi2, Muhammad Bilal2, Inzamam Haider2,
and Sahibzada Muhammad Shuja1
1Comsats University Islamabad, Islamabad 44000, Pakistan
nadeemjavaidqau@gmail.com
2NCBA&E, Multan, Pakistan
http://www.njavaid.com
Abstract. Electricity load forecasting plays a vital role in improving
the usage of energy through customers to make decisions efficiently. The
accuracy of load prediction is a challenging task because of random-
ness and noise disturbance. An extreme deep learning model is applied
in proposed system model to achieve better load prediction accuracy.
The proposed model used to extract features by combining the mutual
information (RF) and recursive feature elimination (RFE). Furthermore,
extreme learning machine (ELM) and enhance CNN are used for load
forecasting based on extracted features from MI and RFE. Addition-
ally, to check the performance of our proposed scheme, we compared it
with some benchmark schemes e.g. CNN, SVR and MLR. Simulation
results reveal that our proposed approach outperformed in prediction
performance.
Keywords: Load forecasting ·Deep learning model ·
Mutual information and recursive feature elimination
1 Background
Rate of energy consumption is directly proportional with an increment in pop-
ulation. In energy consumption, an office building or building in a corporate
organization reserves a significant amount of energy [1,2]. In China, 2011, statis-
tical estimation in a report identifies that 28% of the total energy consumption is
account by the corporate buildings, which is going to reach 35% in 2020 [3]. Same
conditions are to be face in United States with energy consumption close to 39%.
Therefore, industry needs new innovative solutions in order to manage the energy
utilization. In this regard, construction of prediction models for building energy
consumption can play an important role. It can prove helpful in decision making
c
Springer Nature Switzerland AG 2019
L. Barolli et al. (Eds.): WAINA 2019, AISC 927, pp. 1157–1168, 2019.
https://doi.org/10.1007/978-3-030-15035-8_111
1158 H. M. Faisal et al.
Table 1. List of acronyms
ANN Articialneuralnetwork
ARIMA Auto regressive integrated moving average
CART Classification and regression technique
CNN Convolutional neural network
DNN Deep neural network
KPCA Kernal principal component analysis
LSTM Long short term memory
MLP Multi layer perceptron
RFE Recursive feature elimination
RNN Recurrent neural network
ReLU Rectified linear unit
SG Smart grid
SVM Suport vector machine
MI Mutual Information
ELM Extreme learning machine
CDA conditional demand analysis
DL Deep learning
AI Artificial intelligence
MSPE Mean square percentage error
RMSE Root mean square error
MAPE Mean average percentage error
for building management to control the equipment in conducive way. A lot of
work has been done in data science. In the existing work, most commonly case
studies are dependent on the time series data of historical energy consumption
for prediction models construction [5]. Normally, energy consumption prediction
methods are divided into statistical and artificial intelligence (AI) methods. Sta-
tistical models for probabilistic models construction, utilize historical data future
energy utilization are estimated and analyzed on its basis. Author’s [6]proposed
principle component analysis for significant inputs selection in energy consump-
tion prediction model. Author’s [7] applies linear regression for electricity utiliza-
tion estimation while two comparative approaches; neural networks and fuzzy
modeling are utilized for linear regression performance evaluation. ARX extra
inputs in [8] was utilized in autoregressive model for building component param-
eters estimation. Similarly, Kimbara et al. in [9] constructs an autoregressive
integrated moving area (ARIMA) model for online building energy utilization
prediction construction. On the other side, ARIMAX model which is the ARIMA
representation with external inputs, is applied for power demand prediction of
the building. Author’s [11] proposed a method based on regression named as
Prediction of Building Energy Consumption 1159
conditional demand analysis (CDA) building energy consumption prediction. AI
and deep learning (DL) based models have the capability of prediction and the
empirical results obtained from these models are accurate more than applications
in real world. Therefore, these AI and DL-based models are being apply widely
in prediction for energy consumption for a building. Authors in research article
[12] proposed an innovative technique called clusterwise regression which results
in clustering and regression integration for foresees building energy consump-
tion. Author’s [13] proposed a clustering method. This method finds similarity
patters sequences which predicts electricity prices. Similarly in [14], k-means
methodology is proffered which analyzes electricity consumption patterns in a
building. On the other side in [15], a survey is done on managing electricity
related time series foreseeing using data mining techniques. To understand the
energy consumption patterns and levels, decision tree method is used in [16]and
[17] propose random forest algorithm to improve energy efficiency which helps
facility managers. Author’s [18] proposed a relevant data selection methodol-
ogy for consumption of energy prediction. As Artificial neural network system
(ANN’s) are also entrenched in prediction applications so [19] uses short-term
(ST) prediction model for electricity demands in a bioclimatic building. Author’s
[20] proposed two algorithms; Levenberg-Marquardt and OWO-Newton’s algo-
rithm for residential building energy consumption foreseeing (Table 1).
The remaining part of the paper is organized as follows. In Sect. 2, problem
statement is presented. In Sect. 3, system model is discussed in detail. The exper-
iments on the prediction of building energy consumption have been performed
in Sect. 5. Finally, the conclusions of the paper are shown in Sect. 6.
2 Problem Statement
In this paper, we focus the problem of electricity load forecasting. The aim of our
model is to predict the exact load forecast with maximum amount of data from
smart grid. To tackle load forecasting problem, four classifiers are applied to
predict the electricity load. CNN is a classifier that splits the data into training
and testing phase. CNN is efficient classifier, however, following problems are
need to answered for good accuracy of electricity load forecasting.
Hard to tune parameters
High computational complexity
3SystemModel
Our proposed system model consist of four parts: normalizing, training testing
and prediction. Four classifiers with tuned parameters as shown in Fig. 1.
1160 H. M. Faisal et al.
Fig. 1. System model
3.1 Preprocessing Data
System load data is collected for years 2015 to 2017 from ISO-NECA. Data is
divided on monthly basis and each years similar month data is grouped together
e.g. January 2015, January 2016 and January 2017 data is grouped together.
First three weeks of any month are used for the training process and the last
week is for testing. Data is normalized with maximum impact values and divided
into three major parts: training, testing and validation. To investigate the per-
formance of the four classifiers, we used three performance evaluation errors:
mean square percentage error (MSPE), root mean square error (RMSE) and the
mean absolute percentage error (MAPE) which can be calculated respectively as
MSPE =1
T
TM
tm=1
|(AvFv)|(1)
RM SE =
1
T
TM
tm=1
(AvFv)2(2)
MAPE =1
T
TM
tm=1
100
Av
Fv
(3)
where Av is the observed test value at time tm and Fv is the forecasted value
at time tm.
4 Proposed Scheme
In this section, we will discuss RFE and enhance CNN classifiers.
Prediction of Building Energy Consumption 1161
4.1 RFE
RFE method is used for feature selection. Feature selection method fits a model
and discards the weakest feature until the specified number of features is reached.
Algorithm 1. RFE Algorithm
1: Start
2: Train the model on the training set
3: for Each subset size A,i= 1 to A do do) do
4: Keep the record of A most important variables
5: data preprocessing
6: Determine the model performance
7: end for
8: End
4.2 Enhance CNN
In the domain of machine learning, CNN is a class of deep neural network. It
consists of more than one convolution layer and connected one by one. CNN has
one input layer, multiple hidden layers and one output layer. Hidden layers have
Convolutional layer, dense layer, flatten layer and max-pooling layer. Convolu-
tional layer achieves Convolutional operation and than input values passes to
the next layer. Feed forward neural network is used to train the network. max-
pooling is one of the pooling layer. The max pooling layer uses the maximum
number of neurons.
5 Simulation and Reasoning
In this section, we are going to explain the following;
1. Simulation setup
2. Simulation results
3. Performance evaluation
Each of the item mentioned above are discuss in detail below:
5.1 Simulation Setup
To present our proposed model, we performed the simulation in python envi-
ronment. The simulation is performed on system with specification Core i7, 6th
generation, 640 GB hard disk and 8 GB RAM. Five years electricity load data
(ISO New England Control Area (ISO NECA) from 2010 to 2015) is used for the
evaluation of proposed scheme. Data is split into training and testing, in which
training and testing days are 822 and 274. The simulation results are organized
as follows
Mutual Information (MI) and Recursive Feature Elimination (RFE) are used
for features selection
1162 H. M. Faisal et al.
Redundant features are discarded also with Recursive Feature Elimination
In deep learning model, the autocorrelation function is used to calculate the
input values foe ELM
Extreme Learning Machine (ELM) is applied as a predictor to obtain correct
prediction results
ELM performance is measured with three evaluation metrics i.e. Mean Abso-
lute Percentage Error (MAPE), Mean Square Percentage Error (MSPE), Root
Mean Square Error (RMSE)
The proposed scheme is compared with existing models, such as the Conven-
tional Neural Network (CNN), Recurrent Neural Network (RNN) Multiple
Linear Regression (MLR) and Support Vector Regression (SVR).
5.2 Simulation Results
MI and RFE methods are applied to select the main features with respect to
target feature from electricity price data and electricity load. The features, which
have very low effect on price and load are discarded from data. In order to extract
the main feature MI is used as shown in Fig. 1. Figures 2and 3demonstrate that
the partial autocorrelation scheme is used, to find the input variables of the
extreme deep learning model (Figs. 4,5,6,7,8,9and 10).
Fig. 2. Feature selection
Prediction of Building Energy Consumption 1163
Fig. 3. Autocorrelation function
Fig. 4. Partial autocorrelation function
1164 H. M. Faisal et al.
Fig. 5. Error rate
Fig. 6. CNN
Fig. 7. Enhance CNN
Prediction of Building Energy Consumption 1165
Fig. 8. Multi layer regression
Fig. 9. NN
Fig. 10. Support vector regression
1166 H. M. Faisal et al.
5.3 Performance Evaluation
For performance evaluation, four techniques are used i.e. MAE, MAPE, MSE,
and RMSE. RMSE has the lowest error values according to all mentioned classi-
fiers and MSE has the highest error value. In order to determined the accuracy
of the proposed model the comparison of performance evaluation is conducted
as shown in Fig. 11. Figure 11 shows that CNN has the highest error values so
proposed enhance CNN has low error values.
Fig. 11. Performance evaluation
6 Conclusion
Proposed deep learning shown its great learning prediction qualities. The
proposed model consists of two key parts; feature selection and classifiers. A
combination of two different techniques are used for feature selection. The pro-
posed model shows the best load prediction results in building sector. Pro-
posed model provided a novel prediction model for building energy consumption.
Comparison results indicate that the enhance CNN method performances better
than the existing machine learning methods.
References
1. Streimikiene, S.: Residential energy consumption trends, main drivers and policies
in Lithuania. Renew. Sustain. Energy Rev. 35, 285–293 (2014)
2. Ugursal, V.I.: Energy consumption, associated questions and some answers. Appl.
Energy 130, 783–792 (2014)
3. Hua, C., Lee, W.L., Wang, X.: Energy assessment of office buildings in China using
China building energy codes and LEED 2.2. Energy Build. 86, 514–524 (2015)
Prediction of Building Energy Consumption 1167
4. Zuo, J., Zhao, Z.Y.: Green building research-current status and future agenda: a
review. Renew. Sustain. Energy Rev. 30, 271–281 (2014)
5. Daut, M.A.M., Hassan, M.Y., Abdullah, H., Rahman, H.A., Abdullah, M.P.,
Hussin, F.: Building electrical energy consumption forecasting analysis using con-
ventional and artificial intelligence methods: a review. Renew. Sustain. Energy
Rev. 70, 1108–1118 (2017)
6. Li, K., Hu, C., Liu, G., Xue, W.: Building’s electricity consumption prediction using
optimized artificial neural networks and principal component analysis. Energy
Build. 108, 106–113 (2015)
7. Pombeiro, H., Santos, R., Carreira, P., Silva, C., Sousa, J.M.C.: Comparative
assessment of low-complexity models to predict electricity consumption in an insti-
tutional building: linear regression vs. fuzzy modeling vs. neural networks. Energy
Build. 146, 141–151 (2017)
8. Jimenez, M.J., Heras, M.R.: Application of multi-output ARX models for estima-
tion of the U and g values of building components in outdoor testing. Sol. Energy
79, 302–310 (2005)
9. Kimbara, A., Kurosu, S., Endo, R., Kamimura, K., Matsuba, T., Yamada, A.: On-
line prediction for load profile of an air-conditioning system. Ashrae Trans. 101,
198–207 (1995)
10. Newsham, G.R., Birt, B.J.: Building-level occupancy data to improve ARIMA-
based electricity use forecasts. In: Proceedings of the 2nd ACM Workshop on
Embedded Sensing Systems for Energy-Efficiency in Building, Zurich, Switzerland,
pp. 13–18, 2 November 2010
11. Aydinalp-Koksal, M., Ugursal, V.I.: Comparison of neural network, conditional
demand analysis, and engineering approaches for modeling end-use energy con-
sumption in the residential sector. Appl. Energy 85, 271–296 (2008)
12. Hsu, D.: Comparison of integrated clustering methods for accurate and stable
prediction of building energy consumption data. Appl. Energy 160, 153–163 (2015)
13. Alvarez, F.M., Troncoso, A., Riquelme, J.C., Ruiz, J.S.A.: Energy time series fore-
casting based on pattern sequence similarity. IEEE Trans. Knowl. Data Eng. 23,
1230–1243 (2011)
14. P´erez-Chac´on, R., Talavera-Llames, R.L., Martinez-Alvarez, F., Troncoso, A.:
Finding electric energy consumption patterns in big time series data. In: Proceed-
ings of the 13th International Conference Distributed Computing and Artificial
Intelligence, Sevilla, Spain, 1–3 June 2016, pp. 231–238. Springer, Cham (2016)
15. Mart´ınez-´
Alvarez, F., Troncoso, A., Asencio-Cort´es, G., Riquelme, J.C.: A survey
on data mining techniques applied to electricity-related time series forecasting.
Energies 8, 13162–13193 (2015)
16. Tso, G.K.F., Yau, K.K.W.: Predicting electricity energy consumption: a compari-
son of regression analysis, decision tree and neural networks. Energy 32, 1761–1768
(2007)
17. Ahmad, M.W., Mourshed, M., Rezgui, Y.: Trees vs. neurons: comparison between
random forest and ANN for high-resolution prediction of building energy consump-
tion. Energy Build. 147, 77–89 (2017)
18. Paudel, S., Elmitri, M., Couturier, S., Nguyen, P.H., Kamphuis, R., Lacarri`are, B.,
Corre, O.L.: A relevant data selection method for energy consumption prediction of
low energy building based on support vector machine. Energy Build. 138, 240–256
(2017)
19. Mena, R., Rodr´ıguez, F., Castilla, M., Arahal, M.R.: A prediction model based
on neural networks for the energy consumption of a bioclimatic building. Energy
Build. 82, 142–155 (2014)
1168 H. M. Faisal et al.
20. Biswas, M.A.R., Robinson, M.D., Fumo, N.: Prediction of residential building
energy consumption: a neural network approach. Energy 117, 84–92 (2016)
21. Ahmad, A., Javaid, N., Mateen, A., Awais, M., Khan, Z.A.: Short-term load fore-
casting in smart grids: an intelligent modular approach. Energies 12(1), 164 (2019).
https://doi.org/10.3390/en12010164. ISSN: 1996-1073
22. Zahid, M., Ahmed, F., Javaid, N., Abbasi, R.A., Kazmi, S.Z., Javaid, A., Bilal,
M., Akbar, M., Ilahi, M.: Electricity price and load forecasting using enhanced
convolutional neural network and enhanced support vector regression in smart
grids. Electronics (2019). EISSN 2079-9292. (IF = 2.110, Q2)
... Decision trees: Classification tree [41,199], Regression tree [9,33,124,199] Boosted tree [33,41,124] Easy to interpret and nonparametric ...
... Decision trees: Classification tree [41,199], Regression tree [9,33,124,199] Boosted tree [33,41,124] Easy to interpret and nonparametric ...
Article
The rapid depletion of natural sources of energy, coupled with increasing global population has triggered the emergence of various techniques and strategies for building energy consumption prediction. According to information from existing body of knowledge, this paper systematically brings to fore the application areas of building energy consumption prediction (i.e. well-established and emerging), the relationships between these areas and the ways in which authors integrate the current spate of techniques. Based on direct implications of buildings on global energy consumption and CO2 emissions, this information makes it possible to identify trends, strengths and limitations in this context, thereby enabling the centralisation of activities required for future studies. This study follows several well-documented guides for conducting logical reviews of primary articles concerning main topics of building energy consumption prediction within popular online databases. The definition of articles’ search keywords as well as inclusion/exemption factors were governed by a combination of principles stipulated by Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and Procedures for Performing Systematic Reviews (PPSR). In comparison to existing review articles in the studied field, the current study is novel in the sense that it provides a very holistic view to building energy consumption prediction, thereby minimising the need to consult multiple individualised studies that are limited to specific techniques, data sets, regions or types of buildings. Another unique feature of this study is its interrelationship network of articles which depicts a quick glance at some of the most influential studies as well as underrepresented areas, thereby aiding research planning, future directions and cross-disciplinary collaborations.
... In general, existing energy demand and consumption prediction approaches can be mainly classified into either intrusive or non-intrusive, with respect to their reliance on a distributed sensing network or a single sensing point, respectively (Ridi et al., 2014). The former utilize an extensive sensor network to predict demand using numerous indicative signals, such as occupancy, indoor temperature and/or sub-metering demand values (Edwards et al., 2012;Ekici & Aksoy, 2009;Faisal et al., 2019;Shabani & Zavalani, 2017;Zhao & Magoulès, 2012). Since the additional information can significantly enhance their efficiency, most work on energy consumption prediction proposed in the literature is generally intrusive (Amasyali & El-Gohary, 2018). ...
Article
Reliable, non-intrusive, short-term (of up to 12 h ahead) prediction of a building's energy demand is a critical component of intelligent energy management applications. A number of such approaches have been proposed over time, utilizing various statistical and, more recently, machine learning techniques, such as decision trees, neural networks and support vector machines. Importantly, all of these works barely outperform simple seasonal auto-regressive integrated moving average models, while their complexity is significantly higher. In this work, we propose a novel low-complexity non-intrusive approach that improves the predictive accuracy of the state-of-the-art by up to 10%. The backbone of our approach is a K-nearest neighbours search method, that exploits the demand pattern of the most similar historical days, and incorporates appropriate time-series pre-processing and easing. In the context of this work, we evaluate our approach against state-of-the-art methods and provide insights on their performance.
Article
The reliable and an economic operation of the power system rely on an accurate prediction of short term load. In this paper, a deep learning based Long Short Term Memory (LSTM) with hybrid feature selection namely RMR-HFS-LSTM, is proposed. The objective of this study is to reduce the curse of dimensionality, reduce the overfitting and improve the accuracy of short term load forecasting. The RMR-HFS is a combination of filter and wrapper feature selection introduced for identifying optimal subset of features. The instance based RReliefF and information theoretic based mutual information filter feature selection are utilized to reduce curse of dimensionality by finding and eliminating irrelevant features. The selected features of filter feature selection is tuned by using Recursive Feature Elimination (RFE) wrapper feature selection to reduce overfitting. The deep learning based LSTM improves the accuracy by handling uncertainty issues. The experiment was conducted on European weather and electricity load data using python on Tensorflow environment. The performance of the proposed RMR-HFS-LSTM model is compared against Multilayer Perceptron (MLP) and Recurrent Neural Network (RNN) in terms of Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE). The result shows that the proposed RMR-HFS-LSTM model outperforms other models.
Chapter
Cascading failure is the most important reason of blackout in power systems. Since the unscheduled and sudden outage of a line is one of the main reasons of cascading failure, evaluating CFOP and predicting the blackout severity resulting from line outage is of great importance. Therefore, an intelligent three-step scheme based on the predictor DT and DOVs with rich information regarding the line status is presented. An intelligent three-step scheme based on three predictive DT and DOVs is presented for evaluating the CFOP and predicting the blackout severity caused by the cascading failure. In the proposed scheme, DOVs are identified for each line by using the entropy-based method along with the theory of mutual information between operating variables and offline simulation. DOVs are used as input vectors for training/testing of the three CSPDTs. The proposed three-step prediction scheme can evaluate CFOP and predict blackout severity resulting from line outage using DOVs and DTs online for three blackout levels. In a real-time environment, in the first step, CFOP is evaluated using identified DOVs for each line and trained CSPDTs; if there exists CFOP, the blackout severity is predicted for three levels. The effectiveness of the proposed scheme is examined using a standard IEEE39-bus test system comprising multiple transmission lines, and desired results with high accuracy are obtained.
Article
Full-text available
Short-Term Electricity Load Forecasting (STELF) through Data Analytics (DA) is an emerging and active research area. Forecasting about electricity load and price provides future trends and patterns of consumption. There is a loss in generation and use of electricity. So, multiple strategies are used to solve the aforementioned problems. Day-ahead electricity price and load forecasting are beneficial for both suppliers and consumers. In this paper, Deep Learning (DL) and data mining techniques are used for electricity load and price forecasting. XG-Boost (XGB), Decision Tree (DT), Recursive Feature Elimination (RFE) and Random Forest (RF) are used for feature selection and feature extraction. Enhanced Convolutional Neural Network (ECNN) and Enhanced Support Vector Regression (ESVR) are used as classifiers. Grid Search (GS) is used for tuning of the parameters of classifiers to increase their performance. The risk of over-fitting is mitigated by adding multiple layers in ECNN. Finally, the proposed models are compared with different benchmark schemes for stability analysis. The performance metrics MSE, RMSE, MAE, and MAPE are used to evaluate the performance of the proposed models. The experimental results show that the proposed models outperformed other benchmark schemes. ECNN performed well with threshold 0.08 for load forecasting. While ESVR performed better with threshold value 0.15 for price forecasting. ECNN achieved almost 2% better accuracy than CNN. Furthermore, ESVR achieved almost 1% better accuracy than the existing scheme (SVR).
Article
Full-text available
Daily operations and planning in a smart grid require a day-ahead load forecasting of its customers. The accuracy of day-ahead load-forecasting models has a significant impact on many decisions such as scheduling of fuel purchases, system security assessment, economic scheduling of generating capacity, and planning for energy transactions. However, day-ahead load forecasting is a challenging task due to its dependence on external factors such as meteorological and exogenous variables. Furthermore, the existing day-ahead load-forecasting models enhance forecast accuracy by paying the cost of increased execution time. Aiming at improving the forecast accuracy while not paying the increased executions time cost, a hybrid artificial neural network-based day-ahead load-forecasting model for smart grids is proposed in this paper. The proposed forecasting model comprises three modules: (i) a pre-processing module; (ii) a forecast module; and (iii) an optimization module. In the first module, correlated lagged load data along with influential meteorological and exogenous variables are fed as inputs to a feature selection technique which removes irrelevant and/or redundant samples from the inputs. In the second module, a sigmoid function (activation) and a multivariate auto regressive algorithm (training) in the artificial neural network are used. The third module uses a heuristics-based optimization technique to minimize the forecast error. In the third module, our modified version of an enhanced differential evolution algorithm is used. The proposed method is validated via simulations where it is tested on the datasets of DAYTOWN (Ohio, USA) and EKPC (Kentucky, USA). In comparison to two existing day-ahead load-forecasting models, results show improved performance of the proposed model in terms of accuracy, execution time, and scalability.
Article
Full-text available
Energy prediction models are used in buildings as a performance evaluation engine in advanced control and optimisation, and in making informed decisions by facility managers and utilities for enhanced energy efficiency. Simplified and data-driven models are often the preferred option where pertinent information for detailed simulation are not available and where fast responses are required. We compared the performance of the widely-used feed-forward back-propagation artificial neural network (ANN) with random forest (RF), an ensemble-based method gaining popularity in prediction—for predicting the hourly HVAC energy consumption of a hotel in Madrid, Spain. Incorporating social parameters such as the numbers of guests marginally increased prediction accuracy in both cases. Overall, ANN performed marginally better than RF with root-mean-square error (RMSE) of 4.97 and 6.10 respectively. However, the ease of tuning and modelling with categorical variables offers ensemble-based algorithms an advantage for dealing with multi-dimensional complex data, typical in buildings. RF performs internal cross-validation (i.e. using out-of-bag samples) and only has a few tuning parameters. Both models have comparable predictive power and nearly equally applicable in building energy applications.
Conference Paper
Full-text available
In recent years the available volume of information has grown considerably due to the development of new technologies such as the sensor networks or smart meters, and therefore, new algorithms able to deal with big data are necessary. In this work the distributed version of the k-means algorithm in the Apache Spark framework is proposed in order to find patterns from a big time series. Results corresponding to the electricity consumptions for years 2011, 2012 and 2013 for two buildings from a public university are presented and discussed. Finally, the performance of the proposed methodology in relation to the computational time is compared with that of Weka as benchmarking.
Article
Full-text available
Data mining has become an essential tool during the last decade to analyze large sets of data. The variety of techniques it includes and the successful results obtained in many application fields, make this family of approaches powerful and widely used. In particular, this work explores the application of these techniques to time series forecasting. Although classical statistical-based methods provides reasonably good results, the result of the application of data mining outperforms those of classical ones. Hence, this work faces two main challenges: (i) to provide a compact mathematical formulation of the mainly used techniques; (ii) to review the latest works of time series forecasting and, as case study, those related to electricity price and demand markets.
Article
Developing energy efficiency strategies for Intelligent Buildings requires a continuous monitoring of the impacts of energy efficiency measures against baseline scenarios. Since not all consumption events can be linearly described, non-linear models based on fuzzy systems and neural networks are successful where linear models either become inaccurate or too complex to be employed in practice. Herein, we study the adequacy of plain-designed non-linear models to predict energy consumption in a real-world Intelligent Building setting by using simple predictor variables such as time-of-day, weather conditions, and occupancy as estimated from WiFi traffic. The developed Fuzzy Systems and Neural Network models achieve considerably better performance and accuracy indexes than linear regression models, with VAFs equal to 79.3% and 75.7% respectively, and MAE of 5.9 and 7.1 kWh, respectively, against 36.6% VAF and 12.8 kWh MAE for linear regression models, thus demonstrating that low complexity non-linear models can be used to accurately describe energy consumption baselines in Intelligent Buildings.
Article
Low energy buildings (LEBs) are being considered as a promising solution for the built environment to satisfy high-energy efficiency standards. The technology is based on lowering the overall heat transmission coefficient value (U-value) of the buildings envelope and increasing a heat capacity thus creating a higher thermal inertia. However, LEB introduces a large time constant compared to conventional building due to which it slows the rate of heat transfer between interior of building and outdoor environment and alters the indoor climate regardless of sudden changes in climatic conditions. Therefore, it is challenging to estimate and predict thermal energy demand for such LEBs. This work focuses on artificial intelligence (AI) model to predict energy consumption of LEB. Two kinds of AI modeling approaches: “all data” and “relevant data” are considered. The “all data” uses all available training data and “relevant data” uses a small representative day dataset and addresses the complexity of building non-linear dynamics by introducing past day climatic impacts behavior. This extraction is based on dynamic time warping pattern recognition methods. The case study consists of a French residential LEB. The numerical results showed that “relevant data” modeling approach that relies on small representative data selection has higher accuracy (R² = 0.98; RMSE = 3.4) than “all data” modeling approach (R² = 0.93; RMSE = 7.1) to predict heating energy load.
Article
It is important for building owners and operators to manage the electrical energy consumption of their buildings. As electrical energy is the major form of energy consumed in a commercial building, the ability to forecast electrical energy consumption in a building will bring great benefits to the building owners and operators. This paper provides a review of the building electrical energy consumption forecasting methods which include the conventional and artificial intelligence (AI) methods. The significant goal of this study is to review, recognize, and analyse the performance of both methods for forecasting of electrical energy consumption. Compared to using a single method of forecasting, the hybrid of two forecasting methods can possibly be applied for more precise results. Regarding this potential, the swarm intelligence (SI) method has been reviewed to be hybridized with AI. Published literature presented in this paper shows that, the hybrid of SVM and SI methods has indeed presented superior performance for forecasting building electrical energy consumption.
Article
Some of the challenges to predict energy utilization has gained recognition in the residential sector due to the significant energy consumption in recent decades. However, the modeling of residential building energy consumption is still underdeveloped for optimal and robust solutions while this research area has become of greater relevance with significant advances in computation and simulation. Such advances include the advent of artificial intelligence research in statistical model development. Artificial neural network has emerged as a key method to address the issue of nonlinearity of building energy data and the robust calculation of large and dynamic data. The development and validation of such models on one of the TxAIRE Research houses has been demonstrated in this paper. The TxAIRE houses have been designed to serve as realistic test facilities for demonstrating new technologies. The input variables used from the house data include number of days, outdoor temperature and solar radiation while the output variables are house and heat pump energy consumption. The models based on Levenberg-Marquardt and OWO-Newton algorithms had promising results of coefficients of determination within 0.87–0.91, which is comparable to prior literature. Further work will be explored to develop a robust model for residential building application.
Article
This paper describes a load profile prediction algorithm that has recently been developed for use in the optimal operation of cool/heat storage systems. It is highly desirable to devise an on-line prediction technique so that the difference between a predicted load and an actual load is as small as possible. To accomplish this, an autoregressive integrated moving average (ARIMA) model is assumed. The modeling is first done for the past load data. Next, the model predicts load profiles for the next day. The load profiles are updated every hour on the basis of the newly obtained load data. Performance of the load profile prediction algorithm, was assessed by recording the cooling load (1987) at an Osaka site. Results show generally good agreement between actual and predicted loads.